Professional Documents
Culture Documents
de Cuernavaca
Powell Draper1; Maria E. Moreyra Garlock2; and David P. Billington, Hon.M.ASCE3
Abstract: The structures of Felix Candela stand as elegant examples of the art of thin shell concrete structural design. However, few thin
shell structures are built today and many of the most enduring structures, such as those of Candela, have not been analyzed. This project
Downloaded from ascelibrary.org by UNIVERSITE LAVAL on 03/04/15. Copyright ASCE. For personal use only; all rights reserved.
performed a finite-element analysis of Candela’s Cuernavaca Chapel in order to better understand its structural efficiency and gain insight
into its design. Analysis results show that Candela’s striking structure has very low stresses resulting from its own self-weight.
DOI: 10.1061/共ASCE兲1076-0431共2008兲14:2共47兲
CE Database subject headings: Finite element method; Thin shell structures; Structural analysis; Hyperbolic parabolic shells.
Fig. 2. Dimensions of Cuernavaca Chapel 共in ms兲; letters represent reference locations in Table 2
The accuracy of the finite-element model of this structure was The force and stress results of the finite-element analysis rein-
first verified through a simplified analysis. The structure was ini- force Candela’s selection of a structurally efficient form. Using
tially modeled with a uniform 4 cm 共1.6 in.兲 thickness. A vertical the uniform shell thickness model, principal shell stresses reach a
section cut was taken from the fixed supports at the base nearest maximum of 4,488 kPa 共651 psi兲 in compression and 1,131 kPa
the taller edge 共Fig. 7兲. The area of the unsupported section 共in 共164 psi兲 in tension. Compressive stress values are well under
front of the supports兲 was determined from the Rhinoceros model what we can assume to be a minimum compressive strength of
of the geometry. Rhinoceros also computed the centroid of this concrete used. Tension is kept low enough so that it can suitably
section. With this information and the thickness of the structure be assumed by the reinforcement in the shell. Maximum defor-
共in the model兲, it is possible to compute the bending moment mation in the model was 0.5 cm 共0.2 in.兲. No significant cracking
created by the self-weight of the unsupported section 共or “over- was observed in the actual structure so we may conclude defor-
hang”兲 at the vertical section cut. mations are kept within reasonable limits.
The area of the shell extending in front of the supports was The highest stresses are found in the structure along the lines
calculated 共from Rhinoceros兲 as 267.7 m2 共2,881.5 ft2兲. The cen- of supports and the back edge of the structure. First hand obser-
troid of this section was calculated at 2.8 m 共9.2 ft兲 from the vation revealed that at these regions, Candela thickened the struc-
section cut. Assuming a concrete density of 2,400 kg/ m3 ture to accommodate the higher stresses 共Figs. 9 and 10兲. The
共150 lb/ ft3兲, the bending moment at the section cut was calcu- model thus serves to provide us with a clearer understanding of
lated to be 707 kN m 共521 K ft兲. The results of the finite-element Candela’s design by demonstrating its efficient distribution of
analysis therefore had to be in agreement with this value. forces and resulting stresses.
The finite-element model of the structure was analyzed under To further investigate the structural response of the Cuerna-
the effect of its own dead load. The results of the analysis of the vaca Chapel, we next altered the finite-element model by first
finite-element model gave joint forces acting at the nodes of each adding the thick edge beam at the shorter lip of the structure and
element. The distribution of joint forces at the section cut is then varying the thickness of the shell. Again, no specific dimen-
Fig. 8. Distribution of joint forces along section cut, plan view 共uniform thickness兲
Conclusion
of Cuernavaca Chapel model and corresponding actual increase in loid minimizes bending in the shell and precludes the need for
thickness ribs or a masking of the thinness. Despite the large size of this
structure, stresses are kept low enough to allow for a remarkable
thinness and free edge.
sions of these details have been found, so approximations were Comparison of three finite-element models of varying com-
made based on measurements observed on the actual structure. A plexity and precision shows that the simplest model, which rep-
comparison of the results may be seen in Tables 1 and 2. The resents the shell with uniform thickness and no edge beams,
model with varying thickness, as the actual structure has, shows a yields conservative stress results. Yet these results indicate that
general reduction of stress from the models with uniform thick- the stresses are well within the strength limits.
ness and a more gradual distribution of stress throughout the The structural analysis also allows for an interpretation of
shell. Candela’s design methodology. The basic form of the hyperbolic
For the model with uniform thickness and an edge beam, prin- paraboloid need not bind the designer into a predetermined ge-
cipal stresses reach a maximum in compression of 4,695 kPa ometry. Within the discipline of the hyperbolic paraboloid, Can-
共681 psi兲 and a maximum of 1,586 kPa 共230 psi兲 in tension. For dela found room for creative design, by maximizing the effect of
the model with varying thickness, principal stresses reach a the free edge with a strikingly tall opening. Candela also under-
maxima of 2,523 kPa 共366 psi兲 in compression and 1,503 kPa stood that the function follows the form, and by stretching the
共218 psi兲 in tension. Thus, in each case, we see that stresses are height he also opened the structure to an increase in dead load and
kept low, verifying the efficiency of Candela’s choice of form. wind load. This he countered by adeptly thickening the edges and
Computational analysis of structures such as the Cuernavaca back of the structure. The result is a structure that adheres to an
Chapel serves to verify what Candela seemed to know at the advantageous geometrical form while allowing the designer to
outset of design: that employed properly, the hyperbolic parabo- manipulate it into a singular work of structural art.
The current dearth of thin shell concrete structures continues
to confound those who are aware of how effectively they provide
an attractive, efficient structural solution. This study has sought to
demonstrate how the medium of thin shell structures was mas-
tered by one structural artist through a thorough investigation of
one of the master works of his oeuvre. Through careful attention
to the design and construction of structurally efficient forms,
Felix Candela was able to demonstrate the strength and durability
of a remarkably thin structure.
Acknowledgments
Table 1. Comparison of Finite-Element Model Results 共Uniform Thickness, Uniform Thickness with Edge Beam, and Variable Thickness with Edge
Beam兲
Uniform Uniform thickness with Varying thickness with
thickness edge beam edge beam
Max displ. 关cm 共in.兲兴 −0.5 共−0.2兲 −0.5 共−0.2兲 −0.3 共−0.1兲
Min principal stress 关kPa 共psi兲兴 −4,492 共−651兲 −4693 共−681兲 −2525 共−366兲
Max principal stress 关kPa 共psi兲兴 1,134 共164兲 1,586 共230兲 1,505 共218兲
Note: Negative values in compression; positive values intension.
References Meyer, C., and Sheer, M. H. 共2005兲. “Do concrete shells deserve another
look?” Concr. Int., 27共10兲, 43–50.
Candela, F. 共1958兲. “Understanding the hyperbolic paraboloid.” Architec- Otero, A. B. 共2001兲. “Félix Candela y el borde libre.” Bitacora Arquitec-
tural Record, 123共7兲, 191–195. tura, 5, 38–47, 71.
Computers and Structures, Inc. 共2004兲. CSI analysis reference manual for Robert McNeel & Associates. 共1993兲. Rhinoceros user’s guide, Seattle.
SAP2000, Berkeley, Calif. Saliklis, E. P., Billington, D. P., and Carmalt, A. W. 共2007兲. “Tedesko’s
Draper, P. 共2006兲. “Felix Candela and the chapel of Lomas de Cuerna- Philadelphia skating club: Refinement of an idea.” J. Archit. Eng.,
vaca.” Master of Science in Engineering thesis, Princeton Univ., 13共2兲, 72–77.
Princeton, N.J.
RESEARCH
Abstract
This paper presents a contemporary architectural working method that encompasses
digitalization and parameterization of existing buildings and optimization of new
buildings designed with ruled surfaces. The method uses parametric modeling
and computational structural analysis in order to simplify contemporary building
processes. As an example of the application of these techniques, in this paper they
are applied to Felix Candela’s Church of St. Joseph the Craftsman, a design which
features hyperbolic paraboloids that are considered difficult to design, calculate and
build. The optimization method introduced in this paper seeks to explore different
possibilities for designing and modifying buildings designed using non-standard
geometry allowing them to be built out of simplified elements but also keep
construction and visual properties of their shape. This method is also useful for
students and young engineers to expand their skills in structural analysis, parametric
modeling and optimization methods with contemporary tools.
* Zlata Tošić
zlata10@live.com
Sonja Krasić
sonjak@gaf.ni.ac.rs
Naomi Ando
n‑ando@hosei.ac.jp
Milos Milić
milos.cicevac@gmail.com
1
Faculty of Civil Engineering and Architecture, Niš, Serbia
2
Faculty of Engineering and Design, Hosei University, Tokyo, Japan
Vol.:(0123456789)
Z. Tošić et al.
Introduction
To talk about architecture without using the word “form” may now seem
inconceivable (Forty 2013). Much of the architecture of the twentieth century is
characterized by its experimentation with the use of geometric shapes in building
design. Reinforced concrete is one of the most used materials for building these
designs. Not only because it has the ability to cover large areas, but also because
it enables physical continuity of the shape. However, in shell structures with non-
standard geometry, the complexity of the geometry, the large continuous surface
area and material use, demands certain phases in the building process that are
costly in order to get the best final results. These include: the cost of wooden
formwork, the time for concrete to stabilize and how to get the right shell shape
(Jovanović et al. 2014).
Topology optimization can be described as a family of computational methods
aimed at finding optimal structural layouts and configurations (Aage et al. 2014).
There are a range of methods for this optimization including solid isotropic
microstructures with penalty (SIMP) methods (Mariano et al. 2009), heuristic
methods of topology design and the homogenization method (Bendsøe and Kikuchi
1988). The method also depends on the type of object. For example, for optimization
of higher buildings there are manufacturing type constraints, in particular pattern
gradation and repetition, in the context of building layout optimization (Stromberg
et al. 2011). For optimal reduction of material the Evolutionary Structural
Optimization (ESO) method for developing conceptual forms of complex
structures can be used also (Xie et al. 2005). Although, it is powerful tool, the
important steps must be chosen in order to have the best overview of the effects
of optimization. Developments in technology have brought new programs which
enable us to optimize, design parametrically and rationalize shapes and elements
of construction, therefore making them more efficient, while still taking care of the
aesthetic component of the final product (Stavrć and Wiltsche 2015). Furthemore,
the structural behavior of shells is developed essentially due to their form. Some
researchers deal with small modifications in the geometry of the form without
modifying their initial aesthetic configuration too much (Cavieres et al. 2011). In
contrast, this paper uses discretization of the shape, for construction simplification,
while keeping the visual effect of the hyperbolic paraboloid (HP).
The aim of this paper is to demonstrate a method for using geometry properties
for optimization and to place constructive elements in the optimal positions. The
method supports appropriate material choice, structural system selection and
the development of additional elements of the façade. The method supports the
examination of these properties of both new and existing construction, including
a consideration of mass, stress in elements, necessary process of building,
architectural properties. It allows for defining analyzed parameters of both models in
the form of percentages for comparison and the analysis of geometrical properties of
elements for potential digital prefabrication (Tomas and Marti 2010a).
The case study used to demonstrate the method is the Church of St. Joseph
the Craftsman in Mexico, designed by architect Felix Candela and built in
Geometry and Construction Optimization: An example using…
1959 (Wortman and Tuncer 2017). The method is undertaken using Rhinoceros
with plug-ins Grasshopper, Karamba, Silvereye and FEMAP with NX Nastran.
Using the method we present the geometry and construction optimization
of the new contemporary variation or model of the case study building. The
main characteristics of this building, which are changed to create the new
contemporary model, are the execution cost and performance process. Although
it will be defined in detail in the next section, the principle of shortened workflow
can be seen in this scheme (Fig. 1). In the workflow the operations are lined up
by order. The first three are dealing with definitions of geometry, parametrization,
construction elements, material, support and loads. Operation 4 is a calculation
of defined construction (using plug-in Karamba) and operation number 5 is
minimizing deformation in function of number “N” (using plug-in Sivereye).
Operation 6 is checking the other geometrical and construction characteristics.
If they are in an acceptable range of numbers (length and number of elements,
mass) and satisfy visual criteria (shape) for the process of prefabrication, then the
number “N” from step 5 is the final solution. Otherwise, the next number “N” is
taken and the operation 6 is repeated. The loop from 6 to 6′ will be repeated until
we satisfy conditions from the operation 6, in which case the number “N” will be
defined in the final step 7.
Fig. 2 Plan and front view of the Church of St. Joseph the Craftsman in Mexico (Source: Radivojević
and Kostić 2011)
quadrangle. The space between two HPs is formed for daylight illumination (Fig. 3).
The other four remaining sides are used as glass walls with thin columns.
Typically, the three observation levels micro, meso and macro are distinguished
in reinforced concrete construction. A new idea is to transfer this principle to shell
construction. It is not absolutely necessary to construct the entire load-bearing
structure on the macro level through one shell, but rather to understand the shell as
part of an overall structure on the meso scale. For example, ceiling structures can be
composed of individual shell elements, which in turn act as a beam (Faber 1965). To
accomplish this it is necessary to think about the elements of the construction and
their connections, which are the main properties. Geometry optimization of the shell
can be done through shell tessellation with many algorithms (Hegger et al. 2018).
Using geometry properties this type of optimization is possible on the double curved
surface of the HP form between two straight lines (generatrices of both systems);
thus, Candela achieved economy of construction by avoiding use of curved panels
for his formwork in construction (Moreyra Garlock and Billington 2010).
To digitize the building’s shape, it is necessary to analyze its dimensions
and geometry first. This building is an example of ruled surface where we can
use its characteristics for optimization of its geometry. In this case, geometry
optimization is defined by using discretization of the geometry by planarization
of the continuous, double curved form of the HP, with triangular panels. The
goal is to keep a similar visual effect of the HP form, achieve good construction
Fig. 3 Rhinoceros drawings with approximate dimensions which demonstrate the geometry of Candela’s
design
Geometry and Construction Optimization: An example using…
properties and at the same time facilitate the process of building with
prefabricated elements.
The whole geometry of the building is then defined in Grasshopper. The idea is
then to modify a new HP surface using its straight generatrices. Parameterization
is defined by number “N” which represents the number of divisions of HP
directrices (guidelines) on equal parts in order to get straight lines (Krasić 2012).
These creates the basis of the grid, for both systems of generatrices are the main
rods in the construction (Fig. 4). The second step is to find the average point
coordinates of every special quadrilateral and connect them with all four points.
These are the second type of rods called diagonals. Triangle modules represent
the planar panels, which are not susceptible to bending during construction (grid)
deformation in order to form stable structural geometry. The third type of rods
which connect two HPs are ties.
Then pre-calculation of the new construction design is necessary, because
even though the geometry optimization represents discretization geometrical
features of the shape in order to facilitate the assembling of building elements,
they are also part of the construction system and all construction properties
have to be satisfied. The pre-calculation process is using Karamba where we
are able to define: cross-section of elements, loads, thickness of the panels,
material and support. For material, steel type S355 is chosen, not only because
of its mechanical properties, but also because of the shape of the elements and
their connections. These loads are taken into account in the decision: mass of
the construction and weight of triangular panels. For determining support three
degrees of freedom (rotation around all three axis) were defined, in marked blue
points (Fig. 4).
We then calculated mass and deformation of the new design and using
Grasshopper all properties of construction are directly related to the number of
division “N”. Changing that value all characteristics (except cross sections) are
simultaneously changing as well. For every selected value “N” the Karamba
calculations are represented in Table 1.
The next step is to define the number “N”, which is directly related to geometrical
properties of construction elements including number and length of elements
(geometrical features) and mass and deformation (construction features). These
are all intended to be minimized for the assembling process and level of stress in
elements. In order to see relations between construction properties and number
“N”, the process of optimization is done in Silvereye. The goal is to find the number
“N” (input parameter) for minimal values of deformation (fitness value) of the
construction system. As mentioned, geometry and construction optimization are
co-dependant so we compare the two most important properties of geometry and
construction optimization (Fig. 5).
From Fig. 5 we can see that the value of deformation varies from N = 1–4 because
of the placement of the supports, length and cross-section of the elements. After
stabilization maximal deformation dramatically decreases when the value of number
N > 4 and then slowly rises after N = 11. Nevertheless, divisions more then 10 aren’t
analyzed further because of the large number of elements and their connections.
If changes in the original shape are made in order to place planar elements, the
lower boundary of the discretization must be defined in order to preserve the fluid
continuity of the geometry as much as possible, as it presents the whole architectural
beauty of the shape. The similarity between curve and polygonal line depends of
scale and individual perception, so it is up to the designer to select the satisfactory
discretization. Authors decided that options 4 and 6 are eliminated as the line from
Fig. 6 Grasshopper drawing of the shape of discretized geometry with planar panels with different “N”
value
the “discretization of the original HP” shape which is achieved with division points
of generatrices (blue full line) does not follow/overlap the “curvature of the original
HP” (yellow dashed line). Therefore, it does not create the same effect and satisfy
the visual criteria (Fig. 6).
In the end, we focus on the number and length of the elements. Even though they
are inversely proportional, it can be noticed that for N = 10, length of the elements
does not decrease much when compared to N = 8, but the number of elements is
drastically higher. The last characteristic is mass, which is increasing with every
number „N”. After overviewing all parameters it can be concluded that chosen
number for division is 8.
Fig. 7 Rhinoceros drawing of the existing Church of St. Joseph the Craftsman with main constructive
elements annotated
effectively optimized and is one of the most advanced buildings, not only in terms of
construction system, but also building technology in that time.
The chosen geometry and material require physical continuity, which makes
the process of building complicated and lenghty. It also requires a lot of precision
especially during placing the formwork (Fig. 8). Structural calculations for the
existing building are done with all the stated elements using FEMAP with NX
Nastran. It is important to define mass and maximum deformation in the system
with reinforced concrete to be able to compare the results with the calculations of
the new designed building.
When analyzing stress distribution we found that maximum intensity is located in
the spikes of the shell as well as connections of shells and main beams, as expected
for this system. Maximum stress is 12.64 MPa as it can be seen in Fig. 9.
Figure 10 shows total translation which is located at the spikes of both shells; the
maximum amount is 0.244 m. The mass of the whole construction of the existing
object is 565.805 t. The structure is analyzed with geometric nonlinear theory and
linear behavior of the concrete. Conditions are formed for the basic load case, which
includes only self-weight of the structure. Deformation on Fig. 10 is enlarged three
times.
Fig. 8 Rhinoceros drawing of
the formwork that was used
for the building process of
the Church of St. Joseph the
Craftsman (adapted from: Smith
et al. 1967)
Geometry and Construction Optimization: An example using…
The authors wanted to analyze results of the calculation for the new model in the
same program as they used for the built Church. All of the inputs (position of
elements, material, load and support) are defined to be the same as in previous
Karamba calculation (Fig. 11).
Points marked “123” in Fig. 11 present supports with three degrees of freedom
(rotation around all three axes). Cross sections are tubular and their diameter is
different for every type of element: main rods (168.3 mm), diagonals (133 mm)
and ties (88.9 mm). Moreover, because of the buckling in the longest ties at the
bottom, they are replaced with simple beams, which will serve only for glass
panel’s position.
After the definition of all elements, connections, materials, loads and supports,
the results of the stress distribution are shown in Fig. 12. Maximum stress is
located on the connections of rods and it is max = 62.72 MPa.
Figure 13 presents the model of total translation, where the location of
the maximum level of deformation is in the same place as it is in the existing
Z. Tošić et al.
building. This value is shown in the contour view and it is max = 0.164 m. Total
mass of the object is 36.039 t. This value is less than stated previously, because
of the reduction of cross sections during the design, but the value of deformation
has not varied a lot.
Geometry and Construction Optimization: An example using…
The comparison of the construction properties of the new and existing or referenced
building is shown in the Fig. 14 according to the previous analysis in the paper.
There are two charts with mass and deformation characteristics presented in
percentages.
The first chart defines the percentage of mass of the new and referenced
construction, where it can be seen that the choice of different material and
construction elements can reduce mass by about 16 times. The second chart presents
analysis of maximum deformations, where the new construction has around 66% of
maximum deformation of the existing construction. This comparison demonstrates
the level of control of the methods used throughout the previous process. Moreover
it shows that wide range of possibilities of different analyses in contemporary
technology can be effective on these shapes.
The use of optimization techniques in the design process of structures widens the
field of use of computers and allows the user to obtain optimum designs for stated
design conditions. There are other researches that are achieving optimization due
to transformation of the entire geometry (dimensions, thickness, curvature), while
preserving aesthetic appearance similar to that initially planned by the designer
(Tomas and Marti 2010a). In future research the combination of these two methods
could be explored.
After analysis of the construction and geometry, optimizing and finding element’s
number and positions, elements are placed in the construction system (Fig. 15).
Optimal design of rods offers, by comparison, a path for closer integration
between optimization and fabrication, and it is therefore arguable that they should
take a more prominent role in the investigations of the field (Aage et al. 2014).
Fig. 15 Scheme of all elements of the new model construction as well as facade expression (Rhinoceros)
Using technology the next step is to digitalize elements for potential fabrication.
This is an indicator of the geometrical properties of the elements.
Firstly, we can define types of rods in the construction. Rods are divided in five
groups regarding their length which is from 1.4 to 6.9 m. Figure 16 demonstrates
how every group of elements is distributed through the model. Table 2 shows
specifications of every group length and the number of elements as well as the
percentage of elements that the group covers in the whole structure. It can be
concluded that most of the elements are from 1.64 to 4.65 m long and only around
6% are from 5.08 to 6.90 m. All of the 5th group rods are ties. The total number
of rod elements is 688. This geometrical configuration is beneficial as the main
rods, which have the biggest percentage of stress, are the shortest and the longest
ties have only tension.
The second phase is prefabrication and division of panel’s area. The groups are
formed depending on the surface area of the panels. This analysis can show the
curvature of the surface, as the zone with the smaller panels has bigger curvature
than the area of the bigger and identical panels. When adding all numbers from
Table 3 the total number of panels is 512.
The most curved part of the surface is the middle of the HP where the first
group is. The difference in the surface areas in every group varies. In the first
group it is 0.19 m2, then it rises to 0.38 m2 in the second, third and fourth group
and finally the fifth group has four identical panels (Fig. 17). Most of the panels
are from the first three groups (90.61%). According to the surface area of the
panels it can be concluded that the distribution during the process of building will
not be complicated, as the panels are from 1.54 to 3.26 m2.
Every panel is attached to three intersection points of the rods. Weight is also
not the issue during assembling, as its amount is around 10 kg/m2. Composite
aluminum panels are fireproof, water resistant, with excellent heat and accoustic
insulation. Some of the aesthetic values are (1) surface flatness and smoothness,
(2) superior weather, corrosion and pollutant resistance, (3) even coating, (4)
super peeling strength, (5) impact resistance, (6) lightweight and easy to process
and maintain. As they can be custom made, and all dimensions are specified, it is
possible to make prefabrication precise (Lee et al. 2018).
Finally, it can be noted that in many topology optimization procedures, certain
regularization is necessary in order to control the physical size of structural
members appearing in the new design (Aage et al. 2014). Although these
dimension values apply only to this example, this process is not only defining
overall properties about construction, but helping in practical digitalization
of the elements as well. It facilitates organization in the whole structure and
gives better input when it comes to construction process. Figure 18 shows the
final result of the new building’s design. Given the change in the construction
system it is possible to cover the spatial grid with lightweight aluminum panels
with openings which can provide more daylight in the Church. There is a lot of
freedom in design options. The planarization of the panels enables placing glass
in place of the panel’s openings.
One deficiency of the building process concerns the precise spatial placement of
the rods and their connections. It is planned to start from the support placement,
and then to be tied with matching rods. It will be necessary to finish the middle
of the object first in the shape of cantilever going to the ends. In future work such
issues have to be thought through in more detail. For example, when it comes to the
production process one of the priorities is to create a safe waterproof shield for the
building, therefore, the shell element’s connections must be designed as continuous.
Conclusions
New developments in technology have spread across all the areas of expertise in
architecture and civil engineering, making new technology and processes available
in the building process. Complex geometric shapes can now be achieved using
prefabricated elements and the visual identity of the final object can also be
preserved at the same time.
The geometry optimization method described in this paper helps in defining
the element’s shape, position and dimensions in order to analyze its construction
properties. Both are co-dependant, as geometric properties have influence on some
of the construction characteristics and vice versa. Together they form elements
which can be precisely analyzed for the process of fabrication. This process is one of
the main benefits of efficient building. Now, not only will the process be quicker and
less expensive, but also less risky when it comes to the process of building.
This paper presents an analysis of a case study building, taking Candela’s built
object as a good example of HP shape design, and offers a contemporary solution
with different materials and changes in the constructive system which can make the
Z. Tošić et al.
building process much easier. Candela’s significant structures were all of HP forms
(Moreyra Garlock and Billington 2010). This research shows methods which can
modify this approach to contemporary designing, optimizing the construction and
geometry in ruled surfaces geometry with strictly defined dimensions. However, it
is not possible to use this method for more free-form and complex geometries as
the starting point. Construction properties, such as mass and deformation, of the
new and referenced HP shape are compared in order to verify the right methods.
In addition it is important to see in which ways are contemporary programs and
technology upgrading the process of modelling, digital fabrication (Tomas and Marti
2010b) and optimization. The process is not only important for existing objects such
as Candela’s Chapel, but also for new design works.
References
Aage, N., O. Amir, A. Clausen, L. Hadar, D. Maier and A. Søndergaard 2014. Advanced topology
optimization methods for conceptual architectural design. In: Advances in Architectural Geometry,
eds. P. Block, J. Knippers, N. Mitra and W. Wang. 159-179. Cham : Springer International
Publishing : Springer.
Bendsøe, M. P., and K. Kikuchi. 1988. Generating optimal topologies in structural design using a
homogenization method. Comput Methods Appl Mech Eng 71 (2): 197-224.
Cavieres, A., R. Gentry and T. Al-Haddad. 2011. Knowledge-based parametric tools for concrete
masonry walls: Conceptual design and preliminary structural analysis. Automation in Construction
20 (6): 716-728.
Faber, C. 1965. Candela and his bowls (Candela und seine schalen). Munich: Verlag Georg D.W.
Callwey.
Forty, A. 2013. Words and buildings: a vocabulary of modern architecture. London : Thames & Hudson.
Hegger J., M. Curbach, A. Stark, S. Wilhelm and K. Farwig. 2018. Innovative design concepts:
Application of textile reinforced concrete to shell structures. Structural Concrete 19 (3): 637-646.
Jovanović N., P. Pejić and S. Krasić. 2014. The use of voronoi diagram in the design of an object
containing a hyperbolic paraboloid, 18th Geometrical Seminar, Vrnjačka Banja, Serbia, Book of
Abstracts. ISBN 987-86-6275-027-3, 55. http://tesla.pmf.ni.ac.rs/people/geometrijskiseminarxviii/
Book%20Abstracts-VrnjackaBanja2014.pdf. Accessed 4 March 2018.
Krasić, S. 2012. Geometrical surfaces in Architecture. Nis, Serbia: Faculty of Civil Engineering and
Architecture, University of Nis.
Lee, A. D., P. Shepherd, M. C. Evernden and D. Metcalfe. 2018. Optimizing the Architectural Layouts
and Technical Specifications of Curtain Walls to Minimize Use of Aluminium. Structures. 13: 8-25.
Mariano, V., M. Pascual and M. Q. Osvaldo. 2009. Topology design of two‐dimensional continuum
structures using isolines. Computers & structures. 87 (1-2): 101-109.
Moreyra Garlock, M. E. and D. P. Billington. 2010. Structural Art and the Example of Félix Candela.
Journal of structural engineering. 136 (4) : 339-342.
Radivojević G., and D. Kostić. 2011. Construction systems in architecture. Nis: Faculty of Civil
Engineering and Architecture, University of Niš, Serbia.
Smith C. B., M. Pani, M. Goeritz, F. Candela, L. Barragán and J. O’Gorman. 1967. Builders in the sun :
five Mexican architects, New York : Architectural Book.
Stavrić M. and A. Wiltsche. 2015. Ornamental plate shell structures. https://www.researchgate.net/publi
cation/268368071_Ornamental_plate_shell_structures. (accessed 5 January 2018).
Stromberg L. L., A. Beghini, W. F. Baker and G. H. Paulino. 2011. Application of layout and topology
optimization using pattern gradation for the conceptual design of buildings. Structural and
Multidisciplinary Optimization. 43 (2) : 165-180.
Tomas A., and P. Marti. 2010. Shape and size optimization of concrete shells. Engineering Structures.
32(6): 1650-1658.
Geometry and Construction Optimization: An example using…
Tomas A., and P. Marti. 2010. Optimality of Candela’s concrete shells: A study of his posthumous design,
Journal of the International Association for Shell and Spatial Structures, 51:.67.
Wortman T. and B. Tuncer 2017. Differentiating parametric design: Digital workflows in contemporary
architecture and construction, In Parametric Design Thinking. Design Studies 52: 173-197.
Xie Y., P. Felicetti, J. Tang and M. Burry. 2005. Form finding for complex structures using evolutionary
structural optimization method. Design Studies 26 (1): 55-72.
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.
Zlata Tošić is PhD Candidate in the Faculty of Civil Engineering and Architecture in Nis, Serbia at
the Department of Architecture. Her field of research implies the geometry, topology and structural
optimization of shell constructions. She is also interested in exploring genetic algorithms and using this
method for optimization of grid shells.
Sonja Krasić is Associate Professor in the Department of Visual Communication, Faculty of Civil
Engineering and Architecture, Niš, Serbia. Dr. Krasić undertakes research in the field of descriptive
geometry and into applications of photogrammetry in architecture.
Naomi Ando is Fulltime Professor, Faculty of Engineering and Design, Hosei University, Tokyo, Japan.
Dr. Ando’s speciatly is Architectural, urban planning and geometry. Various forms of architecture and
cities (e.g., plazas and streets) and the qualities and characteristics of the spaces that those forms produce.
Milos Milić is PhD Candidate in the Faculty of Civil Engineering and Architecture in Nis, Serbia at
the Department of Civil Engineering. His field of research implies steel constructions, materials and
structural engineering. Resent research is regarding clinging timer slabs with steel beams and their
structural analysis.
JOURNAL OF THE INTERNATIONAL ASSOCIATION FOR SHELL AND SPATIAL STRUCTURES: IASS
Editor’s Note: Manuscript submitted 9 September 2005; revisions received 20 September and 17 November 2006;
accepted for publication 18 November 2006. This paper is open for written discussion, which should be submitted to the
IASS Secretariat no later than August 2007.
SUMMARY
Studying landmark works of structural engineering is essential to the advancement of the field and to the
continuing education of the engineer. This report presents a study of such a structure, Felix Candela’s Los
Manantiales shell in Xochimilco, Mexico City. The shell’s structural action is discussed, and then examined
through finite element analysis using SAP software. Analysis results are supplemented by field observations of
the shell’s current condition, and examinations of archival photos of the shell taken during its construction.
Keywords: Felix Candela, concrete shell, hyperbolic paraboloid, free edge, creep, shrinkage
2. CANDELA’S CONCEPT
technical writings, Candela never boasted of his Table 1. Finite element model reactions, 1k = 4.45kN
structures as ‘works of art,’ but in many ways, he
considered himself an artist. He enjoyed integrating Model Reaction (moment reactions ≈ 0) Value (k)
Vertical, V_model 57.21
aesthetics and sound structure, using his imagin- Horizontal, H_model 35.92
ation to find new solutions. He saw the free, thin,
crisp edge as the ultimate shell builder’s challenge,
software was used to tabulate the shell’s surface
and his solution, found after less than a year of
area, as shown in Fig. 3.1. The surface area was
research, is simultaneously an elegant and efficient
then used to calculate the shell’s total dead load.
innovation. Colin Faber describes how, in 1955,
This method deviated from ideal hand-made
Candela began working on the free-edge problem
calculations, but as the software accurately
for the roof of the Mexico City Stock Exchange [5].
produced surface areas for simpler shapes, its use
The project ended with a ribbed shell, but Candela
was decided an acceptable break with intended
continued exploring the problem on his own. His
procedure. The results are presented in Table 2
next large project, the San Antonio de Las Huertas
below.
Church, built in mid-1956, was completed with free
edges.
Prior to a full stress analysis, SAP2000 was first From Fig. 3.2, tanα = 4d/L, where α is the angle
used to calculate the shell’s support reactions. between N and the horizontal, d is the rise, and L is
These are presented in Table 1. Moments about the the span.
supports were negligible.
tanα = 4d/L = (4 x 5.8m)/(32m) = 0.725
A global equilibrium check was conducted against Hhand = V/(tanα) = 57.2/0.725 = 79 k (351 kN)
the vertical reaction, V, to test the finite element
model’s accuracy. First, Rhinoceros 3D drafting Hmodel = 35.92k (160 kN), less than half of Hhand.
VOL. 47 (2006) No. 3 December n. 152
0
The line just behind an edge parabola, the area of
-2 0 20 40 60 the shell most vulnerable to tension cracking, was
Stess in local x direction
Top Surface
-8
-10
Bottom surface support. Compression then decreases steadily to
-12
zero at the apex. S22 compression (Fig. 3.2.2)
-14 begins at zero at the apex, and steadily increases to
-16 a peak of –68 psi (–469 kN/m2) two-thirds down the
Distance from center of shell (ft)
edge. Compression then decreases again, but
returns to a final peak of –118 psi (–813 kN/m2), in
Figure 3.1.1. S11 stresses along groin the V-beam at the support. These stresses are
small, even for the shell edge. However, their rise
and fall suggests that as the shell leans outward past
V-beam S22 Stresses [(+) Tension]
its supports, its cantilevered weight bends the edge
2 slightly inward, as if pulling it into one wide inward
Stress in local y direction
0
curve. Compression rises to a peak two-thirds down
-2 0 20 40 60
-4
the edge, the deepest point of the curve, and then
(k/ft^2)
-6
Top Surface steadily decreases as the edge moves away from
Bottom Surface
-8 that point. At the groin, the full weight of the
-10 saddle bears down on the support to create a final
-12
compression peak. Bending in the thin shell is
-14
Distance from center of shell (ft)
negligible, but the stress pattern itself is notable and
will be discussed later, with respect to evidence of
slight creep in the concrete.
Figure 3.1.2. S22 stresses along groin
The Smin and Smax stresses (Fig. 3.2.3 & Fig.
V-beam Minimum Stresses [(+) Tension]
3.2.4) follow the same trend as those of S11 and
S22, respectively. Of special interest are the Smin
0 stresses, which begin at –122psi (–841 kN/m2) at
0 20 40 60
Minimum Stress in
4
2
Candela’s design successfully eliminated tensile
0 edge stresses. (The full analysis appears in [2].)
(k/ft^2)
Top Surface
-2 0 20 40 60
Bottom Surface
-4 4. PRESENT-DAY LOS MANANTIALES
-6
-8 Figure 4.1 shows Los Manantiales in its present
-10 condition. It is currently under renovation, but
Distance from Center of Shell (ft) many sealed cracks can still be observed on the
shell’s inner and outer surfaces. The most probable
Figure 3.1.4. Smax stresses along groin
causes of these cracks are shrinkage and creep.
VOL. 47 (2006) No. 3 December n. 152
2
Top Surface
0
0 10 20 30 40
Bottom Surface
-2
-4
-6
Vertical distance, from support to apex
along perimeter arch (ft)
Edge S22 Stresses [(+) Tension] Figure 4.1. Present-day Los Manantiales
5
Figure 4.2, a photograph of the shell taken shortly
0
0 10 20 30 40
after concrete had finished being poured, shows
-5 faint, damp cracks on the shell’s surface. Figure
-10
Top Surface 4.3, a photo taken after construction was completed,
Bottom Surface
-15
shows a network of cracks spreading in hard lines
over the shell’s surface. That the cracks became
-20
more prominent from the initial pouring to the
-25 finishing of construction indicates they developed
Vertical distance, from support to
apex along perimeter arch (ft)
as the concrete dried. Los Manantiales was poured
by hand, one section at a time, and as workmen
Figure 3.2.2. S22 stresses (k/ft2) along edge parabola moved over the formwork, finished areas were left
to dry uncovered. The shell’s exterior surfaces thus
Edge Minimum Stresses [(+) Tension]
dried faster than its interior, creating “mud-cracks”
5 as shrinkage pulled the outer surface into tension.
0 Over time, moisture and pollution deepened some
0 10 20 30 40
-5 of the cracks, and today, a number of drip stains can
-10
Top Surface be observed where cracks once broke completely
Bottom Surface
-15
through the shell, as show by the patched leaks in
-20
Fig. 4.4. A number of these cracks are still in the
process of being sealed, but none has done any
-25
Vertical Distance, from support to apex
significant damage to the shell.
along perimeter arch (ft)
A second crack pattern can be observed on the
Figure 3.2.3. Smin stresses (k/ft2) along edge parabola
underside of the shell. In Fig. 4.5, a now-sealed
Edge Maximum Stresses [(+) Tension]
crack propagates in from an edge along the
underside of each saddle. These cracks appear on
8 each side of every edge, about 10ft (~3m) up from
6
each support and symmetric about the saddle’s
4
2 Top Surface
center.
0 Bottom Surface
-2 0 10 20 30 40 While the cracks occur systematically, they are not
-4 the result of form. Their correspondence to the
-6 edge stress trends of the finite element analysis
Vertical Distance, from support to apex
along perimeter arch (ft)
instead suggests they have occurred due to creep.
Xochimilco is hot all year round, threaded with
Figure 3.2.4. Smax stresses (k/ft2) along edge parabola canals, and suffers the air pollution abundant
JOURNAL OF THE INTERNATIONAL ASSOCIATION FOR SHELL AND SPATIAL STRUCTURES: IASS
Figure 4.2. Photograph showing shrinkage cracks during Figure 4.3. Photograph showing more diffuse shrinkage
construction. The cracks are most visible on the left side of cracks after the structure has been completed
the image.
Figure 4.4. Patched leaks and drip-stains on the underside Figure 4.5. Patched creep crack and crack pattern layout
of the shell, and their location in plan
VOL. 47 (2006) No. 3 December n. 152
throughout Mexico City. Heat, moisture, and elegant structure that avoided what he called
pollution all accelerate creep, and for decades, Los “extravagant architectural dreams”[4] that can lead
Manantiales’ small mud cracks increased its to great expense and questionable structural
susceptibility to these three elements. As with any performance. It is critical to the future of structural
concrete structure in permanent compression, Los engineering to recognize the high quality of such
Manantiales most likely experienced some small work, significant for its technical virtuosity and
degree of creep independent of its environment. sensitivity to aesthetics.
This natural creep, however, would then have been
magnified by its unfavorable surroundings to REFERENCES
produce the cracks. It is important to note that the
cracks have in no way compromised the structural [1] Billington, David P., “Felix Candela and
integrity or safety of the shell, and that any Structural Art,” Bulletin of the International
deformations in the shell’s present state are so small Association of Shell Structures, no. 88,
as to be invisible to the eye, even in the cracked January 1986, pp. 5-10.
areas. After 46 years, the concrete is still in good
condition, and with vigilance and occasional [2] Burger, Noah, The Xochimilco Shell, Senior
maintenance, will most likely last through another Thesis, Department of Civil and
46. Environmental Engineering, Princeton
University, 2004.
5. CONCLUSION [3] Candela, Felix, personal letter to Anton
Tedesko, December 5th, 1963, Princeton
Because the finite-element analysis in this report is University Tedesko Archive.
relatively straightforward, it has substantial
educational value, especially when connected to [4] Candela, Felix, New Architecture, Maillart
field experiences and simple calculation to explain Papers, Ed. Billington, D. P., Mark, R., &
structural behavior. However, the analysis Abel, J. F., Dept. of Civil Engineering,
ultimately demonstrates not that such analysis was Princeton University, 1973, pp. 119-126.
necessary for the design, but rather that the design
renders the stresses essentially insignificant. It [5] Faber, Colin, Candela: The Shell Builder,
confirms the correctness of the design while Reinhold Publishing Corporation, 1963.
reinforcing Candela’s insistence on the PHOTO CREDITS
unimportance of such analysis when the designer
chooses a structurally appropriate form. As Figure 1 – Erwin Lang. Photo taken from [5, p.
Candela was fond of saying, “the quality of a 159].
structure is in inverse proportion with the amount of
calculations…”[3] (For further insight into Figure 2.3, 2.4 – Cubiertas Ala. Drawings obtained
Candela’s ideas, see [1] and [4].) from Avery Library Archive of Architectural
Drawings at Columbia University.
Even more important is the first-hand inspection of Figures 4.1, 4.4, 4.5 – Noah Burger
the 46-year-old shell, and its evaluation as still in
fine condition. This is a dazzling accomplishment Figures 4.2, 4.3 - Antonio Candela. Photos
for such a thin shell, and is further testimony to obtained from Avery Library Archive of
Candela’s high merit as a structural artist. With Architectural Drawings at Columbia
Los Manantiales, Candela succeeded in creating an University.
DEVELOPMENT AND PERFORMANCE OF A PRECASTING
SYSTEM, FOR CONCRETE HYPERBOLIC
PARABOLOID SHELL CONSTRUCTION
By
Ronald Tacie Noyes
\\
Bachelor of Science
Oklahoma State University
Stillwater, Oklahoma
1961
t
•i
•
'
Thesis Approved:
570282
ii
ACKNOWLEDGEMENT
iii
TABLE OF CONTENTS
Chapter Page
INTRODUCTION • • • • • • * • • • • • • .. fll l
The Problem • • • • • • • • • • • • • • • • • 2
Objectives • • • • • • • • • • • • • • • • • 4
Limitations • • • • • • • • • .. • • • • • • • 4
IL LITERATURE REVIEW e • • • • • • • • • • • • s • • 6
Introduction • • • • • ,, • • • • • • • • • • 6
Precasting. ~ • • • • • • • • • • • . • • . • 7
Hyperbolic Paraboloid Shell Structures • • • 13
Major Problem Areas in H-P Shell
Development • • • • • • • • • • • • • • • • 17
Results and Conclusions of Testing • • • • • 23
III. THE EXPERIMENTAL STUDY • • • • • • • • • • • . . .. JO
Research Outline • • • • • • • • • ·• • • • &> 31
IV. DESIGN OF STRUCTURAL ELEMENTS .... • • 33
Shell Design • • • • • • • • • •_ • • • • • * 3;
Support System Analysis • • • --· • • • • • • • 4o
Tie Bar Design • • • • • • • • • • • • • • • 53
Footing Design • • • • • ~ • • •- • • • • • • 54
v. DEVELOPMENT OF ASSEMBLY COMPONENTS • • • • • • • • 59
Introduction • • • • • • • • • • • • • • • • 59
Formwork and Precasting • • • • • • • • • • • 59
Curing Precast Elements • • • • • • • • • • • 67
Footing Steel • • • • • • • • • • • • • • • • 68
Assembly Supports • • .. • • • • • • • • • • • 68
Lifting Frame • • • • • • ~ • • • • • -• • • • 76
Column Supports for Foundation Casting • • • 79
iv
Chapter Page
VI. ERECTION PROCEDURE • • 0 • • • 0 • • 0 0 0 0 0 80
Site Layout • • • • • • . • • • • • • 81
Column Erection • • • • • • • • • • • • • 84
Tie Bar Connection~ • • • • • • • • • 87
Support System~ ~ ~ • • • • • • • • • • • 87
Shell Assembly. • • • • • • • • • • • • • 89
Welded Shear Connections. • • • • • • • • 90
VII. TESTING APPARATUS AND PROCEDURES • • 0 0 • 0 • 95
Tie Bar Calibration Tests • • • • • • • • 95
Concrete Test Samples • • • • • • • • • • 96
Structural Testing; • ~ • • • • • , • • • 97
Testing Procedure • • • • • • • • • • • • 102
VIII. PRESENTATION AND ANALYSIS OF DATA. • • • • 0 0 105
Analysis of Construction Costs. • • • 10 5
Analysis of Load Test Data. • • • • • • • 112
IX. DISCUSSION OF RESULTS. • • • • • • • • 0 0 • • 135
Assembly Components and Techniques • • • o 135
Discussion of Cost Analysis • • • • • • • 145
Discus~ion of Load Test Results • • • • • 164
v
LIST OF TABLES
Table Page
I. Labor Costs • • • 0 0 • 0 0 0 • 0 0 0 0 0 0 • 106
II. Equipment Costs • • • • 0 • 0 • • 0 0 0 0 0 0 0 0 109
III. Material Costs • • 0
• • • 0 0 ... • 0 0 • • 0 111
IV. Structural Tests Applied to Shell •• • • • 0 0 0 113
v. Properties of Column Sections •• • 0 •
• • • • 0 115
VI. Test IV Strain and Stress Data • • 0 0 0 0 0 0 0 116
VII. Axial Strain Under Uniformly Distributed Load •• 121
VIII.· Stress Conditions Under Uniformly Distributed
Gravity Load from Test IV Data • • • • • • • • 122
IX. Test III Strain and Stress Data. 0 0 0 0 • 0 • • 126
X. Summary of Maximum Stress Values from Test III • 128
XT. Material Savings by Concentric Load Design • 0 0 146
XII. Adjusted Material Costs for Twenty Foot
Square H-P Shell , • , • • • • • • • • • • . • • • 0 148
vi
LIST OF FIGURES
Figure Page
1. Four Basic Hyperbolic.Paraboloid Shell
Configurations • • • • • • • • • • • 0 0 • 0
• • 5
2. The Hyperbolic Paraboloid Shell Configuration
Used in This Study • • • • • • • • • • • • • • 0 5
3. A Typical Parabolic Arch with Horizontal Load
Distribution • • • • • • • q • • • • • • • • • 0 35
Quadrants Showing Dimensions and Stress··
Distribution in the Parabolic Arches • • • • • • 35
5. Shear Distribution in the Shell 0 0 0 0 0 • 0 0 0 37
6. Cross-sectional Views of Exterior Edge Beam
Sections • • • • e o o o o o o o o o o .. • • 0 38
Shell and Edge Beam Reinforcing Steel •• • • 0 0 0 42
Tie Connection Welded to Bearing Plate on Top
of the Column. • • • • • • • • • • • • • • • 0 • 42
Precast Quadrapt Showing Lift Rings •• 0 0 0 0 0 0 45
Lift Ring Detail • • 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 45
11. Side View of Structure Showing Lift, Drag, and
Moment Reactions • • • • • • • • • • • • • • • • 48
Column Steel Arrangement • 0 0 0 0 0 0 0 • 0 0 0 0 48
13. Haunch Detail • • • • • • • 0 0 0 0 0 0 0 0 0 0 0 0 52:
14. Column and Haunch Steel Cage • • 0 0 0 • 0 0 • 0 0 52'
15. Fboting Details. • • • • • • • •• 0 0 0 0 0 0 0 • 55
16. Soil Reactions to Footitig atid Wing Walls Against
Overturning Moments. • • • • • • • • • • • • • • 55
17. Shell Form Ready for Precasting. • • • • 0 0 0 0 • 65
vii
LIST OF FIGURES (Continued)
Figure Page
18 o As:sembly Jig • • • • • • • • • • • • • • • • .. o .. 70
19. Temporary Center Support • 0 • 0 • 0 0 0 0 75
zo. Lifting Frame •• 0 0 0 0 • 0 0 0 0 0 0 0 0 0 0 78
21. Erected Column Showing Cribbing. • 0 0 0 0 0 0 0 0 78
Worksite Layout Showing Construction Control
Points o o • o o c o o o o o o o o o o o o 0 0 0 82
23 0 Material Layout for Assembly • • 0 0 0 0 0 0 0 0 0 82
viii
LIST OF FIGURES (Continued)
Figure Page
37. Roof and Column Deflection Curves for Test II •• 132
38. Roof and.Column Deflection·cu.rves for Test.III.
and IV. • • '.~ • 0 0 0 0 ~ 0 0 0 0 0 ~ 0 0 0 0 133
39. Stress Versus Load and Time Curves, ~est IV • • • 134
40:. Labor Requirements.Adjusted by Three Methods
for Comparison. • • • • • • • • • • • • • • • 0
ix
CHAPTER I
INTRODUCTION
11
built-in 11 fire protection and low maintenance costs, this
iv versatile product Vi has become popular with designers as
a structural material.
Until the late 1950us the principal precast concrete
construction consisted of the use of concrete blocks~ which
limited construction mostly to modular·units. Some exper=
imental work in tilt-up construction was also conducted.
However, during the past five years the use of concrete
shells utilizing the four basic geometric shapes previously
listed, has broadened the outlook of the concrete industry
considerably. A major advantage which the use of concrete
shell roofs has introduced is the large savings in material
and reinforcement while covering large floor areas. Due to
1
2
The Problem
Objectives
Limitations
(8)
(C)
LITERATURE REVIEW
Introduction
6
7
Precasting
tions, two curved wall sections~ two roof beams and two
roof sections. Components were precast in IBEC 1 s casting
yard near San Juan and trucked 45 miles to the site. A
test house was erected in one hour using a seven man crew.
Faerber (3) describes the construction of a precast
folded plate roof for a residence in Naples, Florida. The
building was designed in the shape of an octagon, incorpo-
rating eight separate gables, each designed as a folded
plate. The roof slabs were cast one on top of the other
as in the lift slab method of construction. Sheets of
polyethylene film were used to facilitate separation. Each
of the 400 square foot sections were four inches thick and
weighed 11 tons.
A slightly different approach toward precast concrete
was conducted by Riley (4). He constructed a barrel
shaped roof by shaping the earth into the desired form and
precasting his roof in place. The columns were placed
prior to casting the roof in order to allow sleeves cast
in the roof to utilize the columns for stability during
the erection sequence. The roof was cured on the casting
bed, then raised by means of hydraulic jacks. After the
roof was raised, the mound was leveled off and the floor
12
loads.
26
aggregateso
concrete a
similar strengths.
shell surface.
ing erection.
position.
(7) Foundations.
30
31
Research Outline
a9sembly.
flection data.
(a) Design live load was 40 pounds per
roof surface.
loading.
Shell Design
JJ
34
xy xy
where
H = the horizontal thrust at the end of each arch
per foot of shell widtho
hxy = vertical rise of each archo
L = horizontal length of each archo
Further simplification of the analysis for horizontal
. ld s H
thrus t , H , y1e W•aob F" 4 h
2 h , 1gure , w ere
=
1------L.--..L-,--L.-.-....1..-.,.......l--4--1-.......-'---'-~..........,_---l......,....---t-----.
fw
~--A---L--L-.L-4---L----J-~~~~
H
I
hxy H
L/2 . .I
L
Figure 3. A Typical Parabolic Arch With
Horizontal Load Distribution.
tt
~
'
....... -..~~~
.cL
t
t~_..~ _..~ i
~
-,
I. b b J
Se ct ion PP
p
Pion
tll~~==~#~6'~-~-;;:::;:::--;-;-··_/_#_2;_'~~~~~ij·
Section A A
24"
B A
Quadrant
Section BB
2
B A
c c
r
24"
·1
Quadrant Slope= 0. 72." I Ft. ~·
I
r
3.5" Slope= 1.62"
I Ft.
_J.
L Section CC
I. 4• .1 24"
--T 18"
yColumn S Hounch
I
I
D
I
I
I
D Quadrant
4
Quadrant
3
r
3.5"
__ JI l_
Section DD
1. 4"
.1
No. 6 bars were used. The compressive steel was also re-
quired to resist bending in the edge beam during eccentric
roof loads.
of dead loads greater than the design load which i1vere in-
duced during the testing procedure.
The welding of this plate to the horizontal interior
edge beams was also noted as a possible weak point; there-
fore, a weld leg width of 3/S in& was made on both sides
of the 24 inch bar. This gave a. calculated allowable load
capability bf q = 4S in~ x 3/S in. x .707 x 5,000 = 63,630
edge beamo
section was even with the top slope of the edge beams for
welding.
Shell Dimensions
Shell Concrete
.
To reduce the over-all weight which was lifted during
assembly of the shell on the column, the shell quadrants
were p·recast of lightweight aggregate concrete which had a
21 day density of 117 lbs./ft~ Lightweight concrete,
45
Column Design
where:
1 lb.f Secz
. . , NewtoJl' s Second Law Coefficient =-~--·--
32.2 lbm. ft.
P = Air Density = ~070 lbm/ft!
48
Wind -
1·
....,::::-----'----flR;\--,.[R_y
.._R_:.;.~---~-::-- --"T"""f
·1
H
l
D
2.5"
# 2 ~ Ties, 10 o. c.
11 I 7''
12
11
_J
11
2.5" 2. 5" 2.5
11
I0
6. Moment, M = Mo 0
Ne °H 0
w2 o;• V2o
40, 000 ft.-lbs.~, or MDL = 20, 000 ft.-lb.s./c.olumn. The ref ore~
the total moment acting on each column will be MT = Mw +
MDL = 9,g 60 + 4 o,goo = 24,530 ft.~lbs.
By the method of ultimate strength design from
Reinforced Concrete 38 (28) a design factor of 2 is used.
The design axial load, P = 14.52 kips, and the ultimate
design load, Pu = 2(14.52) = 29.04 kips. The design
moment, MT = 24.58 kip-ft., thus, the ultimate moment,
Mu= 2(24.58).= 49.16 kip-ft.
From Table 6, "Eccentrically Loaded Tied Columns uu ( 28)
for f c = 3,000 psi, f y = 40,000 psi~ and column size = 10
in. by 12 in., p = 30 kips, and l'1u = 56 kip-ft ~ 4 No. 10
u 0
~ Design
-- Sv
122:0.9 x 6.__ = 61L700 == "' 98 sq. in ~ Two No • 7 bars,
20,000 x 7/8 x 9 63,000
52
11
20
11
6 o. c. Eoch Side, Buttwelded
To Angle 11
18
Welded
Colu mn Steel
#8<P
2 • H • aa
cos =2 · 10,500 · 10.43 = 21,950 lbs. The horizontal
17
component of this thrust was H0 = 2 · H · a = 21,000 lbs.
"He
As = 1'S = 21,000 The As was_ furnished by
20 , 000 = 1.05 sq. in.
one No.• 1.0 bar which had an area of l. 47 sq. in.
The length of weld which was necessary to develop the
p
full strength of the tie bar was L = • 707 h •S. =
9
21,000 The tie bar was welded
.707 (3/8)(14,000) = 5.66 inches.
to the inverted "T" sections, which were welded to the
bearing plate on each column, after the column footings
were cast.
Footing Design
t
I
s,
I
'
Pion
-I I
L 3
/
I
I
I
J,.. j
03 !'/ I
I .34 0
I
01 I
I
0=4.5
I
\ I
I
I .560
..,
-...
I
'
- I
I
I
S2
- I
I
.32 0
I
I
I
I
I I
-- ~
~ 0'2
L - --
·~
\1(2 '\'
~
---..:! >0
:------__
Section AA I. 20
11
.I
Figure 16 . Soil Reactions to Footing and Wing
Figure 15. Footing Details. .Walls Against Overturning Moments.
56
= 12 (1.310) [. 1 + (1 + 43.6
1 • 310 )
1
;2 J = (0.656)(6.86), or d =
1 .$i
6
As = ~ 5 14 = .076 sq. in. Two No. 4 bars top and
2
bottom, As= .40 sq~ in., were used to satisfy As •
1
The As was the governing value of steel area, so
1
this area was checked for bond. For f'c = 3,000 psi~ and
Vmax 3,~00
vmax = s1 = 3 ' 500 1b s • ' u = E •jd = 3.1 x 7 8 x 6 =
0
215 psi < 300 psi. Four No. 4 bars were used ;for the wing
wall steel, Figure 15.
CHAPTER V
Introduction
and columnso
Column Forming
59
60
Shell Forming
Column Curing
first four days, and in the mornings only during the next
four days. No water was added after the eighth day. The
sun shining through the plastic covering during the day
raised the curing temperature and vaporized the moisture.
This produced a curing condition similar to factory con-
trolled curing. After 14 days~ the plastic covering~ bur-
lap material and side forms were removed. Figure 9 shows
a typical quadrant ready to be removed from the form.
Footing Steel
Assembly Supports
Assembl;y Frame
1~"x 12
11
r:
Turnbuckle
3 -0
~
1 11
..,
Screw Jack
Adjustment Joint
f
2{ x 2f x Angle
1 11
3 -6
l Figure 18.
1
20 -0
Assembly Jig.
11
1-
:I
-.J
0
?l
(3) A turnbuckle mounted on the top horizontal
brace which could be extended or retracted
over a range of approximately 12 incheso
(4) Three screwjack legs in the base of each
tower to provide fine adjustment vertically
and to plumb the towerso
(5) Large increments of vertical adjustments in
the corner towers by overlapped tower leg
sections and braces with spaced bolt holeso
The design of the assembly frame was based on a
weight per quadrant of approximately 3,200 poundso By
using a design factor of 2.0, the working load per quad-
rant was 6~400 poundso During the assembly~ each quadrant
was supported by the assembly jig tower, the concrete
column, and two wooden supportso The maximum stress con-
dition for the tower would probably occur with the quad=
rant supported by the tower and the wooden support at the
centero Then, the tower would support one=half of the
working load, or 3~200 poundso
The corner leg of the tower was designed to carry the
full load of 3~200 poundso The slenderness or l/r ratio
governed the design; thus, for steel columns (26),
18/000 =
(84_o 39) 2
18,000
Center Support
edge beam.
leg while the support was being placed over the bar or
End Supports
the support.
75
11
36
11 11
12 12" 12
f l T :1
c
_i;i_ - -
c
_si _ _ - _c_
D c
-'1- - I 12
11
_J 12
11
11 11
4 x4
11
32
11 11
12'-6
11
2 x4
[J
11"
2
x 16 11
Screw Jock
Lifting Frame
less than the 325 psi allowable. The section modulus re-
. d was S = M
qui.re f = 6,600 in.-lb. = 5 •5 in.
. , wh'i.c h was l ess
12200 .psi
than the value of 7.94 in.' for a 4 in. by 4 in. member.
CHAPTER VI
ERECTION PROCEDURE
80
81
Site Layout
Roof Outline
Outline Of Footing
,........l(f-+-t'-E~xcovo~~ _ J ____
1
19 -2"
9 7
I. 20'-o"
.I
Figure 22. Worksite Layout Showing Construction
Control Points.
[I](
~
,
®
D Support
Column
Storage CD System
Storage
~
®
Assembly Sequence
CD -Precost Columns Erected
® - Support System Assembled
@ - Roof Assembled
columns and tie had been erectedo Each quadrant was lo-
ment. For this study, only one modular unit was erected,
time.
84
Column Erection
Support System
Shell Assembly
also noted that the tie rod would keep the third and fourth
quadrants from fitting against the vertical web of the tie
connection. The first quadrant was lowered and a portion
of the corner was removed. The lower corners of the re-
maining tl:).ree quadrants were also corrected in the same
manner.
The first q11adrant was , lifted into position and t~e
interior edge ~eam was visually aligned along the tie bar
as it was lowered. The second quadrant was lowered into
position in the same manner. Approximately 3;4 inch sepa-
rated the two quadrants along the horizontal interior edge
beam after they were initially set in place. The thirdand
fourth quadrants were lowered onto the support system with
their sloping edge beams against the first and second
quadrants. The total assembly time required to connect
the lifting frame to all quadrants and set them in posi-
tion was one hour and twenty minutes. Figure 26 shows the
shell immediately after assembly and prior to welding.
the testing of the tie bar material and the concrete sam-
ples will be discussed. Then, the procedures and equip-
ment used for load tests on the structure will be
explained.
Structural Testing
two gages were placed on each haunch arm along the center=
line of the bottom side, one gage was centered 4 inches
directly beneath the haunch on each side of the column~
99
Deflection Apparatus
4 13 14
Figure 30. The urn was clamped to the inner face of the
south column so that all movement of the structure could
be related to one point.
A datum line was scribed around the glass water level
tube on the urn. A 10 inch section from a scale with 50
divisions per inch was mounted vertically against the tube.
· A 3;s inch outside diameter plastic tube was attached to
the spigot directly below the water level tube to connect
the manometer reservoir to the movable end of the manome-
ter. The movable section of the manometer consisted of a
glass tube clamped to a 1 in. x 2 in. x 18 in. board, a
50th scale which was attached to the board behind the
glass tube, the 3;s inch plastic connector tube~ and a 1/4
inch steel rod approximately 7 feet long, Figure 30. The
bottom end of the steel rod was rigidly attached to the
top end of the board. The top end of the rod was formed
into a ring to use in suspending the manometer board from
hooks which were clamped to the edge of the shell.
The manometer was open to the atmosphere on both ends
so that no pressure differences were developed. The manom-
eter reservoir was filled with approximately 31/2 gallons
of a water and alcohol mixture to prevent freezing during
cold weather. During the testing period, water level
readings were taken each day from the datum line on the
water level tube to correct for evaporation. Readings
were taken on the movable end of the manometer at all 8
points before and after each roof load change.
102
Testing Procedure
105
106
Labor Costs
TABIE I
LABOR COSTS
----
Item Skilled Unskilled
1. Column Construction
(a) Forms 2 40
(b) Steel Forming 5 54
(c) Casting and Curing 4 12
SUBTOTAL (Man-hours) 11 106
2. Shell Construction
(a) Forms 16 96
(b) Shell Steel Forming 8 78
(c) Form Preparation and
Casting 4 16
(d) Curing -2 10
SUBTOTAL (Man-hours) 31 200
107
TABLE I (Continued)
5. Site Preparation
(a) Leveling and Smoothing 0 2
(b) Survey and Layout 2 2
(c) Foundation Excavation 2 --2
SUBTOTAL (Man- hours) 4 13
6. Site Layout
(a) Hauling Columns and
Placing 2 2
(b) Construction of Shell
Supports 2 4
(c) Removing Forms, Loading ,
and Transporting Shells
to Site 4 7
(d) Moving Support System 1 4
SUBTOTAL (Man- hours) 9 17
7. Column Erection
(a) Development of Column
Support System for
Stabilizing Column 4 25
(b) Cutting and Bending
Footing Steel 0 4
(c) Column Erection and
Plumbing 4 14
(d) Casting Column Footings 0 1
(e) Removal of Braces and Site
Cleanup 0 4
108
TABLE I ( Continued)
7. (Continued)
(f) Welding Tie 2 2
SUBTOTAL (Man- hours) 10 50
8. Support System Erection
(a) Initial Erection of
Corner Towers 3 9
(b) Final Alignment of
Tower s f or Shell
Erection 4 12
SUBTOTAL (Man- hours) 7 21
9. Erection of Structure
(a) Initial Assembly 4 12
(b) Prepar ation for Welding 2 4
(c) Welding Edge Beams and
Column to Shell
Connections 15 0
(d) Support Removal and Site
Cleanup 0 6
(e) Grouting Top of Columns 0 4
(f) ·Waterproofing Interior
Edge Beams 1 5
(g) Final Cleanup 0 4
SUBTOTAL (Man- hours) 22 35
Equipment Costs
TABLE II
EQUIPMENT COSTS
TABLE II (Continued)
Materi al Costs
TABLE III
MATERIAL COSTS
TABLE IV
STRUCTURAL TESTS APPLIED TO SHELL
I Uniformly Distributed
Gravity Load, 1 Load
Increment 21~6 psf 73 1/2 hrs~
II Uniformly Distributed
Gravity Load, 3 Load
Increments 61@7 psf 117 1/2 hrs.
III Half-roof Eccentric Load,
Uniformly Distributed, 2
Load Increments 41.3 psf 3 hrs.
IV Uniformly Distributed
Gravity Load, 4 Load
Increments 57~0 psf 74 1/2 hrs*
·-----..-------. ----·
The strain values of gages 1-20 for Test III were not
adjusted due to the residual strain which remained in the
structure after the test was completed. Strain values for
gages 21-25 for Test III and gages 1-25 for Test IV were
adjusted by using a ratio of the time of reading against
the total time of the test and adjusting the final zero
load values.
114
TABLE V
Property Values
Uncracked· Section · Cracked Section
~ =0 ~ = 90° ~ =0 ~ = 90°
(2) d (or d ) 9.5 ."in. 6.~5 · in. 9.5 in. . &..25 .:i.n,.
a:ve
(3) t 12. 0 "in. 10.0 in. 12.0 in. 10.0 in.
(4) kd 6 . o :i:n. 5.0 in. 4.15 in. 3 . 45 in.
(5) z 2.14 in. 1.81 in. 1.69 in. 1.35 in.
(6) jd 7.36 in. 4.4o in. 7.81 in. 4.90 in.
(7) 4 4 4 4
It 1957.0 in. 1176.o i n. 921.6 in. 410.2 in.
For Test III and Test IV, which were loaded on successive
TABL;E VI
Test IV Strain and Str e ss Data
11
h =113
M
M~
..Gage-4:- _- -
ELEVATION ELEVATION
I
b=IO
0
II
0
2
I
0
L...___0_10
Kd=4.15'~9.5J I
i:=, =12''~
PLAN PLAN
(a) Bending in ~ = 90° Direction (b) Bending in ~ = 0 Direction
Figure 31. Columns Showing Directions of Bending,
Dim~nsions, and Reactions.
118
0 0 0 0
l I l l l Jec=33.5 l J 111 Jec= 29.5
- - - -
(b) Time - 50 Hours
0 0 0 0 0 0
OJIIJec=55 mec=60 [ 1111 J ec =40
ebt =-80
eot11=-40
~~ ...............
-
+¢= 90°
Figure 32.
+¢ = oo
- e =Ix 16 6
in./in.
q q
11 I I I I JA H ~ I I I II
I. 11
36
I
-T -T
11
113
h~
114
11
~
l R
v
q q
~A
A
-H H
fR
R
(c) Condition 3., Table VIII (d) Free Body at Base of, Tie Connection
Figure 33. Free Body Diagrams of Theoretical Stress
Conditions Acting on Column During Unifor'mly
Distributed Roof Loading.
121
TABLE VII
AXIAL STRAIN UNDER UNIFORMLY DISTRIBUTED LOAD
prototype does not hold the column base rigid, thus, the
footings deformed and the column could not resist the
entire force in bending.
Table VIII summarizes the stress conditions of the
columns under the uniformly distributed load during Test
IV for several structural conditions, Figure 33. Condi-
tion 3 was the nearest to the actual conditions at the
site. Condition 2 assumes pinned connections at the
haunch and ridge. All of the values in Table VIII were
determined by using the maximum values from the Test IV
data.
TABLE VIII
STRESS CONDITIONS UNDER UNIFORMLY DISTRIBUTED
GRAVITY LOAD FROM TEST IV DATA
The strain and stress data for Test III are shown in
Table IX. The values of the moments calculated from the
base gage strain data are tabulated in Table X. The cal-
culated maximum overturning moment due to the eccentric
roof load, Figure 3l(b), assuming idealized conditions
was M0 = 248,000 in.-lbs. for each column. Assuming an
uncracked section, Mb = 360,000 in.-lbs. for the north
column and Mb = 318,000 in.-lbs. for the south column.
Assuming the section was cracked, Mb = 249,000 in.-lbs.
for the north column and Mb = 220,000 in.-lbs. for the
I
south column. A check of the ~ ratio indicates that the
columns should be investigated for the cracked section
condition. The comparison of the strain values against
the theoretical value shows that the cracked section
values check very closely with the idealized moment.
To compare the observed strain to the calculated
values, the maximum bending strain in the direction of
overturning for both columns was derived from the data.
The values of actual bending strain were derived from the
observed values, Figure. 34, by the following relationships.~
Figure 35: (1) ebc = t~~d • ebt' (2) e 0 t = ebt + eac' and
126
TABLE IX
TEST III STRAIN AND STRESS DATA
Gage No. € er I E I er
1 +30 +83.4 +20 +55.6 1. E.c = 1 x lo-6
2 +60 +167.0 +65 +180.5 in./in.
4 +150 +417.0 +280 +778.0
4 -70 -194.5 -420 -1168.0 2. CT= psi.
5 +10 +27.8 -10 -27.8 3. Ect = 2.78 x
6 0 0 -30 -83.4 106 psi.
7 -25 -69.5 -185 -514.0 4. Es = 30.48 x
8 -250 -695.0 -550 -1529.0 106 psi.
9 0 0 +40 +111.1
10 +200 +556.0 +430 +1195.0
11 -20 -55.6 +5 +13.9
12 -10 -27.8 +5 +13.9
13 +90 +250.0 +230 +639.0
14 -120 -334.0 -350 -972.0
15 -50 -139.0 -50 -139.0
16 -40 -111.1 -50 -139.0
17 -100 -278.0 -240 -666.0
18 -230 -639.0 -470 -1308.0
19 -40 -111.1 -90 -250.0
20 -150 -417.0 +365 +1015.0
21 -26 -792 -10 -304
22 -78 -2378 -134 -4090
23 -60 -1830 -110 -3358
24 -64 -1950 -117 -3570
25 -64 -1950 -117 -3570
-"'=
+ 0
ebt =-515
+"' =0
----
et= -499.I
1~.-1--J......L.LIO + 01~11-+~~
ebc=401
eA: 15.9
Theoretical Values
ec= 416.9
It Kd By Similar Triangles
eot
( 3) e = e +e
oc be AC
0
e
Observed Values oc
e = 1 x lo- 6 in./in.
Figure 34. Axial and Bending Strains Figure 35. Definition Sketch for
in Direction of Gverturn- Heavy Bending Strain.
ing During Test IIIo
128
where:
ebc = compressive bending strain (unknown)
ebt = tensile bending strain (unknown)
eac = compressive axial strain (unknown)
e0 c = observed compressive strain
e 0 t = observed tensile strain.
TABLE X
SUMMARY OF MAXIMUM STRESS VALUES FROM TEST III
Notes~
(1) Maximum load = 41.3 psf.
(2) Assume an uncracked section.
(3) Assume a cracked sectiono
(4) See Figure 36 for location of N.A. for eccentric loading.
-
u
Q) 0
Q)
c -.I
-.2
-.3
0 10 20 30 40 50 60 0 20 40 60 80 100 120
Uniformly Distributed Load - Lb./Sq. Ft. Time - (Sustained Max. Load) Hrs.
Figure 37. Roof and Column Deflection Curves
For Test II.
0
S. E. Corner
t:. E. Mid-Span
o N. E. Corner
• N. W. Corner
"" W. Mid-Span
• S. W. Corner
0
S. E. Corner
.6 t:. E. Mid-Span
a N. E. Corner
.5 • N. W. Corner
... W. Mid- Span
.4
• S. W. Corn er
A----~
A-----,
.3
"'
Q)
..t::.
u 0 .2
-= -I .I
c:
0 "'
Q)
..t::.
t; -2 u 0
-
Q) c:
Q)
Cl
-3 c: -.I
.9
u -.2
-5 -
.!!
Q)
Cl -:3
-4
. 0
0 10 20 30 40 50 10 20 30 40 50 60 0 20 40 60
Eccentric Load - Lb. I Sq. Ft. Uniformly Distributed Load - Lb./ Sq. Ft. Time -(Sustained Mox. Load) Hrs.
(a) Test III Data (b) Test IV Data
0 20 40 60 en -2 0 10 20 30 40· 50 0 20 40 60
0 10 20 30 40 50
Uniformly Distributed Load, Lb./ Sq. Ft. Time ( Sustoined Max. Lood= Uniformly Distributed Load, Lb./Sq. Ft. Time (Sustained Max. Loo d =
5 7 Lb.I Sq. Ft.I Hrs. 57 Lb./ Sq. Ft.I Hrs.
60 0 20 40 60
Time (Sustained Max. Load= Uniformly Distributed Load, Lb. I Sq. Ft. Time (Sustained Max. Load=
Uniformly Distributed Load, Lb./Sq. Ft.
57 Lb. /Sq. Ft. I Hrs. 57 Lb. I Sq. Ft.) Hrs.
Figure 39. Stress Versus Load and Time Curves, Test IV.
CHAPTER IX
DISCUSSION OF RESULTS
Column Forms
i'35·
136
Shell Forming
Footing Methods
The footing used in this study performed satisfactor-
ily, however, by not knowing the exact soil shear strength
and bearing capacity, the footing may have been overdesign-
ed both in regard to size of wing walls and the depth of
the footing. The footing reinforcement cage used in this
study functioned we1'1 in the assembly of the column but
required special bending during construction. This prob-
ably could be eliminated by another type of footing. For
a smaller design load and no overturning moments, a cylin-
drical footing with sufficient bearing area would be
adequate.
Lift Frame
Column Erection
column which extend into holes in the footing which are pre-
viously filled with grout, and (3) a slotted bar, cast into
the footing and column, which is welded and the joint groutedo
Cogan (36) developed an effective pipe connection by pre-
casting into the footing a 4 inch pipe sleeve which was cut
off at the correct elevation; the 4 inch pipe sleeve fit
over a 3 inch pipe insert cast into the column. The insert
had a steel ring or shoulder welded around it to give the
exact elevation. The reinforcing bars were overlapped be-
tween the footing and column, and welded. The joint was
completed by grouting with an expanding grout mix which pre~
Weld,ed Connections
Both the tension bar at the roof center and the edge
beams should have been welded in sections or strips. This
part of the shell design should have been examined more
critically. The factor of safety of the weld on the ten-
sion bar was 5.0; this indicates that the welding should
have been reduced to half of the amount used.
144
The observed cost data from this study have small sig=
nificance in its present form; however, if this data can be
adjusted by appropriate estimates based on the experience
gained from this study, the adjusted data may serve as a
useful guide for construction estimates on this type of shell.
The next four sub-sections will consider the data by
(1) adjusting the material cost where appropriate, (2) ad-
justing the labor cost data by an appropriate skill factor
based on the experience that a crew would have after becom-
ing familiar with the construction routine, (3) adjusting
the equipment costs that are related to the labor and mater-
ial reductions, and (4) converting the values to a cost per
square foot for the shell used in this study.
The final sub-section will be used to adjust the pro-
totype data to a 40 foot square shell. Only the variable
cost factors will be considered, Interest rates will not
be considered in this study; the cost data will be consid-
ered as capital costs.
TABLE XI
TOTAL $56.22
1. Welding Material
(a) Welding rod 40 lbs., $.20/lb. 25 lbs., $.20/lb. $8.oo $5.00
(b) Acetylene 7/8 bottle, $5.70 5/8 bottle, $5.70
per bottle per bottle 4.98 3.56
(c) Oxygen 7/8 bottle, $5.65 5/8 bottle, $5.65
per bottle per bottle 4.94 3.53
2. Concrete
(a) Standard weight, 3 cu. yd., $14.75 3 cu. yd., $14.75
3,000 psi per cu. yd. per cu. yd. 44.25 44.25
(b) Lightweight Ag- 3 1/2 cu. yd.' 3 1/2 cu. yd.,
regate, 3,750 $18.25/cu. yd. $18.25/cu. yd. 63.87 63.87
psi
3. Steel Material 2056 lbs., $.097/lb. 1538 lbs., $.097/lb. 199.84 149.49
4. Form Oil 5 gal., So80/gal. 5 gal., $.80/gal. 4.oo 4.oo
TOTAL $329.88 $273-70
149
TABLE XIII
LABOR ADJUSTMENTS BY SKILL FACTORS
i; Column Construction
2. (Continued)
(c) Form Preparation
and Shell
Casting .60 .Bo 1.2 1208
(d) Curing .4o .75 1.2 7o5
SUBTOTAL (Man-hours) 12.4 159·5
3. Support System Construction
(a) Fabricating Parts .50 .Bo Oo5 5.6
(b) Welding Tower
Frames .Bo 2B.B
(c) Assembly of Bolted
Components .B5 2.5
(d) Wooden Support
Fabrication .40 .75 o.B 12.0
(e) Final Adjustments
on Steel Supports .50 .50 0.5 2.0
5. Site Preparation
(a) Leveling and
Smoothing .90 1.8
(b) Survey and Layout .75 .75 1.5 1.5
(c) Foundation
Excavation .50 .50 1.0 4•.z
SUBTOTAL (Man-hours) 2.5 7.8
6. Site Layout
(av Hauling and Plac-
ing Columns .90 .90 l.B 1.8
(b) Wooden Support for
Shell Transport .50 .75 1.0 3.0
(c) Removing Shell Forms
Transporting, and
Placing Shells .40 .50 1.6 3.5
151
6. (Continued)
(a) Construction of
Cribbing for
Column .50 .75 2.0 18.8
(b) Cutting and Bending
Footing'Steel .75 3.0
(c) Column Erection .60 .75 2.4 10.5
(''.~~(:;). (d) Casting f:ooting .85 o.8
(e) Removal of Braces
and Site
Cleanup 085 3.4
(f) Tie Er~yction .80 .80 1.6 1.6
SUBTOTAL ( Man-hours) 6.o ·38.1
8. Support System Erection
(a) Initial Assembly .50 .50 1.5 4.5
(b) Final Alignment .50 .50 2.0 6.o
SUBTOTAL (Man-hours) 3.5 10.5
9. Erection of Shell
(a) Initial Assembly .40 .40 1.6 4.8
(b) Adjustment for
Welding .50 .50 1.0 2.0
(c) Welding .50 7°5
(d) Support Removal .70 4 .. 2
(e) Grouting Haunches .85 3 .. 4
(f) Waterproofing In-
terior Edge
Beams .50 .,75 0.5 3.8
(g) Final Clean-up .. 75 3 .. 0
SUBTOTAL (Man-hours) 10.6 ?l.2
FINAL TOTAL (Man::..hours) 86::.·ra 352~7
-~-------- -=
152
Symbol Item
100
----e A. First Unit Total Labor Requirements
Adjusted by Skill Factors
0 0 B. First Unit Shell Construction Labor
Requirements Adjusted by Skill Factors
80
C. Percentage Item B of First Unit Total
en
.... Labor Requirements Unadjusted
::I
0 68.9
..c
I
c:
0
60
:E
- c:
- -- -- . -----·-- ----
-....
::::>
en 40
--
0
c:
Q.>
u.... 20
Q.>
a..
TAJB:LE XIV
LABOR REDUCTIONS DUE TO CONCENTRIC LOAD DESIGN
lo Column Construction
(a} Forms 1 10
(b) Steel forming 2 JlO
2. Shell Construction
(b) Steel forming 2 10
7o Column Erection
( b) Cutting and bend-
ing footing
steel 0 .4
9o. Erection of Structure
~e) Welding
d) Grouting haunches
0 0
0 --1±_
TOTA:L· (Man-hburs) 5 58
T'.ABLE XV
ADJUSTED EQUIPMENT COSTS FOR SECOND SHELL UNIT
lo Acetylene Welder,
$3.00/hr.
(a) Forming column
·steel ,,,,.10.4 0 $31•20 0
(b) Forming shell
steel "~ 4.8 4.o 14.40 $12.00
2o Electric Welder
(a) Forming column
steel, $2.00/hr. r~:". 2.4 " 1.0 0 ,.4 ;80 2.00
(b) Tie bar erection,
$3.00/hr. ~ 1.6 ' 1.6 0 4.80 4.80
(c) Forming shell steel,
(1) Shop welding,
$2.00/hr. ,,11.2 .. 8.o .22.40 16.oo
(2) Field welding,
$3.00/hr. 3.2 2.7 9.60 8.10
(d) Footing cage, .. e
TABLE XVI
PROJECTED LABOR ESTIMATES FOR FORTY FOOT SQUARE H- P SHELL
1. Column Construction
(a) Steel Forming 5 54 1.0 1.0 5 54
(b) Casting and :
Curing 4 12 1.2 1.4 4.8 16.8
2. Shell Construction
(a) Steel Form-
ing and 8 78 1.1 1.3 8.8 101.4
Tieing
(b) Form Prepa-
ration and
Shell
Casting 4 16 1.0 2.5 4.o 40.0
(c) Curing 3 10 1.0 1.5 3.0 15.0
161
3. Site Preparation
(a) Leveling and
Smoothing 0 2 1.5 0 3.0
(b) Survey and
Layout 2 2 1.3 1.3 2.6 2.6
(c) Foundation
Excavation 2 9 2.0 2.0 4.o 18.o
4. Site Layout
(a) Hauling and
Placing
Columns 2 2 1.0 1.0 2.0 2.0
(b) Removing
Forms, Loading,
and Transport-
ing Shells to
Site 4 7 1.2 lo2 4.8 804
(c) Moving Support
System 1 4 1.1 1.1 1.1 4.4
5. E+ecti9n of·S~pports 7 21 1.1 1.1 7.7 23.1
('.;
Erection of Structure ..
6.
(a) Initial
Assembly 4 :12 1.0 1.0 4.o 12.0
(b) Preparation for
Welding 2 '.4 1.1 1.1 2.2 4.4
(c) Welding Time 15 0 1.5 22.5
(d) Support Removal
and Site
Cleanup 10 1.1 11.0
(e) Grouting
Haunches 4 1.0 4.o
(f) Waterproofing
Edge Beams 1 5 1.0 1.3 1.0 6.5
TOTAL (Man-hours) 64 252 775 3~6.6
Symbol Item
oo~~~o A. Labor Costs Adjusted for Job Experience
100
A .6. B. Shell and Column Form Costs Adjusted Over
Number of Shells Constructed
t:s------6 C. Combined Labor and Form Costs (Labor ad-
justed for experience, forms adjusted
80 for use)
- --- D. Erection Apparatus Cost Adjusted for
Number of Units Erected
c:
60
-,,,
:::>
....
0 40
Q)
-
C"
0
c:
Q)
u
~
a... 20
TABLE XVII
VARIABLE COSTS FOR FORTY FOOT SQUARE SHELL
Item Cost
on the tie bar just after the shells had been welded
together and the supports lowered, which averaged 21805
micro-ino/in., plus a maximum strain of 405 micro-in. /ino
measured during Test II.
By recalculating the value of tensile force which
should have been taken by the tie according to design,
using the actual dead load of 32.0 lbs./fto 2 , the tensile
force should have been 32,540 lbs. Assuming that the
strain of 218.5 micro-in./ino had not changed during the
four months period between the shell completion and the
testing period, the difference between 32,540 lbso and
27~955 lbso, 4,585 lbs., was the calculated maximum
shearing force resisted by each column. The calculated
unit shearing stress in the concrete by the method for
flexural members was 43.8 psi, which was approximately
one-half of the allowable shear stress.
An i nteresting correlation was noticed between the
was placed on the roof after the column haunches had been
grouted. If the load- strain trend had continued linearly,
and the columns had not been grouted, an expected strain
value of 422 micro- in . /in. would have been obtained . The
difference between 422 and the obser ved str ain value of
405 micro-in./in., or 17 micro- in./in. was considered to
be a measure of increased stiffness of the joint due to
grouting . The bending resistance was increased by approx-
imately 4 per cent by gr outing the haunch.
It was believed that the rigidity exhibited by the
construction joint was due to the wide haunch and shell
connection. This connection including the edge beam to
column weld spanned a horizontal distance of 4 feet and
was "V 11 shaped with a side slope of 3 to 10. Thus, the
joint could not react as a pinned or simple connection,
which was the design aSSUlJlption.
The tie bar area could be reduced by 14.0 per cent
because of the difference between the design tensile
stress and the measured stress. However, changes in soil
conditions due to ground water and moisture infiltration
should be considered before changing the design.
The shell quadrants were assumed to transfer all roof
lo ads, including eccentric loads, as shearing forces
through the parabolic arches into the e.dge beams of t he
quadrant . However, from visual observation during the
cantilever load of 41 . 3 lbs . /ft. 2 , the shell was subjected
to bending stresses. Cracks perpendicular to the edge
167
169
170
tion joint .
(7) The cribbing used for supporting the precast columns
during erection was an inefficient means of erection.
(8) The steel assembly jig effectively controlled the
vertical and horizontal positions of the shell
quadrant corners.
(9} The configuration of the haunch connection to the
shell produced a joint capable of resisting moment
stresses between the columns and edge beams. Grout -
ing the hauncheG increased the rigidity by 4 percent.
(10) The exterior sloped edge beams on the loaded quad-
rants were subjected to bending stresses when the
shell was eccentrically loaded. Bending cracks
occurred along the edge beam under the maximum cant-
ilever load of 41.3 lbs . /ft}
(11) The measured tensile force in the horizontal interior
edge beam was approximately 18 percent of the design
stress, 19 , 000 lbs . for a uniform roof load of 57.0
lbs./f~~ The remainder of the stress was resisted
by the splice plates joining the sloped interior
edge beams.
172
beams.
Conduct a construction engineering study for a prefab-
ricated h-p shell utilizing new methods of column erection
which were suggested in Chapter IXo
Analyze the stresses in the edge beams, tie, and column
on an h-p shell to determine the stresses resisted by each
member.
SELECTED BIBLIOGRAPHY
175
176
179
TABLE A-1
-6
'Reading Load (lbs.) - Strain (in./in. x 10 )
I "
No. Bar #1 Bar #2 Bar #1 Bar #2
. (rage #1 Gage #2 Gage #3 Gage #4
1 2,100 1,830 . -190 +105 0 -70
2 3,950 3,990 .· -310 +145 ..;..10 -155
3 . 6,100 5,640 -300 +150 -15 -230
4 7,980 8' 070 • -480 +135 -30 -330
5 9,.680 9,820 -525 +100 -35 -400
6 11,540 13,000 -560 +50 -75 -510
7 17,640 15,180 -675 -1~0 -110 -580
8 18,860 17,880 ' -690 -145 -160 -650
9 22,800 19,500 -770 -260 -190 -700
I
181
182
TABLE B:-1
STRAIN GAGE DATA FROM TIE BAR
DURING SHELL ERECTION
Quadrants Drawn
Together -200 -120 -80 -260
Quadrant Edge Beams
Welded, Column Weld
ed to Shell Angles -135 -85 +30 -130
Support System
I Lowered -250 -225 -115 -275
Remarks:
(1) Shell erection was begun on October 8, 1963.
(2) Supports were lowered on October 25, 1963.
(3) Tension is indicated by minus (-) signs.
APPENDIX C
183
184
~efl. =Deflection
Acc .. Defl. =Accumulated Deflection
S.E.C. = Southeast Corner of Roof
E.M.S. = East Mid-span of Roof Edge
N.E.C. =Northeast Corner of Roof
.N. Col. =North Column
.N.W.C. = Northw~st Corner of Roof
W.M.S. = West Mid-span of Rooi Edge
S.W.C. =Southwest Corner of Roof
C. of' R .=<Center of Rbof-
Deflections ~were_measul'.led ...::.in' inches.
Positive (+) "qeflection was measured downward.
TABLE C-1
'•
c. of R. +.06 +.06 +.06 +.12 -.18 -.06 +.02 -.04 -.08 -.12
,.;.,
TABLE C-2
TEST II ROOF AND COLUMN DEFLECTION DATA
Load 25.3 psf 46.9 psf 61.7 psf 61.7 psf 61.7 psf 61.-7 psf 61.7 nsf 61.7 nsf 0
'
·Time . . l .l
2 hrs. 3 hrs. 4 hrs. 24 hrs. 49 hrs. 74 hrs. 96.5 hrs. lla hrs. 117! hrs .
Loe a- .,
W.M.S. ""'". 08 -.08 ft.42 ft 34 I+- .18 fto52 +.06 +.58 ..... 16 +-.42 ft- .'"02 I+-. 41'.:l
0 0 It-. 44 -.02 11-.42 ... 46 .:. . 04
s.w.c. ft.02 +.02 ft-.20 H-.22 ft-. 08 +.30 +.10 +.40 .... 16 +.24 ft-. 02 i-. 26 +-.02 f\-.28 II-. 02 It-. 30 ... 30 0
..
C.ofR. ft .14 +.14 ft,04 ft .18 --.to +.08 +.06 ft- .14 0 if- .14 ft-.10 i- .24 -.02 ft- .22 .... 04 fl-.18 ... 14 ~.04
'
TABLE C-3 186
TEST III ROOF AND COLUMN DEFLECTION DATA
Location
cc.
efl .:
S.E.C. 2.5 3.06 1.10
E.M.S. 3.0 3.66 1.34
N.E.C. 2.5 3.16 1.20
~.Col. 0 -.02 -.06
N.W.C. 2.5 3.26 +.92
W.M.S. 3.1 3.90 1.10
s.w.c. 2.7 3.36 +.98
C.ofR. -.0 +.02 0
TABLE C-4
TEST IV ROOF AND COLUMN DEFLECTION DATA
L-oad 19. 3 psf 35.4 psf 49,0 psf 57.0 psf 57.0 psf 0
Time l~ hrs. 21 hrs. 3l hrs. 4!_ hrs. 20 hrs. 51 hrs.
'Loca-
tion ·- Deflection (Inches
Defl. Acc. pefl. Acc. Defl. Acc. Defl. Acc. Defl. Acc. Defl. Acc.
Defl. Defl. Defl. Defl. Defl. Defl.
S.E.C. +.10 +.10 +.12 +~-22 +.14 +.36 -.io +.26 f+.04 +.30 0 +.30
E~M.S. +.30 +.30 +.04 +.34 +.18 +.52 -.16 +.36 f+.02 +.38 -.04 +.34
_N .E .C, +.16 +.16 +.04 +.20 f+.22 +.42 -.28 +.14 1+.04 f+,18 +.14 +.32
N.Col. 0 0 -.02 -.02 0 -.02 1+.02 0 ""'. 04 ~.04 0 -.04
N .W.C. -.10 -.10 : ,,0 -.:.. .i..i:F· 1--'~ 1:4: I- .2,4 f+: 24 - ... ,0 ::....02~ + .02·' -·30 -.32
W.M.S. ..... 10 -.10 +.14 +.04 -.12 i-. OR.
l -
f+.32 +.24 t- .10 +.14 -.40 --.26
s.w.c. -.12 -.12 f+.06 -.06 -.12 -.18 f+.26 +.08 -.06 +.02 -.30 --.28 I
C.ofR. 0 0 f+.04 +.04 f+,10 +.14 1+.04 f+ .18 i-,06 +.12 I - .10 f+.02
APPENDIX D
187
TABLE D-1
TEST II OBSERVED STRAIN
(Inches x io-6)
Gage No.
1 0 -20 +20 +60 +50 -40 +10 +35 +20 +100
2 -35 +10 +45 +20 -65 -10 +20 0
3 -40 +25 +70 +40 -40 +10 +30 +10 I -10 ,.
4 -40 -5 +40 +10 -50 -10 +15 +15 +5
5 -30 +10 +40 +30 -50 0 +30 +120 +90
6 +250 +220 +280 +335 +220 +220 +185 +100 +115
7 -50 -20 +15 +10 +220 -25 -40 -50 -85
8 -40 -20 +5 0 -45 +10 -10 0. -5
9 +10 +10 +50 +35 +55 +25 +10 +20 +10
10 -5 +20 +60 +55 +50 +20 0 0 -40
11 -10 +10 +50 -10 -45 -10 -10 -30 0
12 -20 -10 +20 -10 -80 -20 -20 -25 0
13 +5 +60 -85 0 -30 -45 -50 -60 -60
14 +20 -80 -15 -90 -160 +30 +15 +30 +60
15 -20 -20 +10 -20 -50 -10 -15 -15 0
16 -10 0 +30 -10 -35 0 0 -15 0
17 -25 -20 +30 +60 -65 +10 -65 +5 +10
18 -30 -40 -5 +5 -50 0 -30 +15 +10
19 -20 -40 -20 -45 -40 -90 +20 -130 -110
20 -5 +30 +70 +80 +15 +20 +20 +15 -10
21 --- --- --- --- -315 -320 -310 -285 -190
22 0 ~320 -415 -415 -420 -430 -440 -440 0
23 -15 -265 -365 -340 -370 -370 -380 -370 -30
24 -5 -290 -395 -420 -420 ~415 -420 -425 -10
25 0 ~5 -270 ~360 -380 -380 -370 -380 -385 -10
Remarks: (1) Test dates were February 24-29, 1964. (2) Loading, uniformly distribut-
ed gravity load. (3) Steel gage factor, F=2.ll. (4) Concrete gage factor, F=2.13.
(5) Steel gage resistance, 120 ohms. (6) Concrete gage resistance, 300 ohms. (7)
Strain data is uncorrected. (8) Minus (-) sign -~-ndicates tens~__o_n__._______________
189
TABLE D-2
Gage No.
1 0 +30 +20 0
2 +60 +65 -10
3 +150 +280 +20
4 --10 -420 -30
5 +10. -10 0
6 0 -30 -10
7 -25 -185 -50
8 -250 -550 0
9 0 +40 +30
10 +200 +430 +60 l
11 -20 +5 +15 I
12 -10 +5 -30
13 +30 +230 0
14 +120 -350 -40
15 +50 -50 -25
16
17
18
+40
-100
+230
-50
-240
-470
-15
-85
-60
I
·I
19 I
-40 -90 -110
20 +150 +365 +85 II
21 -30 -15 -10
I
22 -70 -120 +20
23 -60 -110 0
24 -60 -110 +10
25 0 -60 -110 +10
I Remarks:
(1) Testdate, March 2, 1964.
(2) Load, uniformly distributed on half of
the roof.
($)Same as Remarks (3) through (8), Test II.
TABLE D-::S
Gage No.
1 0 +10 +10 +20 +35 -10 -55 -70
2 0 -10 0 +20 -30 -85 -85
3 +10 +20 +40 +80 +80 +40 -90
4 +25 +20 +40 +20 -30 -70 -80
5 +5 0 +10 +20 -30 -75 -70
6 +10 0 +10 +20 -15 -60 -60
7 +20 +20 +30 +40 -25 -95 -155
8 +30 +20 +40 +20 -10 -80 -80
9 +5 +10 +20 +25 -20 -40 -90
10 -10 +10 0 +45 0 -10 -130
11 +10 +10 +10 +20 -30 ... so -45
12 +10 +5 +10 +20 -40 -70 -45
13 +5 +20 +20 +40 +25 +20 +50
14 +20 +20 +40 +30 -30 -65 -75
15 +10 0 +10 +10 -40 -75 -70
16 +5 0 +10 +10 -30 -65 -70
17 0 0 -10 0 -20 -40 0
18 +20 +15 +30 +15 +170 +130 +130
19 +20 +50 +50 +60 +105 +50 -70
20 +5 +40 +30 +60 +20 +40 +40
21 -35 -50 -65 -75 -60 -50 +5
22 -160 -260 -360 -420 -430 -380 +25
23 -110 -190 -270 -320 -320 -310 0
24 -130 -230 -310 -375 -390 -380 -10
25 0 -120 -210 -280 -340 -360 -345 -5
Remarks: (1) Test dates were 3-5 March, 1964. O ) Loading was by uniformly distributed
gravity load. (3) Same as Remarks (3) through (8)' Test II.
VI'F'A
Ronald Tacie Noyes
Candidate for the Degree of
Master of Science
01 Jan 1971
P. V. Banavalkar
J. E. Parker
Recommended Citation
Gergely, Peter; Banavalkar, P. V.; and Parker, J. E., "The analysis and behavior of thin-steel hyperbolic
paraboloid shells" (1971). Center for Cold-Formed Steel Structures Library. 160.
https://scholarsmine.mst.edu/ccfss-library/160
This Technical Report is brought to you for free and open access by Scholars' Mine. It has been accepted for
inclusion in Center for Cold-Formed Steel Structures Library by an authorized administrator of Scholars' Mine. This
work is protected by U. S. Copyright Law. Unauthorized use including reproduction for redistribution requires the
permission of the copyright holder. For more information, please contact scholarsmine@mst.edu.
CCFSS LIBRARY Gergely, Peter, Banavalkar, P.
22 3 * 178 V., Parker, J. E., THE ANALYSIS
1971 AND BEHAVIOR OF THIN-STEEL
HYPERBOLIC PARABOLOID SHELLS
Technical Library
Center for Cold-Formed Steel Structures
University of Missouri-Rolla
Rolla, MO 65401
THE ANALYSIS AND BEHAVIOR OF THIN-STEEL
HYPERBOLIC PARABOLOID SHELLS
by
Peter Gergely
P. V. Banavalkar
and
J. E. Parker
January 1971
the surf ace the internal stresses in the deck are generally
low and the def tections are small. 2) Since a hypar surface can
be generated by straight lines, thin-steel or light-gage
panels may be used to form the shells furthermore such panels
are well suited to carry the in-plane shear forces in hypar
shells. 3) Basic hypar units can be combined in a large
variety of ways to produce attractive roofs (Fig. 1-2, page 216).
4) The dead load to live toad ratio ia very low in the case
- i -
The following are the major problems associated with
the design of thin-steel hypar structuresa 1) The deflections,
stresses and the stability of hypars depends greatly on
the shear rigidity of the thin-steel deck. This property
must be evaluated experimentally for each combination of
decking, connections to edge members, and seam connections.
Furthermore, in the case of hypars the deck is warped
and thus the shear rigidity may be different from that
of an equivalent flat diaphragm. 2) The deck may buckle
due to the shear a~resses1 and the bucklinl0 "'1st be
evaluated for highly orthotropic shells. 3) The design
of thiR•steel h}"P&r structures ia generally governed
by stiffness (deflections or buckling) requirements. The
evaluation of thP. deflections is a very complex matter
because it depends on the deck rigidity, the edge member
axial and bending stiffnesses, and on the eccentricity
of the deck-to-edge member connection. 4) lf the curvature
(or rise-to-span ratio) of a hypar is small, the deflections
may be very large and a conaidera~le portion of the load
is carried by bending oather than by membrane shear.
5) Partial or concentrated loads may cause large local
deflections, eepecially if single-layer decks are used.
The present investiaation studied all the above-
mentioned factors. The experimental and the analytical
studies are summarized briefly in the following paragraphs.
- ii -
The experimental investigation consisted of four
types of tests& a) Four medium-scale (12 ft by 12 ft in plan)
inveJrted umbrella tests to study the stresses, deflections,
&nd the deck bucklings b) Test on a small-scale (2 ft by
2 ft) inverted umbrella structure to •tudy scaling effects
and the overall buckling of hyparss c) Sixteen flat shear
tests to determine the shear rigidity of the decks used
in the hypar testes d) Twelve saddle-shaped hypar tests
(5 ft by 5 ft in plan) with various rise-to-span ratios to
evaluate the effect of rise or warping on the shear rigidity
and to study other factors such as partial loading and
single versus double layered decks. Photos of the various
types of tests are shown in Figs. 7.1 to 7.5. The experimental
program is described in detail in Chapter VII.
Prior to the main test program, several small-
scale (2 ft by 2 ft) four-quadrant tests and medium-scale
single-quadrant tests were also conducted. These tests
were however discontinued because of the severe .. ecaling
effects in the case of the small-scale models and the
violation of the synvnetry ronditions in the case of
single-quadrant experiments where the neighboring
quadrants were missing. Nevertheless, these teats produced
useful qualit•tive information and experience with
manuafacturing and testing thin-steel hypar structures.
The edge members of the umbrella-type specimens
were made of tubular members since this afforded easy
connection of the warped surface to the straight edges.
- 111 -
The decking consisted of single or double layers of
standard corrugated panels. One layer was connected
to the edge members with sheet metal screws at various
spactngs. The seam connections between the panels
was also by means of sheet metal screws. In the
case of shells with two layers, the top layer was
connected to the bottom layer in a similar manner.
- iv -
The incl!"eaae in shear stiffness due to the addition
of a second layer of deck was found to be only about
1/3 if the aecon4 layer was connected only to the
first layer and not directly to the edge members. Similarly,
the deflections of a double-layered shell are more than
half of those of a corresponding single shell. If the
two layers are interconnected with sheet metal screws
(on an 8 in. grid in the present saddle-shaped hypar
tests), the deflections are further reduced by about
10 to 20J., depending on the rise ratio.
- ,, -
for the design of flat hypar structures. However, the
design of the connections (seam or edge) may be based
on the shear forces obtained from the simple membrane
theory.
- vi -
shear rigidities the deck may buckle. As an example, a
12 ft by 12 ft model having a single layer 24 gage
corrugated sheet deck buckled at a uniform load of 70 psf
(see Fig. ). Thia model had relatively stiff edge
members (3 in. dia. tubular sections). The corner
deflections remained linear with increasing load beyond
the buckling load.
- vii -
The stiffness of the eccentric edge members were
properly accounted for in the mathematical representation
of the structure. The connection of the decks ~6 the
edge members may allow rotation about the axis of
the edge members and movement normal to the edges due
to slip at the connections. These possibilities were
also considered in the analysis.
- viii -
part of the investigation. The stresses in most types
of hypars are low and th design is usually controlled
by deflection or buckling limitations.
- ix -
are about four times greater than those due to symmetric
loading. This increase of deflections obviously depends
on the type of structures in this case much of the
flexibility was due to the bending of the central column
of the umbrella structure,
Since the instability analysis of hypars by the
finite element method involves considerable amount
of computer capacity and expense, approximate methods
were developed for the calculation of buckling loads.
The buckling of the compression edge members was studied
by isolating them from the structure, The instability
of columns loaded by tangential axial forces that
to the member
remain parallelAduring deflection was evaluated. The
results are tabulated in Fig. 6-13, page 287.
- x -
-INDEX
Pase
NOMEl':CLATURE l
I. INTRODUCTION 7
1. HYPAR ROOFS 7
2. LITERATURE REVIEW 9
1. INTRODUCTION 31
2. ELEMENT STIFFNESS 33
A. DISPLACEMENT FUNCTIONS 33
ME.rh~D
'( ,, )
B. CURVED ELEMENT ( ct - 3:6
6. SOLUTIONS OF EQUATIONS 56
7. STRESS ANALYSIS 57
A. DECK STRESSES 58
B. BEAM STRESSES 59
IV. GENERAL COMPARATIVE STUDY 61
1. INTRODUCTION 61
2. COMPARISON OF 'METHOD 'a' WITH OTHER
NUMERICAL SOLUTIONS 62
Z. GEOMETRICAL SHAPE 92
Page
3. DECK PROPERTIES 94
A. SHEAR RIGIDITY 94
B. THICKNESS OF THE CORRUGATED DECK 95
C. NUMBER OF DECKS 96
4. BOUNDARY CONDITIONS 97
A. EDGE MEMBER PROPERTIES 98
·-
.# ... ' .....
-vi·
CLflSSAPY OF TERt'S
-1-
-Z·
Ii'JTRODUCTION
I. 1. HYPJl..R ROOFS
- ----
T!-ic hyperbolic paraboloid shell ro6f, like any other
form of shell is one of the types of construction that makes
efficient use of materials by depending primarily upon the
form or shape for strength rather than on ~ass. The doubly
curved surface of a hypar shell is composed of straight lines
in two directions (Fig. 1.1). From the construction point of
view, this property is very attractive. It facilitates the
use of straight members for formwork and reinforcing steel in
the case of concrete hypars. This very feature also allows
the use of light gage steel deck panels, which could be easily
warped to the required degree to form the hypar surface.
The hypar surface shown in Fig. 1.1, can be pener-
atcd in two ways 1 : (1) The surface can be defined by woving
a convex parabola one in a direction parallel to itself, over
a concave parabola BDA. The parabola ODC lies in the plane
perpendicular to that of BDA. (2) The surface can also be
defined as a warped parallelogram. The surface can be gener-
ated (Fir. 1.1) by movinr along y-axis, a strai~1t line that
remains parallel to the xz-plane at all tiwes but pivots while
sliding alonr. the straight line AC. Physically the surface
can be visualized as a warped parallelogram OBCA, obtained by
depressing the corner H through a distance CH. By means of
s i mi 1 a r tr i an g I cs {Fig . 1. 1) ~ it can b c ea s i 1 y sh own that the
I. 2. LITERATURE REVIF.hl'
built-up sections. The loading was applied with sand bag•. Strains
Yu and
Kriz 31 tested a concrete inverted umbrella shell 24'x24' in
plan, in which upturned edge beams were used. The symmetri-
cal and unsymmetrical loading was simulated by the discrete
loads. The measured strains and deflections were presented.
Three large scale tests were conducted for hypars using
orthotropic deck as a shell surface:
Nilson 3 Z tested a lS'xlS' hypar quadrant with simu-
lated boundary conditions of the adjacent quadrant. A single
layer of cellular deck was connected to the channels, which
were used as the edge nembers, by ~eans of a warped plate. Uni-
form loading was applied by 25 jacks. The load deflection
curve and the measured m~mbrane stresses were reported.
Two large scale hypar models SO'x30' in plan, were
tested rcccntly 33 . Two different cellular orthotropic single
layer decks were used in each case. 14 WF sections were used
as the edge members, uhich were allowed to move freely in the
plane of the hypa.r but were supported against the vertical
movement. The normal uniformly distributed load was applied by
creating a vacuum in the enc.lo!:;ed chamber.
I. 4 ACKNOWL~DGMENTS
DECK PR0PERTI ES
I I . 1. l1JTYHlDUCTION
-18-
-19-
E
y • ~C
x x
+ ~y 2-lb
xy -~
y 2-lc
c;ff
Solving the equations for the stresses we get,
C1x Ex vxyEx 0
1
C1
y • vyxEy Ey 0 2-2
Tf·vxyvyxl
't'xy 0 0 {l~vxy"yx)Geff
The terms "xy and "yx represent the coupling effect o the
actions (stress or strain) applied in two perpendicular direc-
tions. From Fig. 2.2, it is obvious that these coefficients
(vxy and vyx) cannot be equal. For example, consider the
cellular deck shown in Fig. 2-2c, where a uniform stress
applied along the bottom plate of the deck in the x-di~ection
\)
yx
. 2-4
• iJ
x •
xt EX + Blt £y E 2-7
- ~
Fx t E1 t
where Nx • axt; Ext • (1-vv ); Elt • (l·vvyxl
yx
. 2-8
1 1 1
V' -= ! a x'cx ' + a '£ y '
!y + ! -r xy
' 'y xy
' ' 2-lOb
for both single and double decks with varying rise to span
ratios.
It was found that for the lower values of a (<0.12),
the center deflection of all-supported hypars was proportional
to the value of a. Both single and double deck hypars were
tested for three different rise to span ratios (1/8, 1/5, 1/3).
The theory developed in Chapter III correctly predicts the
effect of a and rise independently on the central defl~ction
III.I. INTRODUCTION
As discussed in Chapter I, the solution of the shal-
low shell equation for realistic boundary conditions is an
extre~ely complicated mathematical proposition. This necessi-
tates the use of numerical methods. The finite element method
Has chosen because of its versatality in handling realistic
boundary conditions, different structural configurations, ortho-
tropic deck materials and any forms of loading, with ease.
The method has also demonstrated good convergence character-
istics.
The finite ele~ent method based on the stiffness
analysis uses the principle of minimum potential energy. The
total potential enerr,y of an elastic system, for a geo~etri
.!!_ •
a~.
a
ati:- (P+V) • 0 3-2
1 1
-31-
P.1 c:
-av
-- 3-3
()[\.
1
3-5
model.
The linear elastic stiffness analysis consists of
four important steps:
(1) The formulation of the element stiffness matrix.
(2) The formulation of the master stiffness matrix for
the entire structure by assembling individual elements.
(3) The solution of Eq. 3-4 for the given boundary con-
ditions and loading.
(4) The interpretation of the deflected shape and the
computation of stresses and forces.
Two alternative stiffness formulation methods are
studied here:
Method 'a': The use of rectangular curved elements, based on
shallow shell theory.
~
Method 'b': The actual shell surface is approximated by the
A
assemblage of flat rectangular elements.
III.2. ELEMENT STIFFNESS
The elements rectangular in plan are selected. These
elements are very simple to formulate and for the structure
under consideration, their limitation of application to the
rectilinear rectangular boundaries, is not considered to be of
any serious consequence.
As shown in Eq. 3-lc, the element stiffness matrix
can be derived from the strain energy U of an element, ex-
pressed in terms of an assumed displacement field.
A. DISPLACEMENT FUNCTIONS
The displacement fields assumed for the analysis are
-34-
as follows:
1
U = ao [(x-a)(y-b)ul - x(y-b)uz + xyu 3 - (x-a) yu 4 ] 3-6a
1
V =as· ((x-a) (y-h)vl - X(y·b)vz + xyv 3 - (x-a) yv 4 ] 3-6b
- ax(x-a) 3 (b 3+2y 2 -3by 2 )ey 1 -a(x 3 -ax 2 )(b 3+2y 3 -3by 2 )ey 2
+ b(3ax 2 -zx 3)(y 3 -ty 2 )ex 3+h(a 3+zx 3 -3ax 2)(y 3 -by 2)ex 4
2 2 2 2
+ abxy(x-a) (y-b) exyl+abxy(x -ax)(y-b) exyZ
3-7
One may also note that the inplane rotation about the z-axis
is being omitted in this formulation.
The function for w, normal bending displacement, is
of the cubic order. The terms corresponding to the degrees of
freedom w, ex and ey are obtained by the cross product of the
corresponding terms for the beam function in x and y-directions.
In order to represent the constant strain corresponding to the
twisting term ~!ay i.e., the term 'xy', the additional degree
of freedom in the form of exy is added to the displacement.
The displacement functions are geometrically syMrnetri-
cal and include the constant strain and rigid body modes for
the flat plate. It is obvious that the assumed displacement
fields for u, v and w are not of the same order. Whereas those
for u and v are linear, as stated before, w displacement is
cubic. If the displacement fields of u and v were of the same
order as that of w, each node would have 12 degrees of freedom
thereby having a total of 48 der,rees per element. Besides
this, the linear edge member elements would have to be given
the same order of stiffness JT!atrix. The additional degrees of
freedom would involve more computational work and this effort
could only be justifi~d if good results, without sacrificing
the required degree of accuracy, could be attained with a fewer
number of elements.
Any combination of displacements which can be accoinp-
lished ~ithout straining the structure are called rigid body
displacement modes. These displacement modes can be easily
-36·
z • b xy 3-8
3-9b
= u ,y + v
tX
3-9c
l(
y • -w
,yy 3-lOb
- 38-
21<
xy • -2w ,xy 3-lOc
Strain Energy
The strain energy for a typical element consists of
two parts: the membrane and the bendinf strain energy.
u • ub + um 3-11
3-12
0 3-13
0 0 y :Jl:;
4 s 6
+ Exyt[u2~Y + v2,x + Zu,yv,x]
7 8
c
- 4ExytCA!) w [u, 1 + v,x]
9
c z 2
+ 4Exyt(AB) w } dxdy 3-14
- 39-
1
b a
Ub 11: ! : : p1x(-lt,xx)+My(-w,yy)+2r·xy(-w,xy)]dxdy 3-15
rT 0 0 Dxy -2w,xy
xy
The bending strain energy is given by 38 t
C. FLAT ELE~"ENT
[K)m 0
-
v
[KJ:flat • ----·-- ·-----·· 3-21
0 [K]bb w
1
4 3-23
+ n Dy)
b4
where,
3-24
D. EDGE tfEFBEP.
1 8 2 2 2 p z 2
Ubb • z0 I [Eiy'"'xx+Elzv'xx+GJe,x+..,rbe'xx+EAbu,x]dx 3-25
3-26a
a = [-xa e2 + x
c1 - -)e
a 1
] 3-26d
Using Eqs. 3-25 and 3-5, one can obtain the conventional beam
stiffness matrix.
However, the warpine restraint is of practical im-
portance, particularly for thin-walled open sections. To in-
clude the warpinr effect, the displaceinent field for e is
assumed to be of the same form as that of v and w58 .
+ a ( x3 - ax 2) eAZ
' ] 3-27a
3-Z7b
beam.
3-28
SA .-
Ab
a 8BY •
4EI.Y
a
.
> 5Bz =
4EI z
a
3-32
12E Il': 12EIZ
5sY • a
5
' sz
• a ., sT = -aC:J
These spring constants are added alonp the main dia-
gonal elements of the master stiffness matrix. This idealiza-
tion is not alway', satisfactory. In Fig. 4.1, a tension bar
connectinr, the lower corners of the saddle shaped hypars are
replaced by four springs in the u and v directions, two at each
corners f and b. This idealization eliminates the interaction
between the nodes f and b. The validity of this approximation
can only be assessed by engineering judgeMent.
3-33
Figs. 3.lSa and 3.lSb show two structures and also give the
-49-
zx
ox' • -1- 0
~x Rx
1 zy
OY' = 0
~ .A 3-34
·Z x .. z y 1
oz' •
l+Z x2+z y2 2+z 2 j+z2+z2
l+Z x y x y
where
and Zy • c x
AB o·
As discussed in connection with the shallow shell
assumptions, the angle between the generators OX' and OY' is
not equal to 90°. Hence a new set of mutually orthogonal axes
OZ, OX, and OY are obtained, where OZ coincides with OZ'. The
procedure for obtaining the direction cosines of ox, OY and oz
is a simple application of the three dimensional solid geometry
( F i g . 3 • 1 7b) •
The transformation matrix for each node can be repre-
sented as
0
3-35
0
tx ty 1:z:
p,] 1 D mx my mz 0 3-36
nx ny nz 0
III.4 LOADING
A uniformly distributed load acting on a rectangular
clement can be replaced by statically equivalent loads of
equal intensity acting at each nodal point. This procedure is
acceptable if the size of the eleMent is small.
The alternative approach knowl' as the work equivalent
load is based on the equivalence of energy. The nodal forces
are so assigned that during any virtual displaceJ'lent the work
done by these forces is equal to the correspondinr work done
by the actual distributed load.
The work equivalent nodal loads for the rectangular
element, with unit normal load, are given 'below:
Boundary ;..
Condit ions u v w ex xy
Hinge 0 0 0 0
Knife-edge 0 f} 0
Fixed 0 0 () 0 0 0
Free
Symmetric 0 0 0
tivc displaccTncnt bett·1een the deck and the edge member, normal
to the boundary (Fig. 3.20b). This type of a connection can
result because of an oversized hole, loosely connected screws,
or due to tearing of the deck. Depending upon the continuity
achieved between t!1e deck and the edge members, different
values of fixity coefficients are used. Tr and TH represent
the torsional and the horizontal fixity coefficients, respect-
ively.
In the case of the moment-free deck to edge Jl\ember
connection, Tp = O. The edre merrber stiffness matrix is modi-
fied by multiplying the columns and rows correspondin~ to the
twisting degree of freedoms (ex and exy) by TF.
The problem is further com~licated by the eccentric
I I I. 6 SOV:'TIONS OF EQPATIONS
[K] [T]R [fi] by [Tl T9 . transforl'!'s the forces to the shear center
or the centroid of the beam.
Because of the mathematical idealization, certain
difficulties are encountered. A beam with an eccentricity in
the z-direction is shown in Fig. 3.23. The variation of t~e
axial forces is shown in Fig. 3.23b and 3.23c. Since the forces
arc balanced at point 0, the axial force also contributes to
the equilibrium of the moments at point 0. This results in the
inequality of the moments along the axis of the beam PQ. The
problem becomes particularly critical in the case of rapidly
changing axial force and a deck with strong bendin! riRidity
(e.g., concrete hypars). No suitable solution is found for
this prohlem as of this moment. In the ab~cnce of definite
guidelines, the deflected shape of the structure should be used
to decide the sign of the mnm~~t.
IV. 1 INTrrODUCTION
·61-
riridity factor a, rise to span ratio, etc. The effects of
these parameters are further discussed in length in Chapter V.
t~."" c~nter and 1'.he diagt..n.:.:1 linr;s '" i and ob respectively. The
central deflec·... ion obtained by method 'b' for both structures
is on the higher side, as compared with the one obtained by
-64-
method 'a'. For St rue. 'l', the difference in the central de-
flcc tion is about O.Bt whereas for Struc. '2', the d~ffercncc
brella shell (Fig. 4.2), provides a more critical test for the
comparison of the different f!1ethods.
It was not practical to compare methods 'a' and 'b'
for all the examples, ther~fnre only a selected number of struc-
tu;-cs were chcsen for comr~ .. : son (Stru'.:.. '6' and ''J' were
used). St rue. '0 is a !;1~'<'." J l scale ccncr~te f'lode 1. In this
structure, the stiffening = ~ge members are located eccentri-
cally, on top of the shell. The idealized edge members are
considered eccentric only in the i !~rection (see Figs. 4.2
and 4.5).
Struc. '9' is also an umbrella shell hypar with 28-G
double layered standard corrugated decks placed perpendicular
to each other. Here the edge member is connected eccentri-
cally to the deck with the deck on top. In the case of Struc.
'6' there is full fixity between the edge member and the shell,
whereas in the case of Struc. '9', the connection between the
-65-
computational work.
The strain-displacement relationships for the curved
. c.
e 1cmen t arc d epen d ent on 1 y on t h c twist curvature AB 1rrcspect1ve
.
of the shape of the actual structure. To explain this further,
consider only the quadrant oabc of a structure of Type I, Fig.
4.1. One could build two cantilever hypars from this quadrant.
The first structure would have edges oa and oc fixed whereas
edges ab and be would be free. In the second structure, the
fixed and the free edges would be interchanged. If both these
structures are subjected to the same loadin,i conditions, method
'a' would give identical deflections and absolute values of the
stresses.
The solutions by ~ethod 'a' for Strucs. 'l' and '2'
did not show any advantage of using a 16-term displacement
function for w-displacernent, which ensures the slope compati-
bility normal to the boundaries of the adjoining element as
against the non-compatible 12-term polynomial used in ~.ef. 19.
The solution obtained for Struc. •2r the inclu-
~ith
'-"' ~·s
sion of complete rigid body rnodes 20 and that obtained
~~"\
without the inclusion, did not show ~uch of a differ-
ence (Fir. 4.6). To study t~e effects of inclusion of rigid
body modes further and also to evaluate the differences in the
Even though part of the vertical load near the column is car-
ried by the concrete shell, in order to satisfy the static
equilibrium for the vertical load it appears that the experi-
mentally ~easured forces in the compression rib are quite low.
The axial and the bending stresses are ~easured along
the diagonal ob at an angle of 45° with the x and y axes (Fig.
4.15). The measured axial stresses show excellent agreement
with the analytically calculated value of 72 psi. An important
point to note is that the calculated and the experimental values
are about 34% higher than those given by the ~embrane theory at
a load of 40.9 psi, the reasons for which are not readily
apparent. The values of bending stresses are very low and are
not compared here. The variation of the shearing force is
plotted all over the shell for both the structures. Though
there are minor differences in the shape of variation of shear·
ing forces, two important observations can be made. The values
of the shearing forces over a substantial portion of the shell,
are larger than those given by the membrane theory. The shear·
ing force near the coluPm is nearly twice as large as that
given by the membrane theory. This sudden increase in the
shearing force clearly indicates that the shell participates
in tr.:nsmitting a certain portion of the vertical load. The
same behavior is also noted in Struc. 'S'.
Struc. '8' is the same as Struc. '6' but it is sub-
jected to an unsynunetrical load (Fig. 4.20), where half of the
structure is loaded uniformly. Only half the structure along
the line cf (Fir. 4.20) is analyzed using 16 elements in each
quadrant. The statically equivalent load is used in one
quadrant. The central column is idealized by means of con-
centrated elastic springs as given in Chapter III, Section
111.2.F. The comparison between the theoretical and the ex-
perimental results is shown in Fig. 4.20. The deflection pro-
files appear to be quite reasonable though the magnitudes of
the deflections ob and ~e are 30·40t on the lower side of the
values obtained experimentally. A static chect for the ver-
tical load is satisfied at the center column though a dis-
crepancy in the overturninp moment is noted.
A highly irregular pattern of axial forces and moments
is obtained which unfortunately could not he verified properly
because of the difficulties encountered during the experiment.
A better solution can be obtained by using a finer grid (64
elements in a quadrant) and also by using work equivalent loads.
It was not possible to check the imp1.ovement in the solution
because of a limited comnuter core capacity. The exa~ple how-
ever, clearly showed that the theory can solve unsymmetrical
loading conditions such as wind load, etc., and can satisfac-
torily predict the overall behavior of the shell.
The corner deflection 6b in Struc. '8' is nearly
three times as large as that obtained for the uniformly loaded
Struc. '6'. The increase in deflections in the loaded quadrant
is mainly due to the twistinp of the shell about the line ah.
C. m.raRELLA SHELLS WITH STANDARD CORRUGATED DECKS
Strucs. '9' and '10' are considered to have flexible edge wem-
bers whereas Strucs. '11', '12' and '13' are considered to have
very stiff edge members. The edge ~embers used in Strucs. '11',
'12' and '13' are 4.37 times stiffer axially and 236 times
stiffer flexurally as compared with the edge members used in
Strucs. '9' and '10'. This large difference particularly in
the bending stiffness alters the behavior of the umbrella
shells.
C .1. INVERTED UMBRELLA SHELL l'TITH FLEXIBLE
EDGE T!f~'BERS
ber ab.
From the consideration of the maynitude of the in-
plane shear and its eccentric transfer, Struc. '11' represents
an extreme case. As shown in Fig. 4.32, the deflection nro-
file along the diagonal ob re~ains practically unaltered for
both hotmdaTy conditions for Struc. '11'. Because of the very
high in-plane rigidity of the 3" diameter pipe, the value· of
the in-plane shear developed at the junction a (Fiy. 4.29) is
quite large (742 lbs.). Thoufh small in ~agnitude, the in-
crease in the compression rib deflection c5c and 6a is almost
200t. The increase in deflection exceeded that which would
have been obtained by considering the edge members oa and oc as
cantilevers, acted upon by the eccentric shears at points a
and c respectively. A small increase in the deflection 6b is
noted and it ~ust be pointed out that the transfer of the
eccentric force also exists at the junction of the tension mem·
bers but it is of minor i~portance.
However, the rnaxim·1J!'I values of the shear force arc ahout 5-10\
lower for Struc. '13'. The only noticable difference for the
variation of the shear force for sinrle and double deck struc-
tures is that, in the case of the former structure, the 1"1axi-
mum value of the shearing force does not exceed the shearin~
to note that the results for the deflections of the deck are
~ .. \I\\.
quite close to t~ose reported in ~ith the gtiff JT1em-
bers, as those usec in St rues. '11', '12' and '13', the def lee-
tions along the free boundaries are small and therefore the
behavior of the shell is quite close to that of an edge-sup-
ported hypar for which, as pointed out earlier, JT1ethods 'a' and
'b' give the same results.
The salient features differentiatin1! the bef\avior of
the hypar with very stiff edge meMbers (St rues. '11', '12' and
'13') and the behavior of tl':c hypars \-ri th very flexible edge
- 91 -
members (Strucs. •0 1
and 1
10 1 ) are further discussed in
dPtail in V. The effect of the edge member weight
Chapt~r
also
on the behavior of hypars is discussed in Chapter v.
V.l INTRODUCTION
C. NUMBER OF DECKS
V. 4. BOUNDARY CONDITI01'1S
the full and zero in-plane fixity (TH) between the edge JTtembers
and the deck (Figs. 4. 30 - 4. :s 2) • Though the results are very
-l'lO-
V. 5. LOADDJG
for the behavior of the shells. The higher the value of this
constant, the more dominant is the membrane action. As dis-
-103-
VI. 1. INTRODUCTION
-104-
results in tl c redistribution of the total stiffness of the
1
shell. The theory used in thi5 chapter does not account for
the local buckling and therefore the effect of local buckling
on the shell stiffness cannot be predicted. The local buckling
can be prevented by choosing proper thickness to width ratio
for each individual plate elemont.
(2) neck Buckltne - In this mode of buckling, the edge
beams re1l'ain stable whereas tho deck, used as a shell, buckles
as a unit. To understand the deck buckling, consider the
umbrella shell in Qip. 4.2. T~e deck acts primarily as a com-
pression arch between the points o and h, and therefore it can
bucY.le along the diagonal ob; but the shell edge members re-
main stable.
(3) Qveral! Buckling - The shell and the edge members
buckle as a whole t1nit. One c-.n iroagine an umbrella shell,
foldinr down as an umbrella turned inside out. Overall
buckling could occur either simulataneously with the deck
buckling or it can happen after the deck has buckled.
According to a siwpl:lfied analysis by Parker 19 , the
possibility of overall bucklinp. for the practical size of edge
memhers is very remote. Very high values of deflections and
stresses for both edge mewhers and the deck will indicate the
possibility of overall bucklinri. T~e conclu~ion that overall
bucklin~ is very unlikely is further verified by Struc. '9'
{Table IV-1) tested at Cornell , where l" diameter standard
pipes were used as edge ~ewber1. The resulting structure was
too flexible to be used in practice. In spite of excessive
-106-
deformations (nearly one l1alf the rise of 14. 4':), the struc-
ture did not show any tendency of overall bucl<linp, though the
deck buckled.
The present study was primarily concerned with deck
buctling. However, the overall instability due to the buckling
of the edge members can also be predicted from the load deflec-
tion curve. The assumptions used during the analysis and the
li~itations of the theory are as follows:
(1) A linearized stability analysis was carried out to
predict the bifurcation point of buckling 51166 • The prebuck-
linp. deformations were within the limits of small deflection
theory.
(2) No attempt was Made to predict Dost-buckling behavior
or the post-buckliny stren~th. To be able to predict the
post-buc~lin~ behavior, one needs to retain the higher order
strain terms in the strain displacement relationships and have
. }
h ig1er or d er matrices
. 67 . It is extremely difficult to formu·
late these watrices explicitly and one has to resort to
numerical integration. The non-linear equations can be solved
by the use of methods such as Newton-Raphson scheme 25 , ener~y
.
searc h tee h nique 67 , c t c.
a2v.T
[ N] i j c: afl . '~ ~ . 6- I
l J
The coefficients of t 11e increwental matrix (N], depend only
upon the geometrical parameters of an element, such as its
len~th. The incremental matrix is identical for both ortho-
tropic and isotropic cases.
For constant values of Nx, NY and Hxy the increynen tal
-108-
matrix for a shallow shell hypc.r element and that for flat
plate elements are identical. The only difference is the ~anner
where '- is the eigenvalue which depends upon the applied state
of me~brane stress e.g. for an uniaxially co~pressed plate,
along the x-direction it will give eifenvalues corresponding
-109-
6-11
-n11 n1~
--------- _;l 6-12
[N]global = w
n21 nzz
VI. 5. DE°TEI'J'INATION OF
- -TI-IE BL'"Cl\LING- _,,..,.,__.
WAD
storage, the author could not use the direct eigenvalue approach.
Instead, a linearized load incrementation Jnethod had to be used.
In the load incrementation Jnethod, as used by the
author, the assumption is made that the in-plane forces Nx' Hy
and I~xy are constant during an incrcrnental step and are equal
in magnitude to the value at the end of each step. The proce-
dure of the solution can be demonstrated by the use of Fig. 6.1.
In the incremental step I, only linear analysis is
carried out by solving the linear part of the equation assum-
ine [N] as a null matrix.
[~] • [K]-l [P] n·l3
From the kno~m values of displacement vector [A], corresponding
in-plane forces Nx, try and Nxy are calculated and. the incre-
mental matrix [N] is formed. The effective stiffness matrix
[K]eff is used in iterative cycle II.
The iterative cycles are continued till convergence
is obtained for the nodal displacements and consequently the
incremental matrix [N] is consistent with the deforJnations.
It is found that for swall incremental loads, convergence of
displacements is obtained d thin three or four cycles.
1 An
incremental load op is applied on the l'lOdified effective stiff·
ness matrix and the increase in the displacef"lents 6 and A is
calculated by finding new values of the in-plane forces at the
end of the step II by the iteration as clescribed. before. The
analysis is continued by applying the increment of the load 6p
on the previously determined effective stiffness ~atrix.
pointed out in Chapter IV, the curved and flat elements, for
hypars with all edFes supported give very close results for
the linear elastic analysis. It appears that the flat ele-
ments give very conserv:•tive results for the buckling of single
decks. As pointed out in connection with Eq. 6-12 for the flat
elements, the transformation of the incremental matrix froJF1
local to the ~lobal axes results in the modification of the
flexural as well as membrane stiffnesses. Since the in-plane
membrane stiffness for a corrugated deck is very small in the
weak direction, a premature buckling could possibly be trig-
gered by the reduction in the me~brane stiffness. As against
this, in the curved element formulation because of the
assumptions used in the analysis, the membrane stiffness is not
modified.
For a double deck or an isotropic deck, because of
hi!~h in-plane rigidity in both directions, hoth methods pre-
dict the buckling loads in the same range. As pointed out in
Chapter IV, the bending action is very dominant in t~e flat
portion of the shell. But in the buckling analysis, the curva-
ture of the hypar in the center of the quadrant may be more
critical. This fact occurs even more in the case of a single
deck and therefore the curved element, which accounts better
for this curvature effect, probably predicts a hi~her load.
-118-
2E
each of
This solution was poslible becausel\..the t'fo differential
one .:."' Cl4.t. I,... e,. ~~c--\' 1 ~ '.
eq11ations contained only elastic constant:.. • F
in one equation and D in the other. Thus the sine terms could
drop out,
In the case of orhbotropic shells these eq11ations
contain several elastic constants and therefore
the sine terms would not drop out. This means that
~Wvl-t
one cannot use title shallow shell equations to obtain reasonable
buckling load. It may be possible to find • very simple
-\ff-
"' "'"
\j =- l ff [t>)( J.S:
0 0
T 2 D , ,L,)>'l< J.S;:
~1)
+ I)'?! ~ ~ -t- 4 D'l-'Q J.f ~ ,_,1
-t- 4 r,; e.(f "C7Pl~ ,.w-1. + 2 i.i ~a
l. )..S)<. µr-i ~ d.,,. ~
The selection of an assumed buckling shape
is ~ A'4'Uilt"
requires very careful attention, A double sine series ·
h>
use because of the complexity' of the resulting arithmetic.
shape:
shear diaphragms.
integration yields: ~
~ab ~
v s ;01 + Q2 + Q3)
where ~ \-:. ~tt+2 D, a'\..~T 2.D l a<\-.s.1.+ D'i \tt' D'"a&.4 ~
+ 2~'-a a,,. 8,_s-i.. -t 4- o ,,.'-a a<L s i. + c'
Qt. -= 4 D "a Cl'"\.. la"\. s.' ;- 4- D~'o Q,..a'"
Q l =-2 \j Y-<o ~:i.. s a = "l"/f'i la" 'ifa d =+G~~ t
Since V represents the change of potintial enefgy under B
a deflection w, the condition V =0 corresponds to the
buckling load. Thus
Q \ +- Q'- '\-- Q J ~0
or
2..4
The analysis of the single-layer Sage structure (~o.t'l)
gave 61 psf for bee buckling load as compared with the
experimental value of about 75 psf and the finite element
analysis of about 60 psf
--
using curved elements. Minimum
s
occurlul for n~ 6 and the buckles formml at an angle of 11.3
(,::..·r.: ~.lZ)
dggrees with the y axi~ The buckling load was not very
sensitive -\.a n anJ. therefore the prediction
I
of the numbe~of buckles may not be accurate.
VH.1
INTRODUCTION
determine the shear rigidity G' of the decking used for the hypar
models. Twelve specimens were 6' x 6' in plan, three were 5 1 x 5'
f lrst one indicating the rise/span ratio, the second the number of
duplicate. For example, for test no. 512, "5" indicates a 1/5
36
\l~~s u, It 115 ~t\ Tc:..W,.
\V- \ . .
1:~
Figul:-8 • They are referred to as "medium scale inverted umbrellas
with very stiff edge beams".
The model was 2' by 2' in plan with a 3" rise. The edge
members were maGe of 3/16 11 o.o. and 0.014 " thick brass tubes,
two ~ere used for the interior compression beams. Two la~ers
~, .... od-..(
ft picture of ls shown
,.\.
in Figure 3t4.
beams - The model was 12' x 12' In plan wt th a 14.4" rise) ~+rv<....\o~
V\\Jv..~~r ct , T0\.9\c. \\1-1. ._
A picture
of it ls shown In Figure -r.s.
The tests are described in the follo,wing sections.
'Ill.
JilR .2 FLAT SHEJ\R TESTS
1) INTRODUCTION
A series of flat shear tests were conducted in order to de-
4S
termine the shear rigidity G' of corrugated steel decking. Luttre14
The formula for G' developed by Luttrell accounts for the de-
f lectlon due to shear alone by subtracting from the measured de-
flectlon the bending deflection of the cantilevered structure due
to axial deformation of the edge members. The value of G' obtained
from a given test can be applied to the following case: Any set-up
For all tests except two the edge members were light gage
channels. For the other two tests. the frame was made up of tubing.
- \ ~' -
\
3: DEClSlNG
decking; including 24, 26, and 28 gage. Either one layer or two
layers, with the second layer placed with the corrugations perpen-
The small scale test had 2 mil corrugated steel foll. This
4 CONNECTIONS
For the medium scale tests, ii8 x %" self-tapping screws were
used to connect the decking to the edge members. For double layered
decking, the second layer was attached to the f lrst layer around
the perimeter only with iiB x %'; self-tapping screws. The spacing
screws connecting the top layer to the bottom layer and the screws
connecting the bottom layer to the edge members was noted. Re-
ferrlng to Figure
1.' one possibility ls shown in (a) where the
1
The other possibility ls shown in (b) where the shear flow now
layer.
For the small scale test, the top layer was soldered to the
the bottom layer was soldered to the edge members at every point
Si LOADING APPARATUS
was pinned. Steel bricks were placed on the loaded edge to prevent
out-of-plane warping.
The small scale traDMt was attached to a wooden frame with
Olsen hydraulic testing machine and load was applied by the machine.
6; DEFLECTI<l'i 1'£ASURE?-£NT
At first, .0001n gages were tried but it was discovered that they
are too sensitive because the readings obtained from them were
If possible the stem of the gage was placed In contact with the
curve was used to obtain the slope, The customary testing procedure
and of a a ""·•
G'/Gnt for each test are presented in Table - ·• The
-y.1(
load-deflection curve for test #14 is shown in Figure •
,1 INTRODUCTION
light gage steel hypars were possible, then it would appear that
tests was the rise, i.e. curvature. Three different rise/span ratios
were employed, 1/8, 1/5, and 1/3# one set for single layer decking
and the other for double layer. Two tests, one a duplicate of the
other, were carried out for each rise/span ratio and number of
plained in Section
\Jn.'·•
The edge beam frames with supports were the same for all the
tests. They consisted of 3" o.D. x ~" wall thickness cold rolled
ratio for each specimen. The diagonally opposite corners, two low
and two high, along with the midpoints of each side were rigidly
"1-2
supported In the vertical direction by steel bricks (Figure ).
This support system together with the fact that the tubes have a
sheets with 2• cover. Three sheets were used for each layer, with
the middle sheet cut to flt the 5' width. Single layered decking
spacing. For the double layered decks, the top layer was fastened
to the bottom layer, around the perimeter only, with 118 x lz" self-
tapping screws @ 8" spacing. The two seams of the top layer were
by wooden sides and screeded to a uniform depth for each 400 lb.
test, a partial load covering an 8" x 12" area was applied at the
center of the same decking. These tests were designated with a "C".
In addition, for each one of the two duplicate models with double
and then the partial loading on an 8 11 x 12" area. These tests were
measured at the center of the decking with SR-4 strain gages with
l" gage length. The strain gages were placed in a valley on the
top side and on the adjacent crest on the bottom side, and in the
mid-length of the tie bar, top and bottom, on some of the models.
The purpose of the gages on the decking was to determine the axial
and bending stresses at the center of the deck and on the tie bar
was to determine the axial force in it.
1 INTRODUCTION
( ~<. ~t \\ J \ t J l ~ / T~~, l V ~I)
Three medium scale modelsA...were designed and tested with the
main purpose being to check the theory for the buckling ot ortho-
The same edge beam frame was used tor all three tests. It
beams, with a 14.4" rise (1/5 rise/span ratio). Each interior edge
beam consisted of two tubular members side by side (spot welded to-
The third model had two layers of 28 gage standard corrugated steel
decking.
For both of the models with one layer of decking, #14 self-
seams together. Three panels, each with 2' cover, were used per
quadrant. For the model with two layers of decking, the bottom
The top layer was fastened to the bottom layer, around the perimeter
only, with /i8 x ~" self-tapping screws @ 8" spacing. The seams for
the top layer were fastened together exactly the same as those in
For all three models, uniform normal air pressure was applied
placed between the floor and the hypar in the inverted position.
measured with SR-4 gages with 1 11 gage length and those on the edge
beams with SR-4 gages with ls" gage length. The vertical deflections,
- \ ~Q-
vu.s
INVERTED mmRELLAS WITH VERY FlEXIBlE EDGE BEAMS
1 ·, INTRODUCTION
J
theory.
rise/span ratio).
The edge beam frame was made up of 3/16" O.D. x .014" wall
each point of contact around the perimeter. The decking was soldered
(16 per quadrant) hung from strings passing through very small
The dlal gages were positioned such that the stems did not touch
the shell. To take a deflection reading, the ste~ of the gage was
pressed until contact with the deck was made; avoiding the
Figure
The test set-up and procedure for this model were almost the
The edge beams were l" nominal diameter standard weight black
The stems of the gages were attached to the shell by means of thread
\)\l. '-
DISCUSSION OF EXPERIMENTAL RESULTS
G'
.1) SHEAR RIGIDITY G' OF STANDARD CORRUGATED DECKS
Comparison of the results for the flat shear tests, which are
\J\\-t
presented in Table , reveals that the variable having the
has the predominant effect whereas the spacing along the corru-
this dimension the larger the shear rigidity. From the results
5' x 5', G' ts less than that for 6' x 6' test no. 5 but slightly
more than that for 6' x 6' test no, 6. Thus, lt appears that
variation in stze,
The spacing of seam connections and number of seams have no
thickness. For example, compare the results for tests no. 5 and 7,
transverse to each other. In each case the top layer was connected
to the bottom layer along the edges, rather than being connected
The reason for this ls that any shear deformation which occurs ln
the bottom layer along the edge members occurs also ln connections
for the top layer which ln turn reduces the effective shear stiff-
ness of the top layer below that obtained by connecting the top
Vil-I
layer directly to the edge beams. From Table ~ tests no. 10 and
and 9 show that the shear rigidity for the double layer tests ts
less than twice that for the corresponding single layer tests. It
only a 33% increase in G' over that of the single layer~ although
,.
two layers were such as to create the condition shown in Figure
~(•)
) then the shear rigidity should be larger than for the case
, , , (q)
shown ln Figure •
'1:-J
Figure shows the load-deflection curve for test no, 14,
The linear portion extends to about 1800 lbs, load which corre-
sponds to a shear Nxy • 25 lbs. per inch. Beyond this point~ the
~ 1/5 is somewhat (20 - 303) less than that of the same flat deck.
\Jl\. '2.
However, as mentioned briefly in Section. , this Information
t·B
Two factors were omitted in the theory used to obtain Figures
-r.~
and , the inclusion of which indicate that the shear rigidity
obtained from a flat shear test may be valid for the hypar shell.
u and v. The other was the spreading of the low corners of the
saddle shaped models. Even though a tie bar connected the low
in its elongation. Duplicate tests no. 521 and 522 were chosen to
as shown In Figure
,.,
ments are included then the resulting a ls .046 Instead of .042
ness of the tie bar Into the analysts then the resulting a ls
approximately .o5 which ls the same as that obtained from the flat
shear test. These results for only one example are not meant to
Table
v.- 42.presents the maximum deflections at 40 psf for the
ls more than half that of the corresponding model with one layer
the single layered deck is more than half that of the double
layered deck.
\J\\ - t.l'•)
Table shows the effect of interconnecting two layers
all over and not just around the perimeter. Test numbers with an
maximum def lectlon due to a 100 lb. load on a single layered deck
with that due to a 200 lb. load on a double layered deck reveals
few beam strips whereas on the double layer it ts spread out and
thus carried by more beam strips in each layer. For all the models
air pressures ln the rubberized canvas bags. The bags were not
entirely air tight so that wtequal rates of leakage from the bags
pressures.
vary from 6200 to 10200 psi for the 28 gage single layer, from
4800 to 5700 psi for the 24 gage single layer, and from 5100 to
10200 psi In the bottom layer of the 28 gage double layer decking.
The wide range of these values for each model indicates a con-
the free corners is much larger than that at the center of each
quadrant.
"!, . HYPAR STRESSES
Q.)
SADDLE SHAPE SUPPORTED ALL AROUND
\J\\- 2.
The experimental stresses presented ln Table - reveal the
ditference tn structural action between a very flat hypar (rlse/span
The bending stresses in the models with a 1/8 rise/span ratio are
much greater than those in the models with a 1/3 rise/span ratio.
with a 1/8 rise/span ratio is much less than that given by membrane
theory whereas the force in the tie bar for the models with a 1/3
rise/span ratio ts almost as large as that given by the membrane
theory. Thus, it is demonstrated experimentally that for very flat
hypers, shear stresses are much smaller than predicted by the mem-.
brane theory and that bending constitutes the major part of the
Mmbrane action.
Experimental edge beam axial and bending stresses as well as edge
beam axial stresses from the membrane theory are also shown In
, • 10(1') , . ll{lf)
1
,. rz fl,)
Figures , , and • It ls difficult to observe con-
ststent trends in the experimental axial stresses in the edge beams.
In some locations they are very close to those values given by the
the results for the three models do not give any Indication as to
teal shear loading to these edge beams. The bending stresses in the
interior edge beams for all three models are small although the
CiiAPT.~~ X
-120-
-\So-
INSr=RT
49. Wright, R. N., et. al., "BEF Analogy for Analysis of Box
Girders," Journal of StructuTal Division, ASCE, July 1968.
52. Cantin, P.• , and Clough, P., "A Curved, Cylindrical Shell
Finite .Element," AIAA Journal, Vol. 6, June 1968.
-\60-
A-1
Ax• fit;:
Pt s z2t ds
-r A-2
s 2
0
J y ds • I '
y
A-3
Prom Eq. A·S, [xh' the elastic extensional constant, for the
hat only,
= Eiot2 A-7
Exh I 1
y
l)r' is the IT'oment of inertia of the hat portion about the base
plate. By compatibility,
A-8
1.-9
A-11
Ax
vyx. vry A·l3
The shear rigidity for the closed deck can be considered equal
to that of the base plate which is usually directly connected
to the edge me!"bers. Even a thin plate interconnecting the
free hat portions will increase the effective shear rigidity of
the section. However. the cost of construction will have to be
taken into consideration in this approach.
A-lib Bending Properties
These properties are given in the Table A-I.
A-Ile Torsional Rigidity
Here again the values of Dxy and Dyx are not equal.
As shown in Fig. A 5 , the value of Dxy is negligibly small as
compared to that of Dyx· The value of Dyx can be obtained by
using Brendt's formula 46 •
4A 2
c A·l4
A-1·5
z
lay
_________. . . .__ x
z •
p
x
u !1x_ j y
(a)
r ~~x.I •
P --------1c:===========::::========::::j~~~-
(b)
I
p Fig. Az
x
t;T
p2
... B f tl c 2-1 p
y
fl·
p
a -1;~
J.
Rotation in
\"'-.an Ideali Zf::d z
System
Actual
Rotation
Pr.OPEPTY: DECKS
:·embrane
Io
Ex E y-r F. E
y
A
s H
F.y F
r F ,;
"xy v v
1o 1 Ax
vyx "rs· v r..y v
y
l
El vE o vE vE
i;-r
2 Ax
l·vyx"xy c L fl 1-v A l-v 2
y
a·: aG iG
Exy
Bending
Et 13 Et 3
1Et 3
DX !T:-·
0
i2~~1-v 2 ) 12(1-v 2)
n··y F.Iyd EI :,'ti Ft 3
12(1-v 2)
vEt 31
nl 12s " vEt 3
12(I-v 2 )
Et 3
!t GIO 0 24(1+v)
·xy
2
GIUs 4Ac Et 3
Dyx -r- a•. 24(1+v)
RrcdCrr)r,
APPENDIX B
STIFFNESS LATRICE$
TA:BLE B-I
e.g. •1 • !b i E xt
+ a E
3b xyt
-\t. -
TABLE B-II
22 b 2 27 a 12 1
- n- 36 1 33 ab 2
~it:,
1· - ~ """"'!
a Ilb"f a ·rra:
-·-·- 13 a 2 12 1 6 1 169 a2b
a, -
78 b
35 Ilb'J - 25 f:) - IT S -r
3 2a
11 b 2 13 a 2 1 3 143 2b2
Bs n;z - '°b! n rr 18 8
54 0 54 a 1~~ I 72 1
69 - 35 ~
- 3Sb3 --ZS ib 25 ab 81 ab
13b
2 27 a 12 1 6 1 - 2!7 abz
alO n~ llb! rr a IT a
13 a 2 12 1 6 1 217 2b
611
27 b
- !! ,. - Il"bJ rr r> !St> -r 8
a
Continued
- \11-
Continued
Coeff. DX Dy D·xy Dl cl
13 b 2 . 13 a 2
B12 - 70
..
;r 70b2 1
25 so
1 169 a2b2
36
54 b' 156 a 144 1 12 1
6 13
35~ - 353
b 25 ifi - rr ao 234 ab
2
6 14 - --
13 b
35 a3
78 a
35 b2
12 1
rr a
6 1
fS a
169 ab2
3
2 2 a2
8 1s --27 b
35 - 2
8 35 b3
12 1
rr r;
36 1
25 b 33 a 2b
i3T-- 11 a 2 l 3 _ 143 a2b2
6 16 - 70 ;r 35 b2 B IT 18
--r-
4 b 52 a 16 b 8 b 52 b3
8 11
35 ~ '31' o 2s a ff i 3a
11 h 2 11 a 2 1 61
8 1a - --
35 ~
a
- Eb! - IT so
121 a 2b2
- g-
2 b3 22 a 2 4 4 g a2b3
8 19 -rr;r 105 b 75 b 25 b 9
6 20
4T 18 a 16 b
fl a
8 b
- IT i 6 ab 3
- 35 a3 35 fi
3 --·-:-z-
621
2 b
35 8 2
13 a
- 105 b
4
75 b rs2 b - 13
9a
2b3
3 b3 9 a 4 b 2 b
8 22 35 8 3 35 F rr i 25 i 29 ab3
3 b
3 1:5 a 2 b b !l a2b3
823 - 70 ;z - ITff.fJ. 75 - 150 12
3 b3 4 b
624 - 35 ~
26
35 0
a - fla - rs2 ab 13 ab 3
3 b3 11 a 2
82s - 'f7J za ms- - b
75
b
- IT
11
6 8
2b3
52 b 4 a3 16 a 8 a 52 a3b
8 26 ~a n-b3 fl o IT o 3
22 b 2
3 4 a3b2
-ma- - rs;r
2 a 4 z~
6 IT a - -25 a 9
--21 Continued
Continued
Co cf f. D Dy D nl cl
x xy
8 2a 26 b 3 a3 4 a 2 a 3 ,)•
35 a J5b3 25 0 25 b
.'- 13 ,l
11 b 2 3 a3 a a .!! a3b2
S29 -ma- 70 b! ff IT 6
---
9 b 3 a3 4 a 2 a 9 a3b
~30 na- 3SbJ IT b IT o 2
-·-· 3 ar-- a
13 ·bz a l12l a3b2
a31 210 a 70 bi"
75 150 _,
18 b 4 a 16 a 8 a
'332 ~- 35 a 6 a 3b
35 ~ 25 0 Eb"
13 b 2 - 2 a~
4 2 13 a3b2
633 ma- Eb! 11! Cl. ~a 9
4 b3 4 a3
B34 ma- mo m16 ab mR ab 4 a3b3
~
2 b3 1 a3 4 2 a3b3
6 35 ma- no - m ab - fl! ab 3
1 b3 - 1 a3 1 1 ! a3b3
636 - 10 a 701> m ab 450 ab 4 i.
1 b3 2 a3 4 2 a3b3
8 37 - na- mo- - m ab - ITS ab - ---r
Cz = Exyt <AA£gB)
The stiffness matrix for the curved element can be
reduced to that of a flat plate by putting c•o; whereby c 1=o
and c •o. The schematic representation of these two matrices
2
is shown in Eqs. 3·20 and 3-21.
- \i4- -
TABLE B-III
7 a 3 7 b __! b2
el m es 20 a 69 20 8
13 30
7 1
- 1 b2 1 2
02 ITlf ab 06 iO ab 610 20 8 14 lPO ab
1 2
03 -21fl a2 67 30 a 0 11 m1 ab2
l 2 3 b
04 m1 a2b 68 -180 a b 8 12 20
SYMMETRIC
!ABLE B-IV
CURVED ELEMENT STIFFNESS MATkI! [K]
APPENDIX C
DECK STP.ESSES
c
l'1 xy • (u ' y + .r ' x - 2 Alf w) EX)' t
w,YY + [(·8x1-ox2•&x3•8x4)/2b+(-9xyl•axyz-exy3•exy4llo1 s
The moments per unit length can be obtained
i· r
x • - [D xw'xx + D1w,yy1 6
l'.1 • 1
'y -CD1"'•xx + Dyl"·r j yy]
I! .
b 1:x,Ny
! , xy
(Mx, My
-·-·-10-·-·- .....
b
! I
-- I
'L~3
l- 'f 'f
TABLE D-1
Coeff.
- Nx Ny Coeff. Nx Ny
-
78 b ' 78 a: . 27 b 78 a
411 ma ffib 4>13 rrr a - !ff 0
11 b 2 13 a ... 13 b 2 13
- •2 ma- m 4114 ma- 350 a
13 11 az
- ms- - m9 11 a 2
41-., - 350 b 4115
b
mo
11 2 11 2 13 2 11 2
2100 a 4>16 - 4200 b 2100 a
. 4> 4 2100 b
-----
78 b 27 a 2 h3 26
415 - 175 i mo- ~17 11s a 525 ab
11 b 2 11 2 11 2
,6 - ma- m9 a
tis - ~lQO b - 2110 a
13 az 1 3 11 2b
•1 - m13 b
m-o- 4>19 1050 b Il'1! 8
4>a
11 b2
-~
13 a2
4'zo
. rrfS b3a mab
s
. ~
- 27 b
- ma- 27 a
- fill> 1 b3 - 13 a2b
99 •21 m1f 3ITO
13 b 2 9 3 b3
- 3 ab
4110 3so a ..
1150 a 4>22 noa- 'Tm)
-
13 a 2
•11 - 700 b
9
- 350 b 4>23
1 3
- Itf olJ b
IS
t2l>oo a b
2
3 ... ~
•12
. 13 b2
mlf
13 a2
- 42lrn"
- 4>z4 - ~ J.·.:. - ~ab
Continued
- 'to--
Continued
Coeff. N H Coeff. N N
x y x y
--r:-3 11 a2b "' 2 a'
-- ?: L _ _=_ I41fO b rnra- IP•
,) 17s ab - 1750
2 a3 13 2 1 3
~26 m26 ab
~o 4>33 3150 ab - 1050 a
11 alJ2 1 3 2 ab3 2 a3b
4>27 InS° . ION a ~34 rsn rn"!
-·-·-.-· 3 a 3-
~28
13 ,,.1
- 1050 ::...;.) 3so b 4>35 - m! ab3 - ~ a 3b
11 2 '.{ a3 1 ab3 1 a 3b
~29 6300 ab ma 4>36 mo 42~0
-
3 a 1 3 1 3
~30 - 100
3 ab
mo- ~37 1050 ab - 3150 a b
13 bt 1 a3
cfi31 ·12600 a TI1fO
}]_~LI D-II
ab
so
------·- - --------·-·-··------
b
IO
·--~- ·--····--------
wl 8x1 9 w2 8x2 0 2 9xy2 0 e~., e>:y:; w . 6x4
-•1 '2
•11
eil
4>3 <1>4
-----
•1s 4>19
xx:1
4>5
¢if
<I> 6
l
4>7
-ct> 8
<l>s
4>21
"'3
4>9
-•10
$10
x3
4>2 2
4>11
•12
ri2-
4>2 3
•13
-4>14
'?
4>14
924
el'.:4
~15
e•I4
•16
?iz s
Wl
e;1
•20 •16
•26 •21
cf>34
-·-· 7
8
's
-4i21
•2s
•29
4>z9
$35
-•11
•12
•12
-+23
4>3(\
-~:;1
4>31
.36
•1s
-¢>16
-•16
cfi2 5
4>32
- <1>3 3
4>33
4>3 7
er1
6
x:rl
•1 4>2
•11
- 4' 3
·•1s -·
·•19
4 •13
-~14
$14
<1>2 4
-411s
·•16
-~16
-•:s
cf>g
-$10
•10
4iz2
-•11
·•12
-•12
"'.~23
Wz
8x2
•2c, <1>2 7 -4>15 •16 4'32 4>33 •11 -•12 4>10 4>31 e:rz
--·
+34 •16 ·cf>2s -+33 4'37 -4>12 •23 -4'31 .36 ijXI2
svn·ETRIC
4>1
-·•11
2 -<1> 3
•1s
<I> 4
-.P19
•s -·
-·_,•1 -·
6
•s6
•20
-cfi 7
8 -•21
W3
6x3
\
cfl26 -•21
+34 8
cfls
cfl21
ct>zs
-+z9
-<1>29
<1>3 s
el3
eXI3
-
DC
•..
•1 -4>z 4'3 ... 94 W4
TABLE D-IJI
(N]
1
t•ATRIX
4>12 <l>p7
t26
-4>1 s
- ,7.1
; ,,6x4
r4
4>34 exr4
. - ·---···· ---·-8 ----
~
8 ex3__ ~~ exy3
. e1 dx1 e,1 ex2 ey2
"'1 xyl w2 W3 \•'
·4 ex4 0,4. exy4
..,·.
..
-e .. .... -8 ...) 0 e2 -e 1 83 -84 -e 2 - 0 -n
.... 84 ez w·
1
0 -62 0 e3 ~ e2 0 -e 3 68 -e 2 -6 6 0 0 ez 06 e
xl
0 0 0 02 0 -e s 94 -e 2 e7 e5 -e 4 e2 0 0 e·y t
0 e2 0 es 0 -e 2 96 -e s -e 9 62 -e 6 0 0 e,1-!_
-a 1 0 0 -e 2 0 0 e4 e2 el -e 3 -a 4 -e 2 w·
2
(J _n
- ,,. \J () 0 ez e6 83 -e 8 -e 2 -eg e72
0 lj -0
4 82 0 0 64 -0 2 -e 7 -6
5 eyz
0 82 -e 6 0 0 -e ... es e(}
2 96 exy2
el 0 0 -e 2
0 e3 0 02 W3
('.; -ez 0 -e 3 0 02 02 6x3
0 0 0 e2 0 es ey3 ,.\
0 e2 0 -es 0 6xy3 .,e
-e 1 0 0 -e 2 W4 "'
SYt-!f. iETP.I C
0 -6
z 0 ex4 '
0 (J
eI4·
.. 0 ~xy4
TABLE D-IV
[N] 2 PATP.IX
APPErlDIX E
STIFFNESS AND INCREt~NTAL PATRICES FOR A
FULLY COiiPATIBLE ECCF.NTRIC STIFFENE~
INTRODUCTION
In the case of light gage hypars the orthotropic deck
is connected at discrete points to the supportinp edge mell'lber.
Because of this discrete connection, there is a certain amount
of non-compatibility between the edge members and the deck. The
stiffness matrix used in Chapter III was developed by the use
of direct co-ordinate transformation. The co-ordinate trans-
formation assumes that the neutral plane of bendiny for the
original beam cross·section rel'f1ains unaltered.. l'Tith the type
of connections used for the light-gage hypars, the assumption
made above represents the tru~ behavior.
However, in the case of stiffeners which are rigidly
connected to a plate or a shell or a monolithically cast con-
crete beam, there is full coll'lpatibility between the strains at
the junction of the deck and the beam. In this compatible case,
a part of the deck also acts along with the stiffener (effective
width concept). The interaction of the deck and the stiffener
results in the adjustment of the neutral axis of the section.
To account for this change of neutral axis, the bending property
of the stiffener will have to be ~odified by arbitrarily assum-
ing the effective width of the deck actin~ along with the beam
(see discussion on Struc. '6' , Section IV- 51') • Depending upon
the relative stiffness of the plate, this kind of arbitrary
adjustment in the stiffener property will present a problem.
This arbitrariness is removed by formulatinp. the compatible
stiffener element.
STIFrNESS PATRIX
A typical eccentric member is shown in Fig. 3.10.
It is assumed that the mer.iber is uniform i.n size and its local
axes i, i and i, throug~ the shear center and the centroid, are
parallel to t~e global axes x, y and z. Let Uc be the average
axial deformation of the section measured at the centroid of
the section (C.G.) and V5 and ~·\ be the bendin rr. deformations
measured at the shear center (S.C.).
Assuminrr a rieid connection between the stiffener and
the reference structural node points, the displacements at the
shear center and the centroid can be expressed as follows:
'~--c = u
,, (aw) . y (av 1 E-la
ax ax
~
"'c c
"s = v - z a 5
E-lb
\'I
s =w + Yse E-lc
aw The total strain energy of the beam ele~ent can
where e • ~y·
be given as
a a21·r 2
~
EAb a 3U 2
-r'0 ( a~> dx I (
ub • + s) dx
0 7
2
Elz a a v·s 2 l
+ -2 r,J
a
I aidx + !2 Er I
a
o" 2dx E-2
+ 2 J C-::-rl
ax
c1.x -·
0 0
E-5
r-6
Bleich 56 .
2
.2E Z
• • ~ [ s Iy _+ rb
1
+ G
2
a J E-7
ac 7 pc ,..2E I~~]
I pc • Polar moment of inertia about the forced shear center o.
The only difference for both the cases is that the T-section
has rb • o whereas rb for the I-section is 1590 inch(. In
both the cases, for simply supported ends, the twist buckling
load was within <<11 in error, from the values given by the
classical solutions. A negligible error is observed in one
element solution. The buc~ling stress predicted with
2 and 4 number of elements showed an ~lmost insignificant im-
provemcnt.
z
t:.y = ·-Ze
t:.Z • Ye
{\ z
z
y
,___ _ _ _ _ _"""", Ay
y
Re-;; train--
,,' '
\.
,,.' ,.."'.")
a u zoo•· \
,,
\\
\\
~'-;/ \ \ .........,,,,. ,,,,. ...
,,,, "
,..-"'
~Tl2 Fig. E2
I 39-95
Twist Buckling
12 WP 40
- ,.... -
\
U. I 'JI w, a')(., I 0 (f I I 9 1'..:( I Lli. lI vl- W::z_ e 1(. :a- e(!-:a- e")(.aJ..
"
A'bE~
AbE.
0...
-
0 0 0
0...
0
-
-AbE.
G,..
0 0 0
-AbE l.,
0..
0
1 2E l
Oa
12EAi.'1'c.~
·-
-12E r.,, ls _,lA!,"'l'c.~ -~er.
a.j.
i-s 0 _,_,
~I U
o l-t'J.£Ai;fJ
l 1.1E1 1 l 5 i-(,EAi.'fc:!J-,Eiil-1.S
11.EI 01 1:z.E L
1
r, -C. El 0 y 'e.r'4'fs
u
0
'-12.E~"'l'c.~ -11E IO'/ -12Eii1's -(.E 101 'Eict"fs
c..- 1
~·
Cl) a.3 a..3 0..3 0.. ~ 0.. 'I. CL~
-·-- -----,:-;--
11u3 l's i!o.3 ,EI,_!s'3./J 1 i-11e11 Ys~ :,ui1 "fs/ ~~
+ 11E11i:/o: -(.ElJ~ HEl~l;/d
0
I I
/ZE!i ~s ~Lt.l'J 'Ii i-12.U 1 1:~ -H !~rj +' Eli'~/d
+ ~ ~-vo. o.:~
+ c,J/ro l
! o..l o.~
- ~ GJ"/a_ J.. + GJ/10
5
+1Hr~/~ + ~ ~1;/c..i -12.ErJo.3 C\. t (,EC'._/d°
0
+- ! GJo.
15 : (l~ i (),,_ - GJ/10 -~+ 2. EI't./o.
+ ,, f!'j,,/(1 I ~ ,J
I , - 'ET'i.-ld -~ u 0..
AbE I AbE~c.i
---
v...
Q I
'
j
0 0 ---1 I
0
SYMMETR.IC l I
-- - Q.
-
12.E I
011
:,l.EA'"Y,i! -1.z.£I 1 ls i' EA'-'l",!J 'E Ia~ S
2 A.~ '
'
~3 o.3 d.~
I
o.~
Ioy • I y + AbZc I
I
12.E I 0 ., 11.E I-;, 'f's ' ' E r.,,. -,a let 'Ys
2
1oz • Iz + AbYc o.~ a. s I o..a. a.. J.
12.i t 1 Y.sf<l.l; -·' E l;J i~~/J
1
- --
TABLE E-I If. 11. E 1~1~~' eL'1T,~ t. E.t~t.:;&-
s TIFFNESS MATRIX FOR A COMP A';' I Br~.
~}Gj/o.I ... ~~GJ/,o
ECCE~1'RIC STJJ™ER i... 11£rb /Q.3· ._,Er~/o.2.
4 E I 0 T -4£ !"' y~
Q. <>-.
~E.l~ Ys/ o..
'I.'
it-'~ El 1i&/ a.
~Ji. GJO..
-1- 4E l'1o/0..
-\C\O ., , .
2 2 a a az
I l a Gz - Ilf Gl - 11f G2 !lf Gl • ka2G2 6xyl
6 6 a a
SYHHETP.IC r ) Gl nr - IO Gl Wz
Gl • ye 6 a a
! Gz IO Gl - i1f Gz 6xz
I
G2 -- ...E£
ft.
COMPl'TEP PROGPAPS
(rtethod 'a' and rrethod 'b'). Since the input data for both
curved and flat elements is nearly the same, a general des-
cription of both stiffness and instability programs is given
here. All information given in this appendix, is pertinent to
!ID' 360/65 wodel available at the Cornell Universit~ Double
precision is used in all programs.
STIFFNESS Al-JLAYSIS
OUTPUT
INSTABILITY ANALYSIS
Instability analyJis is done only for uniformly dis-
tributed loading for the grid size of t5x6 with 36 elements
though results can be obtained for any general type of loading.
As pointed out in the stiffness analysis, the grid size can be
varied to meet the requirertents. The compiling til!'e is roughly
so secs, whereas the time for individual iterative cycle
ranges a?ong 20 to SO secs.
The input stream is basically the same as that for
the stiffness analysis, the addi.tional information needed here
is the s tartinp; load point (I S.t. increment t see Chapter VI)
subsequent increl'l'ental loads and the number of iteration
-\ C\4- -
TABLE II-2
DOUBLE DECK
2 1/2" x
1/4"
A 0 0 0
Ai. E
- 0..
0
W1
0 0
e(f 1
(L
0
w'l.
0
'7(.
0
2-
GL
I. 0..1.
0
0.
.. o.'"6
-12.E•i y~1
I a, o.2. C.,.
i. I
I
+ G;/.:L
.....__. - Go/"o. j
'J.
""'Aa.E·Zc.
.. I
t..1:,Ei,
G. '-A'1Y,1rAi,E. l,c
0
i;>EI;J 'f I~ Ya c.\. tA.oE"Y, "1i:.
I
i- 4El~ .:L CL o.2- o.'"
!f-2.EI;r ,".\..
0. a..
A&e 'Yc.
1 '<-AloE l',-:
~i>E "f, -,EI ~ A.F~~~
- 'EI~ls
(l.. C)I,..
0
.. .o.£li
C1. 0..1. a. J. .... :e.ri:
,. ()..
0... '
I
+ GJ/o. I
- Ar.E i.t'° l
a.. i-A&,EYc.1,! I
~!-·4'Ii I a.
CL. I
'-•''ft I
ct I
+ 4:_11 II
--·---
e., ~I~ e_1t._:.z..-+l_e_<1_J.-+-s_1 • J G
,+--v-:l..--'--w·-J.-+___ -~J
0
A&,E Z.~
--..- •At/Yr..
_..._-...:;.... 0
__t]_,
- A i l
0 0 0 ·Abf'. '!( AbEYe! 0
0.. a. OI. 0.. 0.. I
-
-12.E.Ii 12f.I1 Zs
0 0 0 0
a. 3 0. '3
SYMMETlUC
_____ _
._..._........-
TAB LE I 1I - 3
--
STIFFNESS MATRIX FOR
.- -
NON ·COMPAT IBLI:l ECCENTRIC
BEAM r:tEMENT (CUSIC
~.:.----
CZ
-CICZcosS+Cxsin~
(A] 1:
}~2x + c2
'z
CyCzsinB+Cxcos~
j crx + c2
z
CJC 0
-C C cosa-czsinB
•-.l- -·· - '
rcr:
../ "':ic
CZ
z
t:r:::zcos B
J"'x""""z
0
0 0
Xz . Xl
ex • L
Y2 - Y1
Cy • L
z2 . z1
CZ • L
9 II 72
., 2 14.4 II 2 28G 0.00047 0.04 29.S 0.30 111~ 2-1 ·~
1
v 40
' Std Std VI
0.0149
pipe pipe
,, ,.
10 II
13. " Ii II
" ii ll
" " "
80
- Z.01 _
- BOUND.ARY LOA
STRUCTURE ~E0~:1:TRY DECK EDGE t~ErBERS CONDITION~ UJ·
i~O.
11
TYPE
It
A
Ii~ O!
72
J:rn 72
c
INOI
14.
TYPE
II
NO. OF THICKNESS
DECKS
1
INCH
28G
I d
INCHt/INOI
0.00047
a
0.06
Ex lo'
LBS/INCH 2
29.S
"
0.3
Bl B2
3 ''C. D. 2-3"0.D.
TYPE
v
PSF
40
40 0.0149 's"Thk l:("Thk VI
pipe pipe
--··
12 •; t;
" 1; It
" 24G 0.000753 i· fl fl
" " VI "
.. 0.0239
- .
,. II 0.00047 o.os " " "
ll II t. ti
2 28G f:
" "
0.0149
t3a
--·- I'
0.06 fl
"
,,
" fl f I
ti
" ti !I
" "
' - h ti
12[40 Z-12 v ti ti
14 1:
240 240 48 ti
'' 16G 0.00192 If
0.060 [40
--
15 i.
12. 12. 5.216 I 1 0.25 - 1.0 I). so 0. 39 +NO BEAMS+ IV 1.: A
92 92
--
x Dimensions are in ems
+ gm/cm
2
STRUCTURE 1 STRUCTURE 2
LOCATION -nethod I a 1 Method 1 b' !~ethod ta I r·riethod 'b I
~2
l(} ems io· 2cms 10 ~ 3 irich 10 • 3 inch
TABLE IV-4
9 28-G
Double Deck 832 848 1460 1870 2292 2718
11 28-G
Single De<.:k 890 62 6820 10700 7710 10762
12 24-G
Single Dede 69 5600 6200 6269
13 28·G
Double Deck 2780 22 6505 4510 9385 4532
EXPERIVENTAL
LOCATION ANALYTICAL AVEPAGE J r•AXIPm,• lr,,If'lIMUt'
STRUCTURE 11
c5 a 0.133 0.20 0.26 0.15
c5b 0.29 0.29 0.48 0.18
c5 c o.16 0.15 0.15 0.15
c5e 0.91 0.70 0.92 0.59
STRUCTURE 12
STRUCTtmE 13
--
c5 a 0.18 0.15 0.26 0.040
----- --·-
AXIAL
--·----------------
BENDING TOTAL (Absolute)
LOCATION EXPT. ANALYTICAL EXPT. ANALYTICAL EXPT. ANALYTICAL
-e
c
TABLE V-1
SELECTED EXAMPLES SHOWING THE EFFECTS OF CHANGE OF STRUCTURAL PARAMETERS
9 g 10 Rise 'C' 14.40" 0.38 1.57 0.38 0.69 -5329 -5329 2747 2747
13.80 11 0.41 1.69 0.41 0.75 -5518 -5518 2752 2752
13 g 13a Shear 0.05 0.18 0.31 0.18 0.56 - 856 - 856 43~ 43~
Rigidity -0~06-~~o-.15-~-o~.2~ai:--_,,,.o~.1~s----=o-.4~s~~---....-ss=1.--~---....-8s~1=--~~4~91..--~~4~8-1~
£actor a
8
11 & 12 Thick- 0.0149 0.13 0.29 0.16 0.91 -1129 - 802 456 1070
ness
o. 02 39 0.116 0.26 0.13 0.57 -1150 - 870 420 1093
No. Single
11 & 13a of Deck 0.13 0.29 0.16 0.91 "'.1129 - 80 2 456 1070
Deck Double
Deck 0.15 0.28 0.15 0.48 - 881 - 881 481 481
EDGE MEMBER STRESSES PSI DECK
MAX. BENDING BENDING STRESS Shear force at
*Structure --~--~----~~--~--~~--~--~~~~-----------------
PS I Center for Nxy
oa oc ab be lbs/inch
52.75
9 fl 10 19,077 19,077 18,179 18,179 2,130 54.96
6,648 6,648 4,228 ~,228 4,510 50.0
13 g 13a 5,899 5,899 4,095 4,095 3t780 51.28
11 & 12 5,225 4,634 3,990 4,480 10,700 36.96
---------~~--------4~,_7_2_3_____3~,_9_7_4~---3~,~8_0_1~~-4,385 6 zoo 37.75
TABLE V-2
STRUCTURE
!>··E~·IBR.AN E
THEn~y
STP.ESS
'
NO. a. PSI oa oc. ab be 0'11 oc ab be
--- -
SINGLE !'EC!
- s 1.0 *19660 10700 907(1 ~440 11~40 54.4 48.? S'l.6 56.1
18640
11
..... ~-
14 0. 06· 3420 1673 1673 1575 1575 49.0 49.0 46.0 46.0
-· DI~NSIONS
: '
b !OS 108 S/8 . - 4.0
~
... 2
106 )08
,_
. a/2 . - -
~a...._
t
t ........ •'b - -
16 16 0.10 9. 34 4.65
__.
-- a ... ""'
t t 70.S . 10. 5
24-G
0.0239
- - 0.670 7.46
Jb
t
-- .
Properties nnd DiD.ensiotts cf Flat ~he~r Tests
1
(ft)
6 x 6
Decking
26G s.c. *
Hembers
3
Connections
screws @ 8 11
Cor-.r.c ct ions
-'
3 6 x 6 26G S .c. 6 11x3/4 11x. l046 11 3 screws @ 8" screws @ every .056 11400 r->
1 layer channels 3rd valley and @8" ,
4 6 x. 6 28G S.C. 6°xl12 1'xe 1046" 4 screws@ so screws @ every .011 12000
1 layer channels 3rd valley and @S"
8 5 x 5 . 28G s .c.
1 layer
6 1'xl':t"x. l046"
channols
3 screws
@ 2-2/3 11
screws @ ev\!ry
3rd valley and @8"
.068 llSOO
* Standard Corrugated
v \\ - \
v 1\-\
Table Z1 Continued
14 6 x 6 28G s .c.
2 layers
l'' std weight
pipe
3 screws
@ 2-2/3"
screws @ every
3rd valley and @8 11
.045 15200 '
15 5 x 5 28G s.c. 6''xl1i"x.1046 11 3 screws screws @ every .oso 15400
2 layers channels @ 2-2/3" 3rd valley and @8"
V\\-2
T.:i.ulc }:xpc~d1;1,~nt.1l Rc.r;ults for S.1dcllc Shaped Hypars
Supported 1\l l Aroun<.l (q ra 40 psf)
Test in
w1:i.:i.x Stress nt Center in Strong Dir. Axinl Force in
No. inches BcndinB (psi) Axial (psi) Tie i3a r ( 1 bs)
811 l. it. 20f•OO -640 1820 (5660)"'*
Sl2 1.14 19100 -1700 1630
511 o.ao lf•70o'"
512 0.77 15900* .
-
311 0.33 2~;00 1260
312 o.32 3.:60 810 1710 (2120)
\J\l- 3
Tnblc Expnriracntnl Results for Saddle Shaped Hypars
Supported All Around (8 11 x 12" area loaded)
,......
3 !/)
c::
0
.....
+-'
al
,....
::J
bO
•l"'I
'+.
s::
0
u
,....
al
p.
....
~
N
I
.-1
-'t.\j -
0
0
s
A
s
A I
J
y
y,Ey,Dy
Fig. 2-1 Typical Orthotropic Deck
a) Sinusoidal
Corrugated
-
b) Trapezoidal
I
J-- 1
... I I
t[J L ·1 1 l .l'\-1
~
1
--1'
c) Cellular
t
:1_ _L I
J- i
•l
-.r l
d) Stiffened
Panel
; 11 JI
,.. 1
11:
~I
I[ : e) Box··type
Effective Width
Varies
z
Fig. 2- 3 Effective Cross-Sectional Area
of a Hat for Axial Force
Max.
Stress
- ]Hean
..1-~1--~------~--------
Stress
e/Z
Effective Width
.I
Fig. 2·4 Effective Width of Compression
Flange in Bending
Principal Directions
----Lower
Deck
Edge ~embe~ ~~
---·--
8 The Author
A Ref. 19
0 .. 08
...0 A.
+'
u
Flat
"'>.
fJ,,
Shear
.,..+' Test
....
"CS G
A
N
'
,.,,
.,..
b(,
A.
°'•
~
... 0 .. 04
aS
u
.a=
(/)
..
. ti
s ..... c"
""
o.oz
c in inches
=3c
4 8 12 16 20 Z4 28
0
0.02
C in inches Jc
4 8 12 16 20 24 28
0
Fig. 2-8 Shear Rigidity Factor 'a' Vs Hypar Curvature
(Double Layer of Decking)
e
xl
r,
ox
1
'"'
ul Uz
wl exyl
exy2
ex ex
?\ "3
71
3
U4 U3
] ey
3
w3 6xy3
W4 axy4 V3
N My
Nxy . r Mxvr
0 2.LJ ___ 0
\ . x
x
1
-~
N
x ( )Mx
Mx
Nxy J
u~?Axy
. Nxy
~
y
My
s-s
l- aL2 '
a/2
y
s-s b) a = 96 inches
b = 144 inches
t = S.O inch
E = 3x10 6 lbs/
F.E. s-s inch 2
v = 0"30
IB
--·-·...,..--------1---x q = 200 lbs/ ft 2
'No. of Elements
• 18 (Half
Plate)
:1 1>.0064ai.
0B . · D " ·
s-s
a. 2 a/2
I Y
;S-S
I. c) a= b = 70.50 inches
t "' 0.0149 inch
s-s 28G Std. Corrugated ·
b/Z ~
I ..
) - Deck · ·
E = Z9.Sxl0 6
lbs/inch 2
v .. o.. 1a .
·q • a• 3 o. p s i.
r{o .. of El~ments
. • Varies (Quadrant)
b/2 ....
o • .S~
• :~
B ·n-4 EI·y
I J ,)
a/2 a/2
1•
y
Fig. 3-4 Plate Bending Problems
Fig, 3-5 Deflection Profile Across Corrugations
0 (see Fig. 3-4C) for Uniformly Loaded
simply Supported 28G Standard
Corrugated Steel Deck
q = 0.30 psi
2 Ill
G>
.i::
u
s::
1-4
s::
•r-4
""r''
4 0s::
•r-4
+J ~
u
G>
r-i
~
G>
Deflection of
'
Q
6 Beam Strip
... 6.96 11
B
A
8 0 The Author
A Series Solution
C.G s.c.
z z
(a) (b) (c)
Fig. 3-6 Typical Cross-Sections of Beams
- - - - - - - - s. c.
y,v
-
x
x,u
z ,w
fig. 3-i An Arbitrary Cross-Section of a Beaw
wl, 8xyl
-
x
~y,v
zc
y
x,u
e-' (
e Plate
(a) (b)
0.25K/"
'C - 4
.....~·
0
- ~~~-il~~-~~~--~~-
No. of Elements
2 4 6
No. of Elements
Fig. 3-13a Effect of Restrained Warping on
Vertical Deflection 1 oq'
. 8. 0
6.0
(Ila)
1-4
0
1-4
1-4
'-1.l
4.0
2.0
0
No. of Elements
Fig. 3.13b Convergence Characteristics for
t:. 0 Q t and t eQt
-LLC\,.
x,u
·--"E-e_z_....,;__ z, w
Column
A A
A =A =A =A =A
1 2 3 4
B1•B 2=B 3=B 4=B
C =C =c =C 4=C
1 2 3
- - - - =o
x •x =x =x
1 2 3 4
- - - -4=0
r1=r2=r3=r
(a)
A A
A =A =A =A =A
1 2 3 4
3 4
B •B •B =B =B
1 2 3 4
x c1 =c 3 =C
C2=-c 4•-C
x 1•x 4 •A
z
--- x' x 2:ax 3 =-A
/R(A,B,O)
y
y• Y1·r2·B
(b)
Y3-r4=-B
Q
pt Qt
Q" x
S' R'
PO+RS
P'Q' • R'S' = 'z
P'S' ... Q'R' • PSiQR
y
(a) (b)
P_ _._ _ _Q.,.__...,...x 0 -
,_:::::::::::::-----..--::::--~.._.. x
---..,.... x '
x'
- z'z t
y y'
(a) (b)
Edge
Member
(a) Moment-Free (b) Sliding-Connection
Connection
Fig. 3-2) Deck and Edge Member Connections
{a) (b)
0 , ... 1
0
x
z
·--· ·--· ;-r--~Y
z
-
x z
(a) M.:. 1
Pxl .. I }7 °~xl
p.x2 (1 l----Px2 EM
0
&'I 0
MoQ, M-oP + (P x2 -P xl)
i:b J ZS
- M- • 0
oQ
Ip xl l MOP ... M-oQ = (P xl -P x2) z,.
~x2
(c) MoQ ~ MoQ
A A
--i h
t
c
b
Fig. 4-1 Structure Type I
Deck
'C >
1 2 3 4 s 6 7 8
b
I I I I I I I a
A -
. '
B
-
c-
D - e
E-
F-
G -
H -
c - 0
g h 1.... z
1" (TYP
B Varies
p
1a.-1 A
-
PA
f
- ti
·_01
I
c
Strucs.
8t
'6' & t
B 1/ 4" Th·c
Sh·!ll
d a b
A A
1
Plan
Strucs.
'6' & tgt
Z-1/2" square---
column __)----'
Struc. '7'
Section P-P z
v-· Symmetry
1.0
2.0
f2.53
A*Z.55
3.0
Scale _
1" = 10
2 cm
·lo -.... b
I Symmetry
I 1.0
0 x 10 2
2.0
r:2. 46
3.0
* Method ta'
** Method t b t- .
+ Ref. 17.
Fig. 4-6 · Deflection Profiles
(Structure I 1 I)
0.035
E
u
s::
\3349
·~
c
'O
=
0
Scale
·~
~
1" = 0.0025 cm
,'
u
G.> f')
r-4 0.030 w
~
11>
~
r-4
C1S
'
....E
0
z
0.02588
0.02530
0.025
2x2 4x4 6x6 8xB
Grid Size
Fig. 4-7 Convergence Characteristics for Deflection ' <5
0 ' (Structure t 1 t)
a
.. I
.
v-Sy:nmetry
4I9. 20"'
10
Scale
l"=Sxl0- 311
Symmetry
s
~9.18+
9. 28 ... 10
* Method 'a'
•~ Method 'b'
+ Ref. 20
Fig. 4-S Deflection Profiles
(Structure '2')
-z..q. ·-
d e
\
\.
\
.o "' ·~
' .
............. ·- /:rs.-·
~ .
2.41 Scale
1"=1"
f g
.o -·Ar·-- Experimental
e Analysis
9.32
7.50
b ~-~------'-----i---'--__.a
9.83
7.50
Scale
1"=7.50 ksi
1/2 The
9.32
7.50
9.83
7.50
a o
-·-·- Experimental
c--·-----------=i
z-L: ::J
v-------
3 - -~
s-~- ~
[----:-:-- ----:------1
6_L ~
7- L:----~--_:_:::---1
8-a o
Fig. 4-12 Shear Force Nxy lbs/inch
_(Structure 'S ')
-l-4r,
0.016
0.02 0.018
a b
o.o
--c-· Q..-• - 9 ·--..;:) 0. 0038
_ _.c _ _ . ca-. _...Q.
.!..
----.-----~:-=-= 1 e . s . Q- · lt: · ~ . o• a2 z
• ~ 0.023
Scale
1"•0.02"
. --
0• or~~-=-~·A::=:::;=::·=4s: - -
·e-- ·--eJ-. - . '""""""
:--s-.:.. ~
0 .. 02
'·"
- ·-A- Experimental
~·-&·~ Method 'a'
o Method 'b'
200
c::::
.....
100
a b
Tension Member
Sea le l" .., 100 lbs
324 .. 0
300
Mern.brane
Theory
~·-4r·~ Experime~tal
'
100
''
''
0
--~~ a
Compression Member
80
Axial Stress
A.
. aJ
~
Cl) 60 ~
~
p..
s::
•rl
Otj
~
'
_,)
0
~
40 MtJmbrane
Theory '
20
4 6 s9
Stress in psi at Point 'e'
Membrane Shear
,. 13. 48 lbs/" ~
c------~--
8-------~~~~----~~--~---
a o
0.025
°"'· .............
A
Scale
l"=0.025"
0.025
o.oso
-·-A-:-·- Experimental
e Analysis
Fig. 4•17 Deflection Profiles
(Structure '7'1
324
300
' ' ,,
en
,Cl
' '' Membrane Theory
' ',/
~
200
....s:: 160
4)
u
$-.
'
''
0
u..
100
+ ''
''
a, Tension Member b
j
Scale
1"=100 lbs.
300
''
en
. '''
,Cl
''
'' , f
.....
....s:: 200 Mcmbrane
Theory
~
u
...
'
0
u..
100
0
Compressio11. Member a
-zs1-
Scale
1"=25
lbs/l'
Membrane
_____ .L:~48lbn"_. __ _
Shear
a o
0.10
4
0.08 --~.
0. 0 s G- ""'--...A.._
dr1~-+~~+-~-+-=~~.:-.~~~~-4-~--+-~--1V
+
~. "'·~ 0.068
0.10 "--£0.098
0
0.] 0
+
~.~
~ ~0.052
" .. o.o9o
0 • .10
~
- · 6-.·- Experimental
Analysis
40 - · -&-- . - Experimental
Anaiysis
....
Deflections in inches
Fig. 4-21 Load vs Deflections (Structure '8')
Grid Size
4x4 6x6 8x8
1.60
II)
Cl> 1.574
.£::
u
s::
•"'4
s:: 1. so
·~
.tl
co
c::
0
..,
•"'4
u
Q)
..... 1. 372
~
Cl>
Q
1. 30
o.so
cS ,
a-------
0. 3 8 <1---A-
O. S1 Scale
0.56 1 11 :110.SO"
1.0
1. so
1.57
0 r"b
- ·/Jr.. - Experimental
e Method 'b'
o.s - -c- -- Method 'a'
1.0
1.5
Fig. 4·23 Deflection Profiles
(Struc:. 1
91) ~
--z.$'t..-
0
- , 1• a
c5
o.s
Scale
l"•O.SO"
1.0
1.S
- ·~·- Experimental
- - - o - - Structure '9 t
1.0
1.50
'.Fig .. t·24 Deflection Profiles
(St rues. '9 ~ & '10)
-2.'SI-
Axial
Stress
. -
S.35 1 tt,. S ks i
S.52
Bending
Stre.ss
-
1"•10 ksi
a
0
--.-- --·A
-G
1• a
Total ..
Stress
10.0 1"=10 ksi
--·--A·--- Experimental
Structure '9 t
20
---iC:I---- Structure '10'
23.70
24.60
a
Fig. 4-25 Stresses in Compression Member
(Structures • 9 t and '10 1 )
--i. s: l-
a
a
Bending
10.0 Stress
l"slQ ksi
17 .14
a
a---- ------,--
:;;-o"
Total Stress
10.0
~- l"s.10 ksi
7.K -·-A· - Experimental
~·~ / Structure '9'
~ . _...I).__
3-c----------~
4 __ C_ _______.A;g~ Scale
l"-10
--
6.75
lbs -inch/
inch
S- , , _ _ - - - - - - - - - - - · - - - - - -
,,,,----
/ 7.69
6 ___ /_~-·--------·~----~~
7------
S-a
~-+-------·--
- ~~o
4-E::~_:._ ______ ~
1"=100 lbs/"
s- ~-=----~---_--_=-_ _:_~-_]
6_E::_-_---_-_-
_ ~: =:J
7_E..;:::
Membrane Shear• 49.50 lbs/"
r--------------=i
s-C
a == ] o
Fig. 4-28 Shear Force Nxy lbs/inch
. (Structure '9')
-'Z. C., I -
Deck
/
(a)
A• Mt 2 ~--L--·-~
ZEI
(c)
()
t5 -o 0. 38
0. 50
o.ss
0.56
...-~~~---~~~~--~~~~----~~~,~~-b
a
o..3&r---..,..
0.5 Scale
0.5
l"=O.SO''
1. 0
1. so
()
0 ... b
-·A-· - Experiinental
- - e - - B.C. V
0.50
--e--- B.C. VI
1. 0
Bending Stress
Shear Loads
1"=25 ibs
so.o
75.0
Shear Loads
Fig. 4-31 Effect of in-plane Fixity on Bending Stress and
Vertical Shearing Loads on the Tension Member
(Structure •gt)
-2.tA- -
0 a
=-
~~-,~~--~-1r~
-~Q...- .!.--~--a0.043
~0.133
0.20
c5
c
-~~g._-..oo.os3
o._.._............ ~ l iif"
------------.0.156
0.20
0
Scale
l"=0.20"
C----,-r-----,--.;....__-.------r ..b
--a.- ---&.-. -Q..- '"'€1
..................
~ .....
0.20 '19-.,.-a 0.24
0.29
b
I =-
Scale
1 "=0. 50"
0.50
Experimental
1.0 9 B.C. \'I
__...._ __ -
cS
----- B.C. V
F ig. 4-32 .
Deflection Profiles
'llt)
(St TUC tu re
0 I .,. b
0. so
"· 0.29
0.29
o.so
0.57
Structure '12'
0
I , ... b
0.50
0.56 - · - 6 · - Experimental
Analysis
Structure '13'
667
+
c b
Scale
1"=667 psi
0
J. c
<1
Max. Membrane Stres~
• 1570 psi
o a
150.0
-t-
0 _ _ _ _ _ _ _J _ _ _ _ __.. _ _ _ _ _ _ _ _ _ _ _..__.._
c
\_
300.0 Scale
l"=lSC lbs
150.0
~
+
0
-----~~-~l----~----
a
IA 1B I c I D IE IF IG IH I
[- - - -- - - - ---- - - -]
1-v:::~- ·~
b c
c---------------·---
2-k:::::. - ~
c-------·--------
3_v...-:~ ~
-J
c------------
4-~
---:.---1
~ -
Scale
1 11 :. 100
lbs
c---------------'1
s-v::= ~
6-V_;:::;=-=::--:------- ~
7- r:-:::::::-=-=-- - --- - - -=-:;J
c---
s-k::::·
---- -------J
~Membrane
~
Shear = 48 lbs/"
e o
Fig. 4-36 Sh~ar Force Nxy lbs/inch
(Structure '11')
0
------------------,.- a
.-.,.,.~. 4. __ Q • -~
~·-A- -=--eo.029
.......A.
-.... ~o. 046
o.os
0
a------.,-------------------b
0.05
0.092*
0.10 0.090
0.123
... b
0.05
--·-4·-- Method I aI
-- • Method 'b t
0.10
---#--- Ref. 21
* Ref. 19
0.123
0.144
Section y =~
10-0
Scale
l" = 10
lbs/n
2 (}()
I
/I \
•
\
~
Scale
1" = 1.0
Section ~. lb• -incft'
a
.~.
x • 2
+
e ~fethod r a'
-·-·-Ref. 21
1. 33
.o
<C
---e-- Single De~k
/
~~·..
o.__...
_,,,,...
0 r.. .,,,,,...,..
b '
..A '
rJ
-
.J
0 • 61 . ~.
~
c;_;.-.().o.:.----- -~ '
._,.,- ,,,,.,,,.,,,,,,.-
•..,...,
.~ ~ .
~..,,,,,. .... ~·
/ ,__,,,,,
/
0. 3 3 / .,,,,..,--
. ~-
~
.~./
·
er
AB
-.~· 10 20 30 40 so 60 70
2
Fig. 5-1 Central Deflection '\S ' vs AB/C (Structures' '3' and '4')
0
-2..11..-
~1
a
Dec
(a)
--- ~~=::::!;:;~~~--~..-!=:-=-=--~·==:l=l._
--·-·-·
_..,,.____....a...~~----·--·-· ~
0
..... -------- ---
(b)
Scale
J"::0.50"
0.4
0. 8111----e----
1.20
0
l. 0
l.50
0.48
Sym.
0.81
Scale
t
! 1. ZS --.._
l"=O. SO"
t;·;-;·-·A--.-___..--.. . . . . -
Def l ec: t icn Along Tension Diagonal
a
., b
...
Scale
3 .. 2 7
l"=S ksi
5.06
1st Trial
op II
Value of De-
fc!'ma t ion at
the end of p I
First Step
z.o
4.0
Scale
C~mpression Diagonal 'ob 1 1"=2 cm
a--~~--~·~--,.--~/"'--><--,-~--------------,·~ r"- c
I
/ '
I
/
/
,/
,,,,,,."""
2.0
- - - - -Buckled Scale
Tension ...
Dia~onal 'act l "• l cm
kgm/cm
z
1.0
2.0
I~
Compression Diagonal 'ob~
Scale
1 11 :.:: lcm
----r---.....,..-----.-·---- c
0.?. 0
Scale
e 1tt=0 • Scm
p.
....
"'co'
~ 1
"O 2 0
~
0
...::I
3 '
i k'
Ill ..
~
0
_ _2.i..-.0- · ~ 2.0 il 2. 0
_,.__ _ _ _,.....,,.._ ~ in inches
Fig. 6-6 Buckling of a ZSG Double Layer All-Supported Hypar (Structure '13')
0
~ • 23.psf e ___/ / ~ b
\ /
so\\ ,,--.......
\ I
/ ------
46 psf _____ /
1. 0
'" C~mpression Diagonal 'ob'
Scale
l"•O.SOtt
1.0
c5
Tension Diagonal 'ac'
1.0
Scale
l"•0.50°
1. 0
~
. i
~~e--- Flat Element~
Cl)
i:l.
....=
't:S
45
I
.I
"
I Laa
Note:
Curved Elements
~
0 see Fig. 6-6
""w""'
'
Point 1
11 Point 'Z' Point '3 t
fig. 6-9 Buckling rif a 24G Single Deck All-Supported Hypar (Structure '!2')
c----- --.....-----,----r---y----.......- b
o.2s
36 psf
o.so
0.75
Cvmpr.ession Diagonal rob'
6
0.25
Scale
o.so P'•O. ZS"
0.75
so
-~
.3 30 Scale
l"=0.10" Deflection
II
0.10 0.20 'l.3J 0.40
.l
o in inches
l1I I
I
I
•'
.' I
I
I
,
I
,•' I
I
''
f
I
,
I
..
•
f
I
I \
''. l \
di" 1
\
gurc 3IA F test
1-Tl
' I.
I "'
.. hs I 1:: :ii> , 1'.I·
'
1.~
Figure~ no.
- Z.'t'I
I
; - . -- ----·- -- -of shear . r. -----·------·---------··-~op
--------, -1
-I · -· .
I l
'\.··~··
'
I
j
·-1\'.
. ,---z·
,
I
1
\
'/
I
--
\
'
1-- - i-- -·-- · · ······ ···-·-···--· -3-------- ··-. -
•· "'
I
l)Tr:-Of siwa:r;.
,..----..r-~---~"-·-
--~- I \
\
J
)'
I
--
~ :rs----~.
\ I; \
...
.
·' .
I
I
,
I
•rop
' ~l.:1:1cr
'
' I
,'
I \
'
'A
I '
:'\,,.
. '
.~ \
'
·- ,
---•
~
\ ,
'
'r"
(layer
l) 0 t •
<
-1 ...... I
--3:>00
-3000
-2500
-<WOO ·o
....r::
rl
I
II
0
·-l
-l '.>00
(.1
F'i[jtlt'e ~ J.oacl versus Deflection
[<)()() !
- j i
~6 in inches
o. t 0.2 0.3 0.4 o.s
____ L·--·-----·-- _ L ____ _J_ - - -
-0.10 16900
'
.....1t;
l
I
~0.08
Figure 'li;l" Shear Rigidity G' a cGnt i
Flat She~r Tests Versus Hypar Curvature for I
No. 3., 9 Cne Layer of Decking
No.
i-o. 06
l
10140 Jl
! I
!
\, Hypnr Tes ts No. Hyp..-ir Tes ts No. I
C!. C' '--s-511,, 512 311., 312 l
'
0
J
i
!
!
t-0.04 6760 I
I
' i
I
I
i I
I
i
I
!
I'
3380
Tc
--1.
t-Oo02
! I
I
I C in inchas
3 12 16 20 24 28
I
I
-0.10 33300 J
! 7. c-,
Figure r;::;::2 She~r Ri~idity G' a r.Gnt
i
I
Versus Hyp.:ir Curvature for
Two Layers of Decking
I
1-o.oa
I
21040
P.yµnr Tests No.
S21, 822
a. G'
0.06
$hear Test
No. 15 No.
Hypar Tests No.
-----~3~21, 322 5r
r-0.02 6760
re
--L
l'
c in inches I
4
I
8
I
12
I
16
I
20
I
24
I
28
I l
--i.,s-
:t .ss
-·. + .22.
•l5 • 11 • O7 -• Ot. -• 01.
··-=.:..T:.:::.=t- t=. +
O
. . *-.
__ __.,......,·:- --- -·--:·--_-----~·04 - ·- ~ ·---
.15
l
.• l}
('!.
.
• 55 .66
••. •
.48
.15-f-
• Ol.-----. 26 · · ·
-t- ;- +·- .::.-.-:±..~.
-. 07 .59 • 66
+ +
• lt
-u,10+.
+i
-140 c=77of·.-. -~: __·.-
·.;4so· . +
+6200
300
+T. 5130
·~236o
.
--····- · ··- ·-·· -+10200 ···
-~---·::-~.: ·_:_-_++ 2290 ·-·~- ..
-·. - . +7690----
+2710 (1550)
·-· +4110
• 0 7 • 1 l • Ol1 • GL1
.l) --+--c: ~-1· ... - i·.. --1- cc- .04
-.01
.07 +
.22 ,29 .18 .04 .22
+ + + +
.• 11
·-17 ·¥
.15 • 7.9 ,110
·43 o_
+ +
, .13 .33 0
.22 + +
.26
·~
·+9 .l10 .o4
f +
• 18 0
+ .15
--··--- -t-----------1'.·--·-----··------·
.22 •oi. •04 .04
(-770) -1030
++.3660
+1270 -1110
·rl1 lJ '.)(1 ·-t- +5080 +
·- --- --~_7)_2~;_7-~~~~-:.-:::::.-,.}_1?:~~IS5 0)
.. _ -·
J
-------~---
- +12t.
+9flfl()'.~O)
strong dir.
• 55 .1~0 • 6 .37
--- --·- -·t· ---·-;--7·-.
• 37 -.01
.33 .40
+:.
t
.29 :
+ .
• ll .n • 1l .ot~ 0 .11
+· + + +
• ta .29 .04 15
.l'... + +
• ta
·~ >¥·
.Ol1
.lH
.2G
+
.11
' '
'f3
.15 .26 .22 -.04
+ + +
.22 : 1I .01
i 0 I . . ; ;. .
.
.... . ·l -·-t ---·-I· + --·-·-!-'---:__,. ·-.----~~-----·-----~·-~-----t
• l v') 0 .15 .01 .01 .01 .04 .15
1.\2(°')
~xperiraontnlVertical Deflections
in inches nt 40 psf
Test No. 3-23G-c2-Tl
- 'SOo -
+930(1550)
+5260
i.
·I· -112oll(-770)
.' +3960
+290011 ' ++t0150 I;
I
I
l.\l(~)
Fi;·,urc :·>\.porh1cntol Strcsse>s in psi
nt 40 psf Load for Test No. J-2~G-C2-Tl
_.,.. f -
.300
--------.135
.026
+
. ~
. ~
···"-~·----·
. -·-
.. ----...........----. ,. ..... --·---·-··--
~~ ,
~--- -- ...·---
' '
- - .020
.029 .015
.063
+ + .032
+· It- .057 +
· .os2
• 315
:·~xpcrhKmtnlVertical Deflections
in inch0s at 40 psf
Test No. 4-2N-C2-T2
1.73 1.01 .48 .33 1.56
--·- -----1----,,...._..---,1.---i-_-__-.-.:-..-.-i------..
.92 1.25
t .J
'
).
.14 .....f2
: .- -.
__ ~ "
1.01
.. ~.
••. • • •
i - . 1
!
'J
' I . I
' '
.05.: .79
I;
1
i ' ' t
'
I
.32
''
'
. ·.36
I'
I:
~33 t.46
' ' ! •
I
+. ' I
I t:
'
----~--~-1---'--'--------'l'------''-l1~--'---ll,_._ __ . 1 . - - - - 1
1,83 .99 .59 .67 1.00 1.39 1.84
1. \4-(c:\j
r;xperi:i1enta 1 Vertical Deflections
in inchas at t~o psf
tast No. 5-2SG·C2-T3
,\:~i.11 :;tt·,~:;:;'-·:; ar•: directly above bcndinp, stresses.
;),•ci-; :;ti-c·:;:~l':; in h<Jtto!'I L1ycr <1rc to the left of those
in lop 1;iy,~r.
- ·-~----
-- -
. -'
.
-;-111~0 +570
+201.. 0+·1·2270
, I
I ,
'-·--~---- --~--·-
I . \ if 'c,» .
i~i:~ut·c · 1 .J::x;x•ri:.\~ntnl Str.e:.scs
in psi <1t t+O psf
Trst No. 5-2SG-C2-T3
PROF. IR. c. G. J. VREEDENBURGH
From the differential equation of the doubly curved shell in terms of the
displacement components u, v and w it is apparent that, in so far as the
.flexural phenomena are concerned, the shell can be considered as a .fiat plate
on an elastic foundation, at a~y rate in a region - of sufficiently small size
to be regarded as quasi-Euclidean - around the vertex of the osculating
paraboloid. The modulus of the elastic foundation is dependent on the shell
thickness and the principal curvatures of the middle surface. The actual load
on the plate must, however, be increased by a deformation load depending on
the displacements u and u in the middle surface.
Various aspects of this approach to the problem are verified by reference to
simple examples and to the results of tests on a large model of an equilateral
hyperbolic paraboloid shell.
0 Introduction
It is a well-known fact that a curved shell can resist loading acting perpen-
dicularly to the middle surface by the agency of membrane forces. In those
cases where the membrane reactions can be resisted and the corresponding
deformations of the shell can freely take place, a force distribution pattern
constituted solely by membrane forces is a good approximation of the trans-
mission of forces that actually occurs - for a statically possible stress distri-
bution in which bending and torsion are avoided will come close to producing
the minimum strain energy in the structure. The usual procedure, therefore,
is to begin by calculating the distribution of forces in the shell according to
the membrane theory. Then a set of correcting forces will have to be applied
so as to take the best possible account of the boundaries (edge beams or sup-
ports) of the shell, where the deformations due to the said state of membrane
stress are prevented from freely developing. Let us suppose the edge member to
be detached from the shell, thus enabling these deformations to develop freely:
the edge of the shell and the edge member would then no longer fit at their
junction. A proper fit can be obtained only if the edge members exerts forces
(normal and shear forces) and moments (bending and torsion moments) on
the edge of the shell and if, conversely, the shell exerts opposite forces and mo-
ments on the edge member in such a manner that the additional deformations
associated with these forces and moments enable complete adaptation (struc-
tural fit) to be achieved. The calculation of these edge disturbances, which
must be superimposed upon the state of membrane stress, is done with the aid
of the so-called flexural theory of shells.
2
Simplified flexural theories are already available for shells shaped according
to simple mathematical surfaces such as the spherical and the cylindrical shell.
A. L. BouMA has applied a theory of this kind to elliptic and hyperbolic para-
boloid shells bounded by four principal parabolas.I)
Even these simplified theories undeniably involve a fairly considerable
amount of arithmetical computation; and for doubly curved shells of arbitrary
shape or combinations thereof (which may or may not be bounded by elastic-
ally deflecting and/or elastically rotating edge beams), such as are rather
frequently employed in modern architecture, there is as yet no serviceable
theory for the analysis of flexural disturbances.
The distribution of the edge disturbance moment from the edges towards
the interior of the shell surface is known in general to have the character of a
damped wave. An obvious question is whether it would not fundamentally
be possible to calculate the approximate magnitude of this edge disturbance
moment by means of the theory of the elastically supported beam - for in a
beam of that kind the edge disturbances display a damped-wave distribution.
Thus J. W. GECKELER 2 ), as long ago as 1926, gave an approximate analysis
for the edge disturbance in a spherical shell which leads to the theory of the
elastically supported beam. Furthermore, in 1953 K. HRUBAN 3 ) analysed the
edge disturbances along the curved edges of a number of hyperbolic para-
boloid north-light shells with the theory of the elastically supported beam,
while in 1958 W. S. WLASsow 4 ), with reference to the spherical shell and the
equilateral hyperbolic paraboloid shell, called attention to the analogy with
the elastically supported plate.
Finally, for the estimation of certain edge disturbance moments associated
with the structural research on the Philips Pavilion at the Brussels World
Exhibition in 1958, the present author also proposed a method of analysis
based on the theory of the elastically supported plate.5)
If permissible, the method of analysis based on the elastically supported
plate or beam undoubtedly constitutes a considerable simplification. This is
because the differential equation of eighth order on which the flexural theory
of shells is based is replaced by an equation of fourth order, which provides a
simpler means of gaining an insight into the anticipated edge disturbances,
this being of great importance to the designer.
1
) BOUMA, A. L. Some applications of the bending theory regarding doubly curved shells.
Proc. Symposium on Theory of Thin Elastic Shells (1.U.T.A.M.), North-Holland Pub!. Co.,
Amsterdam. 1960.
2
) GECKELER,j. W. Uber die Festigkeit achsensymmetrischer Schalen. Forschungsarbeiten
Ing. wesen, No. 276, Berlin, 1926.
3
) HRUBAN, K. The general theory of saddle-shaped shells (Czech), Techn. Univ. Brno, 1953.
') WLASSOW, W. S. Allgemeine Schalentheorie und ihre Anwendung in der Technik, Aka-
demie-Verlag, Berlin, 1958, pp. 330 and 372.
5
) VREEDENBURGH, C. G. J. The hyperbolic-paraboloidal shell and its mechanical proper-
ties, Philips Technical Review, Vol. 20, No. 1, 1958/1959.
3
On the other hand, when applying an approximate method of analysis it is
essential to have a clear conception of the quantities that are being neglected
and of the assumptions made, in order to be able to judge whether useful
results are likely to be obtained in any particular case.
kx = a2z l
ax 2
a2z
kxy = -- . . . . . . . . . . . . . . . . . . . . . (3)
axay
a2z
ky = ay2 J
4
For P 0 we furthermore have:
au w nx
Ex
ax rx Eo
av w ny
(5)
Ey
ay Ty Eo
au av azz nxy 2nxy
Yxy -+-
ay ax 2w--
axay Go Eo
or:
Notation
5
By substituting the relations (6) into (4) and having regard to (3), we obtain:
Essentially this equation is valid for the shell in the immediate vicinity of 0,
but it is approximately valid also for points of the shell situated not too far
from 0. For practical purposes we may consider this region to be bounded
by a circle having 0 as its centre and having a radius equal to half the smallest
principal radius of curvature at 0. Then the ratio between the rise and the
chord of all the normal sections of the osculating paraboloid will be smaller
than 1 /7 so that the corresponding region of the shell can, as a rule, be assumed
to be quasi-Euclidean.
Putting:
If the factor v is also taken into account, the relations ( 11) and ( 12) become:
( 13)
( 14)
6
while the plate stiffness in (10) becomes:
1 Eo 3
K=--- ( 15)
12 l-v 2
1) For the significance of"extensionless deformations" see: BouMA, A. L., Stijfheid en sterkte
van schalen, Waltman, Delft, 1960.
7
Fig. 3. Hyperbolic paraboloid shell, bounded by
four generating lines and subjected to uniformly
distributed load, considered as an elastically sup-
ported plate. The edge members are very stiff
beams. The shaded area is the edge disturbance
zone. In the region surrounded by this zone the
deflection w of the plate is constant and the bend-
ing moment is therefore zero in all directions. The
edge disturbance zone has a maximum width
of H. B
~!!11, ID P .11Llll~
~ '-,y~~"-''-"''"'·''-Y//"'-'-.Y/O''-'y'/ ,
A-B
The quantity A, which is known from the theory of the beam on an elastic
foundation and has the dimension of a length, is sometimes termed the char-
acteristic length. It can often be assumed that, depending on the mode of
support, the edge disturbances will, for practical purposes, no longer be per-
ceptible beyond a strip having a width ranging from A to 4A along the edges of
the shell.
8
2 The foundation modulus c
Consider a circular cylindrical shell rectangular on plan, which is subjected
to a uniformly distributed radial load p acting over its entire surface and
which is so supported at the edges as to be free from disturbances.
Let the radius of the circle be r (Fig. 2). Since, as already stated, in this case
the displacements u and v are both everywhere zero, the deformation load is
zero. Furthermore, the displacement w is constant, and - having regard to
( 10) - we therefore find:
p
W=- (19)
C
1
Introducing ki = - and k2 = 0 we then obtain, according to ( 14) :
r
Eo
c = ------
(l-v2)r2
so that finally:
( l -v2 )pr2
W= . . . . . . . . . . . . . . . . . . . . (20)
Eo
In this case the displacement can also be calculated directly, as follows. The
compressive (tangential) "hoop" stress in the shell is:
pr
(J =- (21)
0
As the lateral contraction is, in the present case, prevented in the longitudinal
direction of the shell, the compressive stress in the direction of the generating
line will be:
(J ' = ')l(J • • • • • • • • • • • • • • • • • • (22)
The specific compressive strain along the circle is therefore:
(l -v 2 )a (l-v 2 )pr
E=--- (23)
£ Eo
9
The displacement w is obviously again expressed by the formula (19), but
the foundation modulus c for the sphere is different from that for the cylindrical
shell. 1
Introducing k1 = k2 = - we now have, according to ( 14) :
r
Ea [-+-+-
1 2v 1] 2Ea
c=~-
l-v2 r2 r2 r2 (l-v)r 2
so that:
(l-v)pr 2
W=--- (24)
2Ea
By direct calculation we obtain:
~ w(1-0pr
a = - and s = = -
2a r' 2Ea
i.e., w is found to have the same value as that given by (24).
These simple calculations may be regarded as providing a check on the
correctness of the formula (14) for the foundation modulus c if the shell is con-
sidered as a plate on an elastic foundation and the deformation load is zero.
Since - as will be apparent in due course - both the actual load and the foun-
dation modulus of the analogous plate are modified as a result of taking the
deformation load into account, the foundation modulus according to (12) or
( 14) (which therefore relates to the case where the deformation load is zero)
will be called the "primary" foundation modulus. The modulus modified by
the deformation load will be called the "secondary" foundation modulus.
I
11
11
11
~_.::-_.::-_.::-~~~~r~~~~~~~ D
11
L_ _ _ _ _ _ _
Pz = ~o [;~ + ~J . . . . . - (26)
As the effect of bending on nx is very slight in the present case, we can write
nx = -pr, so that equation (10) becomes:
Eow
Kl::i!::iw =P-P+--cw
r2
which, on introducing
Eo ow
c =-and - = 0
r2 oy
gives:
(J4w
-= 0 (27)
ox
4
In this case the loading of the analogous plate and the foundation modulus
have both become zero in consequence of the deformation load, and the dis-
placements w of the cut-out strip CD are governed only by an edge load at
the points of restraint C and D of an unloaded beam. This restraint reaction
is not difficult to determine. It is a horizontal tensile force H passing through
r sm rp
the centre zb of the arch, zb
being located at a distance - - - from the
rp
centre M.
For v = 0 the magnitude of H can be calculated from the relation:
12r3 ( 2 sin 2 rp) 2pr2 •
H - - rp+l/ 2 sin 2rp - - - - = E'~ sin rp . . . . . . (28)
Eoa rp u
11
Fig. 5. General view of test arrangement for a hyperbolic paraboloid shell in the Stevin
Laboratory. Overall dimensions of the model on plan 5 m X 5 m.
If His known, the displacement w 0 at the crown of the arch can be calculated.
By considering an elastically supported beam strip with its axis along OY we
then obtain - with }c = 0.76Vro according to (18) - the following expressions
for the moment and for the shear in the shell at 0 1 ) :
Eb 3 wo
mo=---- (29)
5,1c2
qo (30)
12
Fig. 6. Underside view of hyperbolic paraboloid shell with loading devices. Shell thickness
3 cm. Cross-sectional dimensions of hangers 10 cm X 10 cm, other edge members 15 cm X
15 cm.
-60
~
c c
-40
-20
80 100 1 I
140 1 0
+20
______.... x in cm
Fig. 7. Reinforced concrete hyperbolic paraboloid shell with all four bays carrying a uni-
formly distributed loadp = 1000 kg/m 2 ; A= 22.4 cm at B.
EF considered as a beam on an elastic foundation, completely restrained at E and elastically
restrained at F.
o measured shell moments
calculated without deformation load
calculated with deformation load equal top
13
which was square on plan and which were interconnected by stiff ridge
beams, while the inclined edges were also provided with stiff beams (see Figs.
5, 6 and 7).
Consider the case where the four bays carry a uniformly distributed load
p = 1000 kg/m 2 over the entire surface.
Having regard to the characteristic length (22.4 cm at the vertex), it was
anticipated that the case envisaged in Fig. 3 would occur. Hence the elastically
supported plate was locally replaced by an elastically supported beam. Con-
sider a beam strip of unit width cut out of the bay ABCD along the generating
line EF (Fig. 7). This beam can be regarded as completely restrained at the
ridge beam. On the other hand, it is elastically restrained at its junction with
the edge beam at F. Let x be the distance measured to the completely restrained
end of the strip. The distribution pattern of the moment in the elastically
supported beam for a uniformly distributed load p will then, as we know, be
given by:
Neglecting the deformation load, we obtain withp = 0.1 kg/cm2 and).= 22.4
cm the following value for the restraint moment at E : m; = -25.1 kg. The
n
maximum positive moment occurs at x = A and is 5.2 kg.
2
In Fig. 7 the distribution curve of the moments calculated m this way
emanating from E is shown dashed.
The moments emanating from the elastically restrained support F can
similarly be calculated and are also shown dashed in Fig. 7. On comparing
the values calculated by neglecting the deformation load with the measured
actual moments (indicated by the small circles in Fig. 7) we see that there is
indeed good agreement in the pattern of the distribution, but that the calcu-
lated values are much too low. If, to allow for the effect of the deformation load,
we consider twice as large an actual load on the shell (load factor = 2), we
obtain the curve drawn as a full line in Fig. 7; this curve is in better agreement
with the measured values. As emerged from a further investigation of the
problem, the order of magnitude of the deformation load could, in the present
case, be explained with the aid of the measured strains of the edge members,
au av
from which it was possible to calculate average values for - and-.
ay ax
As regards the peak values of the moments at E and F it is to be noted that
these are usually not so serious as they may appear to be.
For one thing, acute angles are always more or less chamfered in actual
practice. Furthermore, ifthe limit of proportionality of the material is exceeded,
a stress peak will be levelled down, while a residual stress of opposite sign will
be produced on unloading. Also, if the load on the shell under present consid-
14
Fig. 8. Underside view of a
fractured bay, showing cracks
extending approximately par-
allel to its edges.
15
at its perimeter. Let r be the radius of the sphere and neglect Y. Then, according
to ( 12), the primary foundation modulus will be:
2Eo
C=-- .
,2 ............ . (32)
:;: ~~~ = 1. . . . . . . . . . . . . . . . . .
qo).3 - mo).2 Wo
(33)
2EI 2El
Hence:
2Elwo
mo= - - . . . . . . . . . . . . . . . . . . . . . (34)
). 2
4Elw0
qo (35)
16
With formula ( 17) we obtain for the sphere:
i5r
;.b2 = V3V2 (36)
Neglecting the deformation load, we must, for calculating the restraint mo-
ment mo, substitute Ab 2 according to (36) for ?. 2 in formula (34).
It should be noted, however, that by considering a detached diametral
beam strip too unfavourable a distribution of forces is found, because in the
edge disturbance zones AC and BD, as part of the complete circular plate,
shear forces and torsion moments are actually present which have a relieving
effect.I)
Consequently the calculated moments will be larger than those occurring
in reality. A further reduction will in this case be caused by the deformation
load.
On substituting for ?. 2 in formula (34) the value Ac 2 according to (37), which
relates to a cylinder, we obtain the restraint moment according to GECKELER's
approximate theory, which is, however, smaller than the correct value. 2 )
Hence, if the deformation load is neglected, it is safer to apply the characteristic
length based on the primary foundation modulus of the sphere. The fact the
characteristic length of the cylinder of revolution occurs in GECKELER's ap-
proximation can be explained with the aid of the deformation load. Because
of axial symmetry, we have:
Eo .
For -
r2
= 1/2c (where c is the pnmary foundation modulus for the sphere)
n
and putting --"' = - 1 /2P, which is an acceptable assumption if the central
r
angle of the spherical shell is not too small, the differential equation for a
meridian strip will be:
d 4w
K- = p- 1 /2P+ 1 /2cw-cw
dx4
1
) In this connection it is appropriate to recall that, for example, in a circular plate uniformly
loaded over its whole area (freely supported or completely restrained at the perimeter), the
moments in a diametral beam strip, as part of the plate, are only 3 / 8 of those in the detached
beam subjected to the same load.
2
) Cf. TrMOSHENKO, S. Theory of Plates and Shells, McGraw-Hill, New York, 1940, pp.
472-475. See also HETENYI, M. Spherical Shells subjected to Symmetrical Bending, Puhl. Int.
Ass. for Bridge and Struct. Eng., Vol 5, 1937/1938, pp. 173-175.
17
or: (39)
2
· 1ar so1ut10n
W e see t h at t h e part1cu · of t h"is equat10n · w0 = p
· remams pr-
- = -
c 2Eb
whereas the value of the foundation modulus has become half that for the
sphere, i.e., equal to that for the cylinder of revolution.
In his plate analogy for the feebly curved shell WLAssow also arrives at a
foundation modulus equal to that for the cylinder of revolution, this evidently
being - as in GECKELER's method of analysis - correct only for axially sym-
metrical loading.
It should, however, be borne in mind that the load which has to be applied
to the analogous plate or beam may differ considerably from the actual load
on the shell.
In more general terms it can be said that, for a shell of revolution with
axially symmetrical loading and a meridian section of arbitrary shape the
secondary foundation modulus will be Eb , where r2 is the second principal
r2 2
radius of curvature (length measured along the normal up to its intersection
with the axis of revolution).
For ov = 0 and neglecting P, we obtain from ( 11) and (5) :
ay
_= ou = -Eb ( -ni + w)
Pz Ebk1 -
ox r 1 Eb r1-
which in combination with ( 10) and ( 12) gives:
5 Conclusions
The conception of the doubly curved shell as an elastically supported plate
has much to commend it. In highly complex cases it would, for the present,
appear to be the only method available to the design engineer for estimating
the approximate magnitude of the edge disturbance moments. It should be
borne in mind, however, that a doubly curved shell can reasonably be re-
placed by a flat plate only if we confine ourselves to a so-called quasi-Euclidean
region, i.e., a region around the origin of the chosen system of co-ordinates
where the principles of plane geometry are still reasonably valid. This can be
18
assumed to be so up to a rise/chord ratio of about 1/7. For calculating the pri-
mary foundation modulus at any particular point both the principal radii of
curvature must be taken into account.
This foundation modulus is valid only if the deformation load is zero or
negligible. If the deformation load is taken into consideration, then the actual
load on the analogous plate or beam as well as the primary foundation modulus
will, generally speaking, be affected. This gives rise to a secondary foundation
modulus. In shells of revolution with axially symmetrical loading this second-
ary modulus is dependent only on the second principal radius of curvature,
this being in agreement with the foundation modulus as applied by GECKELER
and WLAssow. In the case of a cylindrical barrel vault uniformly loaded over
its entire surface, presenting an elongated rectangular shape on plan and
supported on all four sides, the secondary foundation modulus may, for a
shell strip extending in the transverse direction, even become zero.
The analysis as a plate, or a beam, supported on an elastic foundation
appears to be capable of yielding useful results in dealing with spherical shells,
hyperbolic paraboloid shells bounded by generating lines, and elliptic para-
boloid shells rectangular on plan.1)
It will have to be further investigated whether this analysis can be applied to
other cases and also whether it can be applied if the deformation load is neg-
lected but suitable load factors are introduced at the same time.
1
) GIRKMANN, K. Flachentragwerke, 5th ed., Springer-Verlag, Vienna, 1959, pp. 462-464.
See also the publication mentioned in note I on p. 229 of that book.
19
DOUGLAS FIR
USE BOOK
STRUCTURAL ·oATA AND PESIGN TABLES
.- .~.
-~
·i •.· .
t
Ii
·I·
Outer
·concave parabola
.
1)
Inner 1htathin9
Convex parabola
HYPERBOLIC-PARABOLOID SHEUS
be the same, except that the perimeter :mem- ing act in the direction parallel to the grain,
bers will be in tension mst.ead of compression which provides the most efficient use of the
and the horizontal thrusts will be in the op- sheathing.
"
·,:,:
posite direction. For small shells, the second system can be "'
used to advantage as the sheathing boards do ·J;.·'
Boundary Shear: not curve and nave only a slight twist from
end to end. If the shell is small enough so
1
:·.··
There is a shearing force along the len~ that the sheathing is without undesirable de- l
'
ol the perimeter members at the junction of flection\ the falsework can be omitted. This
the sheathing and the ~rimeter members. system has one disadvantage since the joints
This boundary shear, "v, ' in pounds per lin- between constant width sheathing boards
eal foot, is found by dividing the compressive would leave gaps varying in width from a
force, "C," in the perimeter member by the maximum at the ends to no gap at the mid-
length of the perimeter member. point. · U this is undesirable, the sheathing
boards may be tapered to fit snugly. ·
Arrengtment of Lumber Sh11thlng:
Stress In Sheathing:
Lumber sheathina is ideal for shells be·
cause the comparativily narrow widths of the Since the stresses in the sheathing result
boards permit easy adjustment to the doubly in boundary shears along the perimeters, the
curved surface without .special cutting and bounda.r y shears can, conversely, be resolved
fitting except for end trimming to length. to determine the stresses in the sheathing.
Lumber sheathing may be applied to hy- The principal forces in the shell are tensile
perbolic-paraboloid shells in two ways: · forces, ''t,' parallel to the direction of the
concave parabolas; and compressive forces, /.
1. One larer of sheathing boards placed "c," paral1el to the direction of the conyex
paralle to the transverse axis where parabolas (see Figure 86): When the hori-
each piece bows to fit the curve of the zontal projection is a diamond shape, the ,1
,,
convex parabola with a second layer principlil tension and compression forces par-'
placed parallel to the longitudinal axis allel to the longitudinal and transverse axes
where pieces bow to fit the curve of can be resolved by proportion as follows:
the concave parabola.
2: One layer of sheathin~ boards placed Referring to Figure 8(), the principal ten-
parallel to two opposite sides of the sile force per foot of width, "t,' is: ·
structure and a second layer _placed
parallel to the other two sides. In this t li/2 Liv
system, each sheathing board twists = -- and t =
slightly. The total amount of twist v a' 2a'
from perimeter member to perimeter
member depends on the slope of the The principal compressive force per foot
perimeter members. of width, "c," is:
If the horizontally projected shape of the c l2/2 12v
shell is a square, the layers of roof sheathing =-·-and c =
are at right angles to each other for both sys- v . a' 2a'
tems ,of placement. If the shap_e of the struc-
ture is that of a diamond, the double layer of When the horizontal projection of a hy-
sheathing boards will be at right angles to perbolic-raraboloid is square in shape, the
each other when placement follows tlie first princif a tension and compression forces per
system. When placed by the second system, foot o width are equal in magnitude to the
the angle between layers of boards will de- boundary shear forces per foot of length of
pend on the angles between perimeter perimeter members.
members.
The unit tensile stress in the sheathing
As· there js a slope, but no curvature, to a lumber is equal to the principal tension force,
series of straight lines from one edge to an "t," per' inch of width divided by the thick-
opposite edge and parallel to a side, falseworlc ness of the sheathing in inches that parallels
placed in diis manner will automatically gen- the longitudinal axis. The unit compressive
erate the doubly curved surface and serve as stress in the sheathing which acts at 90 de-
support for the placing of the sheathing. grees to the tensile stress is equal t!) the
The first system of placing sheathina principal compressive force, "c," per inch of
boards is advantageous beciluse the principal width divided by the thickness of the sheath-
tension and compression forces in the sheath- ing in inches paralleling the transverse axis.
DOUGLAS FIR USE BOOK 279
HYPERBOLIC-PARABOLOID SHELLS
....':ii.C ti
280 DOUGLAS FIR USE BOOK
HYPERBOLIC-PARABOLOID SHELLS
Unloaded end-
designed accordingly. If the sheathing is
cumulative comprHtion sandwiched into the perimeter members with
force varl11 uniformly half of the perimeter members above and half
from ztro ot ll'le unloaded below, there is no eccentricity and the peri-
end to ma•lmum ot ll'lt
btarln9 1nd . meter members are subjected to axial com-
Pet mtter pression stresses only. The latter method
mtll'lber
permits a somewhat smal)er perimeter mem-
ber to be used, but increases the number of
pieces to be framed and handled in erection.
. .. i
As a hyperbolic-paraboloid shell becomes
flatter, it becomes more flexible .with an in-'
1.
- ~
creasing tendency to buckle. For this reason
it is desirable to place a limita~on on the
flatness which can be expressed as the ratio
t, of the rise to·the length of a side. A minimum . ..
..
rise-length of side ratio of 1 to 5 is suggested. ' '
. ~~·
B1ori"9 end Method of Determining Twist In Perimeter
:ro1a1 cumulative
comprntion force Members:
at blarlnia end. Becaufie a hyperbolic-paraboloid shell is a
doubly curved.surface, the slope of the sheath·
ing at the junction with the perimeter mem-
Fl1ure 87-Column Action of bers is con.~tantly and uniformly: changing
Perimeter Member.
•'
along. the length of the perimeter mem'Oers.
As the surface of the perimeter members
The perimeter members transfer all loads must be tangent to the slieathing where they
to the bearing point and must have sufficient connect, the contacting face of the eerimeter 1 1 i
cross section to resist the cumulative axial members must be shaped aJ>propriately. ll ;~
compressive forces. As the boundary shear the perimeter memben are glued laminated ).
forces are distributed uniformly along the members the .changing slope can be obtained I
length of the perimeter members, the com- .by building in a twist to the whole member. ' :
pression force in the perimeter !llembers For each shell, two right hand and two left
varies as a cumulative sum of the boundary hand pe~eter members are required. ,'i
shear from zero at the high point to maxi-
mum at the sup.P.orts. The perimeter members
can be tapered if desired. The sheathing pro-
vides lateral restraint to the perimeter mem-
bers in the direction parallel to the plane of
the sheathing. In the din!ction perpendicular
to the plane of the sheathing the j>erimeter 90• onat1
members receive no lateral support, and the
Angl1 ABC
slenderness ratio related to this plane must
be considered. As the compressive force varies
·" uniformly from zero at the peak to a maxi-
mwn at the ~mpport, the peruneter members
are considered as a series of columns varying
.,;
~· :
from a long column with no load to a short
~. column with maximum load. At any point
i along .the length the induced compression
1: · parallel to grain stress due to the accumulated
load must not exceed the allowable unit stress
as determined by the standard column formu-
la for a column length equal to the distance
t: from the support to the point being consid:
h
·''
ered (see Figure 87). If the sbeath~ng is Perim1ttr member
!
placed on the top or the bottom of the
perimeter members, the boundary shear for<:es Angle of twiat
cause bending stresse.s in the perimeter mem-
bers due to eccentricity. Hence, the perimeter
members 'a re subjected to combined bending
and axial compression stresses and must be F11ure 88-Twl1t In Perimeter Member.
,.
DOUGLAS FIR USE BOOK 281
HYPERBOLIC-PARABOLOID SHELLS
The total change in slope, or rotation of the . mond shaped horizontal projection, the
contact surface of the perimeter members formula for the angle of twist is ·
from one end to the otlier, is the ang1e of ha
twist. As the ratio of the rise in respect to the tan of angle of twist =
length of the perimeter member increases, · (a') 2 cos ABC
the angle of rotation also increases.
where angle ABC is the angle shown in Fig-
The procedure for computin~ the angle of ure 88.
twist is applicable to h)'l'erbolic-para bOloids
having square or diamond shaped horizontal When the horizontal projection of the hy-
projections. perbolic-paraboloid is square, the angle ABC
Using the mid-point of the length 'ot a becomes zero and the formula reduces to
perimeter member as a convenient reference, ha
and considering the twist at this point to be tan of angle of twist =
zero, the angle of twist from the reference (a')Z
point to either end of the perimeter membeJ'
as shown in Figure 88, can be determined The total angle of twist from one end of
from the following formulae : the perimeter member with respect to the
other is twice the angle determined from the
For a hyperbolic-paraboloid having a dia- preceding formula.
f
['
•'
,;
•1
t'•I
;
.r
iX~9A
J>6~~·
f C f1Vll ENGINEERING STUDIES
0
r r, . ,- STRUCTURAL RESEARCH SERIES NO. 304
by
B. MOHRAZ
and
W. C. SCHNOBRICH
A Report on a Research
Program Carried Out
under
National Science foundation
Grant No. GK-538
UNIVERSITY OF ILLINOIS
URBANA, ILLINOIS
MARCH, 1966
THE ANALYSIS OF SH.ALLOW SHELL STRUCTURES
BY A DISCRETE ELEMENT SYSTEM
by
Bo MOHRAZ
and
Wo Co SCHNOBRICH
A Report on a Research
Program Carried Out
under
National Science Foundation
Grant Noa GK ... 538
UNIVERSITY OF ILLINOIS
URBANA, ILLINOIS
MARCH, 1966
A'.cKNOWLEDGMENTS
This report is. based on a thesis by Bi.JC!I1 M()hJ'."El.Z tl.119.~:r tli~ d:i.ir<=ction Qf
Page
ACKNOWLEDGMENTS . . iii
LIST OF TABLES . vi
LIST OF FIGURES. vii
JNTRODUCTION . . 1
l.l. General . . . . . . . 1
l. 2. Objective of Study. 3
l.3. Nomenclature. 3
METHOD OF ANALYSIS . . 6
2.1. General . . .6
2.2. Description of the Model. . . . . 7
2·. 3. Coordinate System . . . . . . . . . 8
2.4. Displacements . . . . . . . . . . . . . . . . 9
2. 5. Strain-Displacement Relations . . , . 10
2.6. Foroes and Moments . . . q •••••••• l3
2.7. Equilibrium:Equations . . . . . l4
BOUNDARY CONDITIONS. . • . . . . . 19
. 3.1. General . . . . • . . . 19
3.2. Simply Supported Edge . . . . . 19
3.2.1. Roller Support. 19
3.2.2. Hinge Support . • r . . ' .. 24
3.3. Free Edge . . . . . . . 24
. . 3.4.
'(',t
Other Boundary. Conditions . 31
NUMERICAL RESULTS. . . . . . . . . . . 32
4.l. General . . . . . . . . . . . 32
4. 2. All Edges Simply Supported . . 34
4.2.1 .. Uniformly Loaded Rectangular Plate . . . 34
4. 2. 2. Uniformly Loaded. Cylindrical Shell. 34
4.2.3. Uniformly Loaded Elliptic-Paraboloid. 36
4. 2. 4. Uniformly· Loaded Hyperbolic Paraboloid. 36
4. 2. 5. Uniformly Loaded Hyperbolic· Paraboloid -Bounded by
· Characteristic Lines of the Surf ace . 37
'· 4. 3. Two Opposite Edges Simply Supported and the
RemainingTwo Edges Free. . . . . . . . . . 38
4. 3.1. Uniformly Loaded. Square Plate . . . 38
4.3.2. ·Cylindrical Shell Subjected to Sinusoidally
Varying Edge Load . . . • . . . . . . . . . . . . . 38
-iv-
-v-
Fage
Number
-vi-
LIST OF FIGURES
Number
-vii-
-viii-
Number
1.1 General
and esthetic value, are fre~uently used to cover large, column-free spaces.
In the interior regions of a shell, away from the supports, the loads
are carried mainly by in-plane forces. The membrane theory provides a good
estimate of stresses for design purposes. However, in the regions close to the
supports, bending stresses' are developed as the results of the sharp change .in
the stiffnesses from that of the shell to that of the edge member.
shells, bending solutions are available for a variety of loadings and support
shells, bending solutions exist only for a limited number of cases with simple
have been found are rarely those in practical applications. Usually the govern-
ing e~uations are reduced to two e~uations in terms of· a stress function and
the radial displacement [ 13] * . A Levy type solution can be obtained when at
least two edges are simply supported. But these supports are usually a poor
/
-1-
-2-
struction economy and the inherent beauty of these shapes, they have been widely
shells and their extensive use have led to numerous studies of their structural
behavior. The edge disturbance from a straight boundary pe~etrates into the
shell further than the edge disturbance from a curved boundary [4]. Therefore)
a bending analysis which takes into account the effect of supporting the sbell
a Levy type solution such as the one used for cylindrical shells leads to
orthogonal functions.
Approximate methods have been used whenever the exact methods have
methods has been used to obtain a solution for a hyperbolic paraboloid with
simply supported and clamped boundaries [5]. For shells of arbitrary shape,
simulate the shell structure. Hrennikoff [12] replaced the continuous structure
by a system of elastic bars. Yettram and Husain [19] used the system of elastic
bars to obtain solutions to plate problems. Parikh [15] and Benard [2] applied
this method for cylindrical shells. A different model which replaces the
continuum by a system of finite plate elements has been used by Clough [7,8] .
.Zienkiewicz and Gheung. [ 20, 21] have applied the method to orthotropic .slabs
and arch dams. Discrete models consisting of rigid bars and deformable nodes
have been used by Ang and Newmark [1] to obtain solutions in plates.
Schnobrich [16] used a discrete model for cylindrical shells. One of the
obtained.
solutions of complex shell problems which previously were not possible because
theory will be used as a guide in the development of the model. The generality
l.3. Nomenclature
The symbols used in this study are defined when they first appear.
E modulus of elasticity
h shell thickness
eXl+
. l"J rotation of bar ij, i+2j in x direction
b
a .. = flexural stress in x direction at deformable node ij
XlJ
'T
xy shear stress
xXlJ
.. flexural strain in x direction at deformable node ij
2 .1. General
Ritz and Galerkin methods) etcoJ have certain advantages for a particular
present any difficulty in the method of analysis. The method should be adapta-
thickness.
bars and deformable nodes similar to that proposed by Schnobrich [16]. Since
the material properties of the shell are concentrated at the deformable nodes,
the method can be used to study the behavior of shells with different material
properties in the two directions. It can also be used to analyze shells with
variable thicknesso The forces and the moments are constant across each node,
thus, non-linear material properties would not present any difficulty in the
equations within a digital computer. Due to the effort involved in the de-
to develop a single general computer program which can be used and can easily
can become very complex) once it is completed many shell surfaces with various
-6~
-7-
additional programming.
the coordinate lines are not necessarily the set which coincides with the
The dis.crete model employed in this study cons is ts of rigid bars and
deformable nodes arranged as shown in Figs. (1,'2). The nodes have extensional
and shear properties similar to those of t.he real material. At the midpoint
of each rigid bar there is a.groove with a circular hole at its center, Fig. 3~
The rigid bars are connected tO each other at these grooves by a pin which is
inserted in the holes. With this type of connection, the two bars move
shells [16]. By specifying· the tangential displacements at the same point, the
extensional and shear behaviors are no longer separated from each other. The
model can thus be used to study the nonlinear material behavior of shells
When the model is used for the analysis of shell structures, the
Fig. 1. For shallow shells the generators are usually approximated by circular
-9-
axes. When the origin of the coordinate system is placed at the deformable
node ij, the x and y axes are set along the bisectors of the angles formed by
the rigid bars in the xz and the yz planes respectively. The z axis is
directed along the perpendicular to the plane containing the bisectors (i.e.,
the tangential plane). The positive direction of z axis is into the plane
this coordinate system through angles ~/2 and a/2 about the x and y axes
respectively, the desired coordinate system at points ij-1, i+lj, ij+l, and
i-lj is obtained.
Throughout this study, the strains, the forces, and the moments are
formulated for deformable node ij. Node ij may be a typical interior node, a
node near the boundary, or a node on the boundary. The equilibrium equation
space~ from the node ij, in the positive x direction. Similarly the equilibrium
one-half space from the node ij, in the ·posi.tive y direction. The equilibrium
and ij may be typical interior points, points near the boundary, or points on
the boundary .
2.4. Displacements
of rigid bars and is directed along the axis of the bar in the x direction.
of rigid ba~s and is directed along the axis of the bar in the y direction.
-10-
11
c. The radial displacement w" is defined at the deformable node
and is directed along the perpendicular to the plane containing the bisectors
y, and z axes respectively. Figures 1, 2, and 4 show the manner in which the
displacements are defined. The above displacements are only the components
displacements.
node ij, the u displacements of the rigid bars adjoining the node must be trans-
formed to the plane containing the node (tangential plane). The extensional
u
E .. (2.2)
XlJ
w -1
E ..
R Cos a/2 wij (2.3)
XlJ
x
1 1
E ..
XlJ L lbs a/2 (11i+lj - ui-1) - R Gos· a/2 w ..
lJ
(2.4)
x x
can be written as
-11-
node ij is
(2.6)
Because of the twist of the surface there is also a shear strain due to the
w d
E ..
xyiJ 2~wi.
x y J
(2.8)
It should be noted that these strains are the average strains through the
thickness.
The above strains are due to extension and shear. There ar.e also
strains due to bending and twist which result from the rotation of the rigid
bars. The rotation of the rigid bar ij J . i+·2j in the x direction can be seen .
from Fig. 9 as
(209)
Similarly the rotation of the rigid bar i-2j, _:_i.J: 1,i'.l the.:U<>di:rec~;i_on•is
-12-
The flexural strain in the x direction at the top and the bottom elements of
1 t
xxi· J. - ·L ( 8 Xl+
· 1 J· - 8 Xl-
· 1 J· ) 2 (2.11)
x
t t
X . . 2R L Cos a} 2 ( u . 1 . - u . 1 . ) + 2 (w . 2 . - 2w . . + W• 2 .)
XlJ x x i+ J i- J 2L Cos a/2 i+ J lJ i- J
x
(2.12)
Similarly the flexural strain in the y direction at the top and the bottom
t t
2R L Gos t3/2 (v ij+l - v ij-1) + 2 A/ 2 (w ij+2 - ~ij + wij-2)
y y 2L Gos f--'
y (2.13)
The twisting strain is obtained from.the relative rotation of the four rigid
(2.14)
Expressions for B's similar to Eq_. (2o9) can easily be obtained by proper use
t ) t
~ij 2R L
x y
Cos a/2 (uij+l - uij-1 + 2R L
y x
Gos t372 (vi+lj - vi-lj)
t 1 1 )
+ 2L L (Cos a/2 + Cos t3/2 (wi+lj+l - wi-lj+l - wi+lj-1 + wi-lj-1)
xy
(2.15)
-13-
nodes are in a state of plane stress" The in-plane forces and the bending
moments are concentrated at the deformable nodes. The in-plane force in the
L
N .. a .. ( 2Y) h (2.16)
XlJ XlJ
From . Eq. ( 2 . 1)
E
a .. - - - ( E .. + VE .. ) (2.17)
xiJ l-v 2 xiJ yiJ
L
N ... -_ Eh i (E .. + VE .. ) (2.18)
xiJ l-v 2 2 xiJ yiJ
Similarly
Eh Lx .
N .. = - l
2 2 (EYiJ' + VEXiJ°l) (2 .19)
YlJ -v
L
Eh i E
N (2.20)
xyij 2(1+v) 2 xyij
L
Eh x
N E (2.21)
yxij 2(1+v) 2 xyij
(2.8), respectively.
-14-
obtained by taking moments about the mi.d-de:pth of the deformable node. Therefore_,
b _][_ h ] t
L
M .. - 2 [ crxi/ 2) 2 2 (2022)
XlJ
b
wµere a .. is the stress due to bending and is given by
XlJ
(2.23)
Thus,
L
M .. ~ Eh _]£_ ~ ( ) (2. 24)
XlJ 2 2 2 Xxij + vxyij
1 -v
Similarly
Eh Lx t
M .. - --2 -2 -2 (xyi" J. + vxxi" J·. ) (2025)
YlJ 1 -v
L
Eh 'y t
M
xyij 2(1+v) 2 2 xxyij (2.26)
Eh Lx t
M
yxij 2(1+v) 2 2 xxyij (2.27)
The expressions for X .. , X .. , and X .. are given byEqs. (2.12), (2.13), and
XlJ YlJ XYlJ
(2.15), respectively.
To obtain the distribution of the forces and the moments, Eqs. (2.18)
L L
through (2.21) and (2a24) through (2a27) are divided by . ; or ~ .
2. 7. Equilibrium Equations
by the internal forces plus the total work done by the external forces is
equal to zero for any arbitrary virtual displacement. For example, to obtain
fixed. The sum of the works done by the internal and the external forces
caused by this virtual displacement is then set equal to zero. The results
displacements. These equations are solved and the obtained displacements are
work is equal to the negative of the change in the strain energy of the four
Wint = - l N 6E L
where N and 6E are the force and the strain due to ·extehs ibn, shear;,: ;bending)-
u Eh L
~ L [ (.
.w.int E • • + VE . . )6E . . + ( X , . + VX . . ). tsx , .
1-y 2 2 - XlJ YlJ XlJ XlJ YlJ XlJ
+ (E.
Xl+ 2·J +VE.
yi+ 2.)6E.
J Xl+ 2·J +vx~-·
Xl+ 2·J + (x. ·-yi+2J.)6X.
Xl+2j
1-v
+ - - ( E • • 6E . . + . . 6X . .
2 xyi+lJ-1 xyi+lJ-1 xxyi+lJ-1 xyi+lJ-1
. l"J+ I 6 Exy1.+
+ Exyi+ . l"J+ I 6 Xxyi+
. l"J+1 + Xxyi+ . l"J+ l)J
-16-
1
f:.E ..
XlJ L Cos a,/ 2 t:.ui+lj
x
1
f:.E . . L Cos a/2 (-t:.ui+lj) (2.30)
Xl+ 2 J
x
etc.
The external work of the component of the load in the x direction at point
i+lj is
Wu X b. (2.31)
ext = i+lj ui+lj
For equilibrium
(2.32)
Substituting the strain-displacement relations into Eq. (2.29) and then sub-
stituting the results i.nto Eq. (2.32), the desi.red equilibrium equation is
obtained. Thus, the equi..librium equat.ion i.n the x direction at poirit i+.lj is
2
LxLy [(. 1 t 1
-2- 2 + 2 0 ) I 2 ( u • ".') • - 2u ' 1 • + u • 1 • )
· -cos a/2 4Rx cos'- cx/2 1 i+ _.,J i.+ J i- J
.x
·2
+ 1;':' ( 1 + t ) l ( u . l . ,-. - 2u . l . + 11 . . )
4R2 Cos2 a/2 L2 . i+. J+~ i+ J i+lJ-2
x ·y
2
(1-v · v l+v t 1
+ 2 + Cos a/2 Cos f3/2 + ~ 4R R Oos a,/2 Cos f3/2) LL (v i+2j+l -v i+2j-l
x y x y
1 1
- v. · 1 + v. · 1) - (
lJ+ lJ- . 2
R Cos a/2
+Ry Cos aJ2 Oos f3/2) Lx (wi+2J' - wiJ.)
x
-l7-
1 d t2 1
(11-v) 1 ) -3 (w . .4 . - 3wi·+2J·
4R Cos 2 aI 2 '·1 L
(w.i +l"J +1 - w.i +l"J - 1) + (
x y y $
l+ · J
x x
2
t (1-v 1 l+V 1 ) 1
+ 3wij - wi-2j) + 4Rx Cos a/2 ~Cos a/2 +~Cos ~/2 2 (wi+2j+2
1 1
x y
- 2w. . + w.
l+ 2 J
.
l+ 2 J- 2
- w ..
lJ+ 2
+ 2w .. - w ..
lJ lJ- 2)·_] = 0
(2.33)
WVext + WV 0 (2.34)
int =
1 1
Eh x y [
w.wint = - --2 --2- ( Ex l. - 2.J + VEy l. - 2.J )6Ex l. - 2.J + (x x l. - 2.J + vxy i. - 2.J )6xx l. - 2.J
1-V -
+ ( EYlJ+
.. 2 +VEXlJ+
.. 2)6EYlJ+
.. 2 + (x.__ .. 2 + VXxi· J"+2)6Xyi· J"+2
'-YlJ+
1-V (
• 1 . 1. 6 E
+ - 2 · Exyi- . 1 . 1 + X . 1 6 Xxyi-
. 1 J-. . 1 +
. 1 J- E • 1 . 1 6.E . 1 J-
. 1
J- xyi- J- · xyi- xyi+ J- xyi+
Z .. 6w ..
lJ lJ
WW WW 0
ext + int
obtained.
points is given in Appendix A. When the equilibrium equations for all grid
points in the model are formulated, they constitute a set of linear algebraic
the equations are those for a cylinder [16]" If both radii are infinite,
3.1. General
2.35). Since the geometric and the force boundary conditions can be expressed
the boundary is similar to the procedure for obtaining the equilibrium equations
of typical interior points, i.e. the desired point is given a virtual displace-
ment and the sum of the internal and the external works of the system associated
obtained according to the restraints at the edge and include only displace-
rigid in its plane but offers no resistance in the direction normal to the
plane. Assuming that the edge parallel to the y-axis is supported by rollers)
-19-
-20-
v 0
w 0
N 0
x
M 0
x
With the above relations, the strains at the deformable nodes on or near the
Node on the e~dge. Using the first two condit±ons of Eqs. (3.1), the
node ij located on the edge are obtained from Eqs. (2.5 and 2.13)"
,€:yij 0
(}.2)
0
Xyij
E .• 0
XlJ
the y direction. Thus,, at the deformable node ij located on the edge, the
(3.4)
Since w = 0 all along the edge, the points of intersection of rigid bars
cannot displace in the radial direction. This can be accomplished by
removing the dowel pin (see Fig. 3) and rigidly connecting the two bars to
-2l-
each other at the edge. The rotation of bar i-lj+lJ ij+l in the x direction
1 Lx 1
8
xij+l = r::J2.
X
(2R
X·
cos a/2 uij+l - cos a/2 wi-lj+l) (3.5)
(3.6)
(3-7)
The twisting strain at the deformable node ij located on the edge can now be
xxyij 1 (e :_ e ) t 1 t
8 (3.8)
- L
y
xij+.1 ' xij-1 2 r::J2.
x
yi-lj 2
Thus)
t t
2R L Cos a/2 (uij+l - uij-1) + 2R L Cos ~/2 (- 2vi-lj)
x y y x
(3-9)
+ t ( .l + l )( 2w 2 )
2L L Cos a/2 Cos ~/2 - i-lj+l + wi-lj-1
x y
Node one-half 'space from the edge. The strain expressions E •• '
XlJ
E •. ) and i .. at the de,formable node ij located one-half space from the edge
YlJ ·-yiJ
are similar to those of 'a typical interior node. The shear strain E .. and
xylJ
the twisting strain xxylJ
.: . are obtained by substituting Eqs. (3.l) into
. l· ·1 " d ;
E -. · = L- (u.··+1 - ufJ·~i) + Lx (-vi-lJ.} - 2 LL wi ·
XYlJ y lJ x y J
t t
x ..
. ~lJ 2R L ,Cos a/2· (uij+l - uij-1) + 2R L Cos f3/2 (-vi-1)
(3 .10)
x y y x
t 1 1
+ 2L L,· (Cos a/2 +Cos f3/2)(-wi-lj+l + wi-lj-1)
x y ·.
.. Node one. space from the edge. At the deformable node ij located
one space from the edge) all strains except X are similar to those of a
x
typical interior node. The flexural strain· xxij is
2R L
x x
~os.a/2 (ui+lj - ui-lj) +
21
2 t
Cos ·a/2
( -2w ij + wi-2j._)
x (3 .12) .
or near· the edge can now···be obtained by the principle ·of virtua·l displacement.
For example, to obtain the eq~ilibrium equation in. th.e y .d;frection at .point
ij+l loca·ted one:half .~p&ce from the edgey.. the ·point is given a vfrtual dis-
placement 6v .. . The internal work of the system is°' thus given .by
lJ'.1- -
LL,;• .
v
Wint = l~y 2
Eh x y.
2_.
[c EY i· •
J
+ VE .. )6E. . . + ( :L , . + VX .. )f:.X__ ..
XlJ .Y:!-,J -· ~YlJ . XlJ -yl,J
+ (E .. + VE .. )6E .. · + (X .. + VX .. )6x._ ..
YlJ+ 2 XlJ+ 2 YlJ+ 2 YlJ+ 2 XlJ+ 2 , . YlJ+ 2
1-~J
+ -2~ (Exyi-lj+l
6
Exyi-lj+l + Xxyi-lj+l 6 Xxyi-lj+l
·1 . 6 1 )'1 (3.13)
+ 2 Exyi+lj+l Exyi+lj+l + 2 Xxyi+lj+l 6.Xxyi+lj+lj
-23-
The coefficient 1/2 which appears in the last two terms of Eq. (3.13) is due
to the reduced width of the strip of the shell at the edge) Fig. 11. The
(3 .14)
For equilibrium
v v
wex t + w.in t = 0 (3.15)
then substituting the result and Eq. (3.14) into Eq. (3.15), the desired
. L XLy [ ( 1 + t2 ) _2:__ ,r {/ . )
2 2 2 2 2 \ ·ij+3 - 2 vij+l + vij-1
- Cos f3/2 4R Cos f3/2 L
y y
2
1-v t 1 (1-v v
+·2(l+ 2 2 ) 2 (-3v .. l+V/. 2·1)+ - 2 +Cosa/2Cosf3/2
4R Cos f3/ 2 L lJ+ i - J+
y x
l~ ~2 1
+ 2 4R R Cos a/2 Cos f3/2) LL (ui+lj+2 - ui-lj+2 - ui+lj + ui-lj)
xy xy
( 1 v ) 1 ( ) (1-v ) d ( )
- Ry Cos f3/ 2 +·Rx Cos a/2 Cos f3/2 Ly wij+2 - wij - · LxLy Lx -w i-lj+l
2
t 1
+ ( 2 ) 3 ( w . . 4 - jw . . 2-~ 3w . . - w . . 2)
4R Cos f3/2 L lJ+ lJ+ lJ lJ-
y y
1-v 1 l+v 1 ) 1
+ -4R--C-os_f3_/_2 ( 2 Cos f3/2 + 2 Cos a/2 L 2L ( - 3w lJ+
. . 2 + w. 2 . 2
l- J+
y x y
2
l-v - 0
+ 3w.lJ. - w.l- 2;·J )] + -Eh y (3.16)
-24-
There~6re, the boundary conditions, when the edge parallel to the y axis is
u 0
v 0
w 0
M 0
x
Taking these conditions into account the strains at the deformable nodes on
or near the edge are obtained and the equilibrium equations for displacement
points near the edge are formulated. The procedure is similar to that of
roller support.
N 0
x
N 0
xy
(3.18)
M 0
x
R 0
x
where Rx is the classical reaction composed of the shear force Q and the
R
x
(3.19)
-25-
ment of the model, the reaction R must be defined at poi.nts along the edge
x
which best simulate a behavior similar to the continuous shell" The most
logical points are the points of intersections of rigid bars along the edge.
For example, the reaction at point ij+l, Fig. 12J is composed of the shear
i
force Q .. which results from the bending moment M . . ; and Q . . which
XJ.J+1 Xl- 1 J+1 XlJ+1
results from· the two adjacent twisting moments M .. and M . . .
xylJ xyJ.J+ 2
As was pointed out in Section 2.2; at the points of crossings of
the rigid bars; the two bars displace independently of each other in the
moments along the edge must be transferred to the ends of the bars intersecting
the edge. This is accomplished by means of auxiliary rigid bars (Figo 12)
which connect the nodes on the edge to the points on the edge where the bars
space from the edge which may result from the displacements of nodes along
geometric relations, both the extensional and the flexural strains in the
y direction at the deformable node ij located on the edge are similar to those
of a typical interior node, .Eqs. (205 and 2013). The expressions for the
extensional and the flexural strains in the x direction at the deformable node
ij located on the edge are obtained from the first and the third of Eqs. (3018).
Thus_,
-26-
.E .. (3.20)
Xl.J
rl t t --,
Xxij -v 2R L Cos f3/ 2 ( v .. 1 - v .. 1) +
- y y lJ+ lJ- 2L
2
Cos f3/2
( wij+2 - 2w ij + wij-2) J
y
At the deformable node ij located on the edge; the shear strain is found from
E •• 0 (3.22)
xylJ
To obtain the twisting strai.n, we first have to consider the rotation of rigid
bars i-lj+l, ij+l and i-lj-1, ij-1. The location of these bars is as shown
in Fig. 13. Since at the crossing points on the edge, the two bars displace
5 ',s (radial displacements of the ends of the crossing bars); are introduced.
Additional equations necessary to solve for the extra unknown o's are formulated
later from the equilibrium of a portion of the model near the edge. With the
1 1 L
8
xij+l = r::J2. (oij+l - Cos a/2 wi-lj+l + 2R C~s a/2 uij+l)
x x
L
1 1 ·x
eXlJ-
.. 1 r::J2. (oij-1 - Cos a/2 wi-lj-1 + 2R Cos a/2 uij-1) (3.24)
x x
The rotation of rigid bar i-2j, ij (which is equal to the y rotation of bar
The rotations of the two auxiliary rigid bars ij-1; ij and ijJ ij+l in the
y direction are
L
1 1
e~. =~(Cos f3/2 wij +
2R
y
Clos f3/2 v ij - 5 .. 1)
lJ-
lJ y y
(3.26)
L
1 1 y
e~. = ~ (oij+l - Cos f3/ 2 wij + 2R cos f3/2 vij)
lJ y y
_!
8 ··· ·· -2 ( 8 · ·
~J . lJ
1
+
2 _
el· J· ) -
.l
Ly [.,. Ry
·y
cL0 s f3/ 2 v l· J· +. ( 0 l· J+
· 1
_
5 · · 1)
l J- .
J
The twisting strain is defined by
1 t 1 t
xY\T i· J. = -L ( e · · 1 - e · · 1) -2 + Ltn2 ( e · · - e · 1 · ) -2 (3.28)
~~ y XlJ+ XlJ- Dx/c YlJ yi- J
Substituting Eqs. (3.23; 3.24; 3.25; and 3o27) into Eq. (3.28) yields
t (u ) t ( l
2R L Cos a/2 -__ ij+l - uij-1 + 2L L ·cos a/2
x y x y
1 2
+ COS f3/2)(..- 2w.l- l"J+ 1+2w.l- l"J- l)+L Lt .. 1 -
(b lJ+ olJ-
.. 1) (3.29)
X y
located one-half space from the edge; all strains except xx are similar to
those of a typical interior node" The flexural strain xXlJ
.. is obtained from
2R L
x x
~os a/2 (ui+lj - ui-lj) +
21
2 t
Cos
a/
2
(-3Wij + wi-2j)
x
t
+ L- 2 o.i+l"J
x
1
= ~ Mxi-lj+l
x
i 1
Qlxij+l = ~ Mxyij
y
I 1
Q2xij+l = r;J2.
y
Mxyij+2
where
Eh Lv t
-- .-JL ~ (x V L )
MXl-
. i·J+ 1 - 2 .2 2 xi-lj+l - yi-lj+l
1-v
L (3.32)
Eh ·y t
M ..
xyiJ 2(1+v) ~ 2 xxyij
L
Eh y t
MxyiJ+
.. 2 2(1+v) ""2"" 2 xxyij+2
I i
0 (3.33)
Qxij+l + Q2xij+l - Qlxij+l
Substituting Eqs. (3032) into Eqs. (3o3l) and then substituting the results
L 1-v
LY (xxi-lj+l + V"Xyi-lj+l) + 2 (xxyij+2 - Xxyij) 0 (3.34)
x
-29-
L
Ly
·x
[
R L
x x
~OS a/2 (uij+l - ui-2j+l) +
L
2
l
Cos a/ 2 (-3wi-lj+l + wi- 3j+l)
x
+
1
l
x2
r· 5
\:? ij+l
) v
+Ry\ Cos fJ/2
(
vi-lj+2 - vi-lj) +
,
t/ v
Cos f'>/ 2
(
wi-lj+3
l 1 1 l
+ R·L Cos ~/2 (- 2vi-lj+2 + 2vi-lj) +LL (Cos a/2 +Cos ~72)(-2wi-lj+3
y x x y
+ 4w. 1 , 1 - 2w. 1 . 1 ) + L
.l- J+ l- J-
{ 45. ,
lJ+ 3
~L
- 85.lJ+
. 1 + 45.lJ-
. 1 )J 0
x y
Nodes one space from the edgeo At the deformable node ij located
one space from the edge, all strains are similar to those of a typical
interior node.
equations for points on or near the edge is similar to that of Section 3.2.
w.wint
-30-
1 1
+ - (E • • + V E • • ) l::,.E • • + - ( X__ . . + 'VX . • ) 6;x__ . ,
2 YlJ XlJ YlJ 2 .. -YlJ XlJ . -yiJ
1-v (
+ ~2 . 1°
Exyi- J- 1
6E
xyi-.l"J- 1 + Xx-,yi- 1°J- 1 l::,.X xyi- l"J- 1
0 0 0
1 E.
+ ~
2 .xylJ
•• 6E .. )
xylJ
J (3-36)
For equilibrium
then substituting the result and Eqa (3a37) into Eqa (3a38)) the desired
LL ,... 2 2
x2 Y 1·_- 1~2v R 1 1 (V 'l - 00
V ••
, 1-v
l) + ~2~ 2 1 2 Wl J
0 0
+ v l- · 1 + v_ lJ=
· 2 J+ · · 1 - v l=
· 2 J-
· 1)
2 1 2
4 + 6 4 + ) + l=v L (Cos 1 a/2 + )
- wij+2 wij - · wi.j=2 wij-4 2 4. Cos f3/2
1
L 2L 2
( -w l J +2 + 2w lJ
0 0
. . 2 + w i. - 2 J+
. . - w lJ- . 2 - 2w i~
. 2 J. + w.•i - 2 J-
. 2)
x y
to that of the simply supported and free case. Fbr a shell clamped along
the edge, all displacements and slopes along the edge vanish. For a shell
4 .1. General
curvatureo It is not within the scope of this study to consider the effect
One.of the advantages of the discrete model is that the various types of
loads are given. Some solutions for shells having different values of Poisson's
ratio are also presented. The obtained results are compared with existing
the coordinate lines do not coincide with the lines of principal curvatureo
The external loads are transmitted to the supports mainly by the in-plane
Gaussian curvature'} the elliptic· paraboloid with two opposite edges simply
supported and the other two edges freea For this shell the magnitude of N
y
forces (edges parallel to the y-axis being free) varies very :rapidly across
a section normal to the free edgeso If the model can predict the behavior
of the above two shells 3 then it can be expected to give good estimates of
-32=
magnitude and distribution of boundary disturbances for general shell
problems.
·lines of the surface J the shells considered herein have both s;~nnmetrical
geometry and loading. ThereforeJ by using only one quadrant of each shell
one-fourth of the original number. The spacing between the two extensional
2
12(1-V )
which gives
h
t = -
0
For flat plates the equilibrium equation in the z direction,
Kea.. (A-3) is independent of the other two equilibrium equations J Eqs. (A-l
reduces to two sets of completely. uncoupled equations. One set contains the
w displacements of the deformable nodes marked as solid circles (see Figo 15).
The other set contains the w displacements of the deformable nodes marked as
hollow circles. Although, the two systems act independently of each other,
the solutions for the two sets were found to be in good agreement with one
rectangular plate is given in.Fig. 16. Alternate nodes belong to one systemJ
the remaining nodes to the othera The agreement between the two systemsJ as
-34-
seen from the figure, is excellent. For shells whose coordinate lines
the two sets of equations a~e also completely uncoupled fTom each other .
. For shells whose coordin~te lines do not coincide with the lines of principal
of the surface, the two sets of equations are strongly coupled through u,
v, and w displacements.
having a Poisson 1 s ratio of 0.3 are considered. The dimensions of the plates
are given in Table 1. Because of symmetry, only one quadrant of each plate
is used in the analysis . . In tpe first plate, the quadrant is divided into
three spacings in each direction . . In the second plate the number of divisions
in each direction is doubled . . The values of the deflections and the bending
moments at the center of the two plates are given in Table 1. The solutions
are in good agreement with those given by Timoshenko and Woinowski-Krieger [18].
number of divisions in each direction, the values of the moments are slightly
improved.
by Bouma [3] is considered. The shell is u,niformly. loaded and has a Poisson's
-35-
ratio of O.Oo The dimensions of the shell are given in Fig. 17. Due to
symmetry only one quadrant of the shell is used in the analysis. The
quadrant is divided into six spacings in each direction. The plots of the
the agreement is good. The small differences between the results could be
due to both the manner in which the two shells are supported along the edges
and the manner in which the two shells are loaded. The edges of the shell
in Ref. [3] are supported by vertical diaphragms and the loading normal to
Z(y) z1 Cos a y
1
(4.1)
where
4 q 1(
z1 1( al - b
Although the transverse edges of the shell in this example are supported by
the shell. The most significant difference in the transverse bending moment,
given in Fig. 18. The dimensions of the shell in the x .direction are the
same as those of the cylindrical shell,. Fig. 17. The dimensions in the
y direction are so chosen that the curved length and the rise of the shell
are 1800 ems and 200 ems respectively. A 6 x 6 grid on a quadrant of the
shell is used in the analysis. The plots of the displacements w, the forces
N , and the bending moments M at the midsection of the shell are shown in
y x
Fig. 18.
N
xy
.:...2. 23 qa
2 2
M -0.069 x 10- qa
xy
directions which are as large or larger than those which exist in the center
The plots indicate that except near the edge the membrane theory
provides a good estimate of stresses for design purposes. Due to low values
of deflections, forces, &.'1d bending moments, this ,shell is very suitable for
roof construction.
are given in Fig. 19. The dimensions of the shell in the x direction are
kept the same as those of the cylindrical and the elliptic paraboloid shells,
·while the dimensions in they directions are so chosen that the curved length
~3T-
and the sag of the shell are 1800 ems and 200 ems respectivelyo A 6x6 grid
on a quadrant of the shell is used in the analysiso The plots of the dis-
The plots indicate that the membrane .theory can not be.used to
moments for· this shell are much higher than those for an elliptic paraboloid
of the surface is used in this example. The shell has dimensions corresponding
to those used by Chetty and Tottenham [5]. The dimensions of the shell are
given in Fig. 20. The-Poissonvs ratio and the modulus of elasticity of the
6
shell are 0.16 and 3 x 10 psi respectivelyo The shell is subjected to a
2
uniform load of 50 lbs/ft . Since the shell is not symmetric, an 8x8 grid
on the complete shell is used in the analysiso The plots of deflection w, the
force N , and the bending moment M at the midsection of the shell as obtained
xy x
with the model together with the corresponding values obtained by Chetty and
Tottenham are shown in Fig. 20. The agreement between the deflections and
the shear forces is very good while the maximum bending moment obtained by
in the test of the model. The results indicate that the model can indeed be
is considered. The dimensions of the plate are given in Table 3. A 5x5 grid
def+ections and the bending moments at the center and at the free edge of the
plate are given in Table 3o The results are in good agreement with those
and the longitudinal edges free is considered in this example. The dimensions.
of the shell are given in Figo 21. The Poisson es ratio and the modulus of
2 respectively.
elasticity of the shell are Ool5 and 4o32 x 10 5 kips/ft The
:rr
Z(y) ·- Cos b y
the differential eQuation is taken from Parikh [15]. The fact that the
results are in good agreement with the analytical solution indicates that the
model can be used to study the effect of edge disturbances in shells . . Parikh 7 s
shells, does not demonstrate the same accuracy as the model presented in this
the displacements w, the forces N • and the bending moments M at the mid-
Y' x
section of the shell are also shown in Figo 220 The plots of the in-plane
shear forces N and the forces N at the support for the two shells are
xy Y
given in Fig. 23.
The magnitudes of N and N at the hinged support are much higher than those
xy y
at the roller support. This is to be expected in view of the fact that the
shell now behaves in a manner like a fixed ended beam. However, as pointed
out below computations based on considering the shell as a fixed beam are not
adequate.
A comparison of the hinged support case with the stresses that are
obtained by multiplying the N stresses for the roller case by the ratio of
y
the bending moment in a fixed ended beam to the center moment in a simply
supported beam is not conservative. The error also may not be insignificant.
the shell and applying the usual equilibrium equations. The summations of
the forces in. the horizontal and the vertical directions vanish and the
summation of the moments about any axis is found to be zero, ioe. the internal
2
forces produce moments equal and opposite to the simple beam moment l/8 wL .
example are .similar to those of the simply supported case and are given in
~42-
reaction of 0.95 qa which is l.7 percent of the N force at the free edge.
y
Reducing the value of the N force at the free edg~ obtained from the model)
y
by l.7 percent gives a N equal to 2l02 kg/cm. As was mentioned previously,
y
this shell is of particular interest in the test of the model. The results
indicate that the model can be used to predict the rapid variation in forces
and also due to rapid variation of N forces, this .shell is not an ideal
y
example are similar to those of the simply supported case, and are given in
Fig. 27. The shell has a PoissonYs ratio of 0.0 and is subjected to a
rr
Z(y) q Cos by
A 12x6 grid on a quadrant of the shell is used in the analysis. The plots
ment at the free edge is very amall which indicates a behavior somewhat
5.1. Conclusions
a variety of shallow shellso The model gives good estimates of the magnitude
ratio" The comparisons of results indicate that the model can be used to
in forces and moments through the shell can be predicted by the model,
general, the obtained results are sufficiently accurate for design purposes.
includes the plastic behavior is possiblea Partial loadings and the exten-
along the edge, different material properties at different nodes, and shells
For elastic shells whose coordinate lines coincide with the lines
of principal curvature, the model can be redaced to the on~ in Refo [16]"
The extensional and the flexural behavior are then defined at the nodes of
every other row, while the shear and the twisting behavior are defined at
-43-
-44-
of two rigid bars. ·The number of equations are thus reduced to one-half,
equations governing the behavior of the model are very complex. An efficient
generate and solve the equilibrium equations within a digital computer. With
gate a variety of loading cases and support conditions which are encountered
paraboloid units which are bounded by the characteristic lines of the surface,
and finally arch dams can then proceed by proper combinations of various
portions.
investigated.
generalized a's and ~·s. Solutions of shells with variable curvatures such
as elliptic and parabolic cylinders and conoids can then be easily obtained.
..1. Ang, A. H. S. and Newmqtrk, N. M., 11A Numerical Procedure for the
Analysis of ContinuousjPlates, 11 Proceedings of the Second American
Society of Civil Engineers Conference on Electronic Computation,
Pittsburg, Pennsylvania, September 1960.
7P Clough, R.· W., "The Finite Element Method in Plane Stress Analysis, 11
Proceedings of the Second American Society of Civil Engineers Conference
on Electronic Computation, Pittsburgh, Pennsylvania, September 1960.
8. Clough, R.. w. and Tocher, J .. L., "Analysis of Thin Arch Dams by Finite
Element Method, 11 International Symposium on the Theory of Arch Dams,
Southampton University,. Pergamon Press, 1964.
9. Das Gupta, N. C., "Using Finite Difference Eq_uations to Find the Stresses
in Hypar Shells," Civil Engineering and Public Works Review, . February
196L
10. Das Gupta, N. C., "Edge Disturbances in a Hyperbolic Paraboloid," Civil
. Engineering and Public· Works Review, February 1963.
-45-
-46-
15. Parikh, K.· S., "Analysis of Shells Using Framework Analogy," Sc.D. Thesis,
Massachusetts Institute of Technology, June 1962.
16. Schnobrich,·W. C., 11A Physical Analogue for the Analysis of Cylindrical
Shells," Ph.D. Thesis, University of Illinois, June 1962"
17. Soare, M., Application des Equations aux Differences Finies au Calcul
des Coques, Editions de L'Academie de la Republique Populaire Roumaine,
Bucarest, 1962.
20. Zienkiewicz, 0. C. and Cheung, Y. K., "The Finite Element Method for
Analysis of Elastic,· Isotropic and Orthotropic Slabs," Proceedings of
the Institution of Civil Engineers, August 1964.
t
I,
SS
a = 6.0"
PIN I
SS
h ::::;
v = 0,3
o. 03 11
LL ____ . ---'>--X
L~_J 2
x =0 y =0
Method
w M M
x y
4 2 2
qa /D qa qa
2 2
E = 2x10 5 kg/cm Cl = O. 019 kg/cm
w cm N kg/cm M kg
' y ' x '
Pos. Neg. Pos. Neg. Pos. Neg.
1 SS
h =
v = 0.3
0.03"
11
mlC\l I Free
l i I
L~J~x
2
x = 0 ) y = 0 x = a/2 y =0
Method ·
w M M w M
x y y
4 qa 2 qa2 4 qa
2
qa /D qa /D
·-
. Model 5x5: 0.01320 0.0269 0.1225 0.01511 0.1309
TABLE 4 .. MAXIMUM POSITIVE AND NEGATIVE VALUES OF w, NY' AND MxAT THE
MIDSECTION OF CYLINDRICAL SHELL, TWO OPPOSITE EDGES SIMPLY
SUPPORTED, THE OTHER TWO EDGES FREE
w cm Ny ' kg/cm Mx kg
' '
Pos. Neg. . Pos. Neg. Pos. Neg.
.,,
(l)
rg
~ ~
~ Q)
.-1
"'.'j
•l1 ~
•rl
a:
~0
~
().)
0
.
d
H
~
$'.:!
•rl
P-i
z
0
H
f;
u
~
0
0
t
H
0
I-;)
q
H
0
H
p::;
I'(\
0
.
H
µ.,
y
t Contribution Area
for Radial Loads
-----111- x
t ',I
,,......
'/
..
\I./
,,......
'~ -
\ /
'
,
(~
,.I'\ .~
j+l ""
\.1..1 '-IJ \1..1 'JJ ~D
,, 11'1. ,.,
j '"
\I/ ''-'
, r'\
\',I 'L-1 \V
j-1
( t'\
v ,. "'
\I/ "'"
\I.)
,. "
\I.I
, ....
\L.J \D
," ,
j-2 \I/
I I'\
\ 1..1
11'\
'IJ '"
'-!J \U
I
! I I
I I I '"
'I.)
,t-,
v (D
I
t I I
I I I , , ..,. ,.f,.
I ',I \I./ \l/
I I
I ,,
3 f '
\.I \',I - - - - l'I"\
\l.J' '"
-\:7
I")
/
,.,
2 ',I - - - - "'"
'I-' '"'
\l.I
,i-.
l/
, .... ,. ' ,,
1 ',I \.__,/ - - --- 11"\
....., -= ~
1 2 3 i-2 i i+l 1+2
tan a/2
w. ~
1.u
L
sin a/2 = Lx/2 tan cx/2 =
x
H
:x
?R
x Cosc:d2
'
wi+2j
displacement
Gosa/2
displacemen
t
\..n.
\.0
(
;
CJ\
0
I
i
/x
I
0\
t-1
&
a
2 u .. tan a/2
lJ+1
w.].- i·J+1
Cosa/2
I
i
~ 71 Mxy. ij+2
l~I/
~I
I
' / .
.....
..... ... ...
\,../
.... .....
'\
.d ..
't""\
,,,,,
0.001
0.003
0.004
0,005
0.006
-LO
q
..............
....:::t
ro -0 .. 5
01 y
....:::t
•0 0
i-1
x
....
~ 0.5
/o.642 z
LO
a = 1300 cm
b = 1800 cm
Rx = 1156 cm
-3 .. 0 h = 7 cm
-2.747 cp,_ = 68° 26'
.E-..
-2 .. 0 v = o.o
ro Uniforni Load
01
.... -LO
z~
LO
-LO
(\J /
a:l
011
(\J
I
0
r-1 0
x
1..0
-0.10
~
'- -0.05
..:::t
o:I x
01
..:::t
s
0 0
r-1
>< 0.,,041
....
~ 0.05
z
0.10
a = 1300 cm
b == 1713 cm
Rx = 1156 cm
-L5
Ry. = 1960 cm
-Ll42
h = 7 cm
-LO
CD = 68° 26 1
m x
0
·y = 51 50'
01
CD
"' -0.5
z» v ::: o.o
Unifo:rm Load
0
0 r:;
"' ~·
..
-50 y
t=l
~
...:::t
0
ro
a1
...:::t
I
0 50
r-1
z
><
.... 100
~
a :::: 1300 cm
150
b 1713 cm
cm
x = 1156'
R
-15.0
R =-1960 cm
y
h = 7 cm
-10.0
= 68° 26'
ro
01
°'x
eoy = 51° 50'
.... -5 .. 0
z» v = o.o
Uniform Load
0
5. 0
-10 .. 0
C\J ro -5.0
01
(\j
I
0 0
.-l
x
....
:;?::
x 5 .. 0
10."0
0 z
..
s:l
•rl
....
0 .. 02 a = 180 in ..
~
b = 180 'in ..
0 .. 04 c = 36 in~
h i:: 2·cr5 in ..
E = 3x106 psi
-18C z = 50 psf
..
s:l v = 0.16
......_
-rl
.. -160
ti)
,0
r-1
....
-140
z~
-120
-20
_.
s:l"
.....•rl
0
c
-M
I
e
Ul
,0 20
r-1
....
>< 33,.2
~ 40
a = 25.88 ft ..
b = 50.00 ft.
R
x = 50 .. 00 ft ..
h = 0 .. 50 ft ..
Q) 30° 00'
x =
v ·- o· . 15
E = 4.32 x 105 ksf
1(
Z = Cos by
01
I
50
Cl
-6.o
+>
.
t"+-f
............
+> -4 . o
'+;
I
Cf)
p.
•rl
,.':I:! -2.0
"'
':t.><
0
q y
-=t~ -5$0
cd
01 -1 .. 89
; 0
0
r-1
x
.....
5. 0 free z
~
a = 1300 cm
10~0
......___11. 75 b = 1800 cm
15.0
Rx ~ 1156 cm
h = 7 cm
-20 .. 0 qJ
x.
::: 68° 26'
v = o.o
1(
10,0
20.0
-27a30
30.0
-1. 5
-1.218
(\j -LO
I
0
r-1
x -0.5
0.5
FIG. 22 THE DISPLACEMENTS w, THE FORCES N , AND THE BENDING
MOMENTS M. AT THE MIDSECTION OF AYCYLINDRICAL SHELL,
TWO EDGESXSIMPLY SUPPORTED AND THE ruMAINING TWO EOOES FREE
'-71-
Roller - - - -
Hinged---
-10 .. 0
a = 1300 cm
Edge Crown b = 1800 cm
Rx = 1156 cm
-35 .. 0 h = 7 cm
Z = q Cos
rt
by
I
-25.0
'.\
\
I
-20.0
'\ I
\
-15.0
tU
a1
"'
'\
z~ \
-10.0 -
\
\
.
-5.0
0
\
\
\
\
\
Roller
,I>"'
A
/
/
..,,,.,.
---
\ /.
/
5~0
\ / I
\ ,_ ~ 8 .. 12
/
/
10 .. 0
12 x6
6 x 6 ----
-50
~
..........._
...::r 0
Ctl
CJ' 43 . 9
...::r
B
0,......, 50
>cl
... a = 1300 cm
~
100
b = 1713 cm
Rx = 1156 cm
150
Ry = 1960 cm
h = 7 cm
-20
(f>x ::: 68° 26'
qJ,,. = 51° 50'
Ctl 0 .y
01 v = o.o
.... 1(
z» 20 z = q Cos b y
·:z;L7
40
56.8
60
-~·01
(\j
Ctl -6 .. 0
CJ'
(\J
I
0
,......, -4 .. o
><
....
>< -2 .. 0
~
-10
0
l I
-
I
I
10
I
I
I
20
I
cd
01 I
"'
z:-,:..;
I
I
I
I
30
31.7
free
z
40
50
56. 8'
60
50
~
~r---_
ro r---_
01
"'
~»
40
- r-----__
~
30
20
10
A
.. 50
...........
...::t
ct1
01
0
...::t
I
0
r-1
>< 50
...
)
100
y
a = 1300 cm
b = 1713 cm
cm
x - 1156
R
R _ 1 nt:..A cm
= -.J...7uv
y
h = 7 cm
mx = 68° 26'
0
(j)y =·51 50'
v = o.o
i('
Z· =q Cos by
C\! 0
cd
01
(\J
I
0
r-1
5 .. 0
><
6.64
"'>< 10 .. 0
~
EQUILIBRIUM EQUATIONS
Alth011gh the eq_uilibriuill- eq_uations ·o;f the model are generated within
LL
t
2 1
~ ( 1 +
2 [ 2 2 2 ) 2 (u.+3· .- 2 u. l" + ui-lJ.)
Cos a./2 4R Cos a./2 L l J ·. l+ J
x . x
2
1-v t 1 ·
+ - (1 + ) - (u - 2ul.+1J" + ui.+l·J·--2)
2 4R2 Cos2 a./2 L2, i+lj+2
x .Y
2
(1-v v l+v t
1
+ 2 + Cos a./2 Cos f3/2 + 2 4R R
.x y
Cos a./2 Cos f3/2) L L (vi+2j+l
x y
- v. 2. l - v .. 1 + v .. il - (_~ \ + R vI I ) L1 (w1·+2j-wiJ.)
i+ J- lJ+ 1J- _dos a./ 2 Y Cos a 2 Cos f3 2 x
x
2
v d (
1 - ) t 1
- Lr:-~ .wi+lJ"+l - wi+lJ"-1 + 2 3 (wi+4J" - 3~i+2J" + 3w. · - w. 2j)
X· y y 4R
x
Cos a/ 2 L
x
l J l- .
12
t (1-v · 1 l+v 1 1
+ 4Rx Cos a,/2 2.. Cos a/2 + 2 Cos f3/2) L L2 (wi+2j+2 -.2wi+2j + wi+2j-2
xy
2
- wl.J+
.. 2+. 2w l.J.. - w..
lJ -
2)] ~
1-v
Eh x 0
: (A-1)
-76-
-77-
1 1 2
x y [ 1 t 1
-2- ( 2 + 2 2 ) 2 (v · ' 3 - 2 v · · 1 + v iJ' -1)
Cos ~/2 4R · Cos ~/2 L lJ+ lJ+
y y
2
(1-V V l+V t ) l
+ 2 + Cos a/2· Cos ~/2 + 2 4R R ·Cos a/2 Cos ~/2 . 1 1 (u.l+ l'J+ 2 - u.l- l'J+ 2
x y x y
2
1..,. v d t 1
- -L L ...--L ( w. 1 . 1 - w.. 1 . 1) + 2 -3 (w. · 4 - 3w · ' 2 + 3w · ·
l+ J+ i - J+ 4R Cos A/2 L lJ+ lJ+ lJ
xy x y f-" y
2
) t (1-V 1 l+V 1 ) 1 (
- wij-2 + 4R Cos ~/2 2 Cos ~/2 + 2 Cos a/2 2 wi+2j+2 - 2wij+2
y 1 x1 y
2
+ W.
. l- 2'J+ 2 - W. .. -wl..- J
l+ 2'J + 2wlJ
2.)] +~
1-V
y 0
(A-2)
2 2
l - v d 1 "". v 2d t 1
- r:;r;- I:- (vi+lJ. - vi-lJ.) + r:;r;-L 1 wiJ. + 2 -3· (u. 3·
x y x x y x y 4.R Cos a/2 L i+ J
x x
-78-
2
t 1
- 3u. . + 3u. . ... u. .) + - . ( v. . - 3v. . 1 + 3v. . 1 - v. . )
i+ 1 J i- 1 J l- 3J 4R Cos 2 ~/2 L3 lJ+ 3 lJ+ lJ- lJ- 3
y y
+
t!2 v I
4Rx Cos a/2 IJ;os ~/2 +
1-V
2
1
(Cos a/2 + Cos
1
~/2)
J 1
L2 (ui+lj+2 - 2ui+lj
1
x y
2
t l
+ 2v. .
1
,... v .. . ) +
2 1 2
=1+ (w. 4 . - 4w. 2 . + 6w. . - 4 w. 2 .
lJ- i- J- 4 Cos a/2 · L i+ J i+ J lJ i- J
x
2
t l
+ w. .) + . ~ ( w. . - 4w. . + 6w. . - 4 w ... 2 + w. . 4 )
l- 4 J 4 Cos 2 ~/2 L~ lJ+ 4 lJ+ 2 lJ lJ- lJ-
y
2 2]
t 1- 2V l-V l l .
+ 1+" i_Cos a/2 Cos ~/2 + 2 (Cos a/2 + Cos ~/2) _ 2 2 (wi+2j+2 - 2wi+2j
1 1
x y
· 2 + 4 w.lJ· - 2w.lJ-
+ wi·+2J·-2 - 2w.lJ+ · 2 + w.l- 2·J+ 2 - 2w.l- 2·J + w.l- 2·J- 2)}
2
l-V
- - -· z
Eh
0 (A-3)
APPENDIX B
the entire process of generating and solving the set of equilibrium equations
· In the first program the shell dimensiqns and properties are read
to the computero The equilibrium equations are then generated and stored
given set of boundary conditions are made available to the computer by means
the FUNCTIONS or the SUBROUTINE, the same program can be used for a variety
the equilibrium equations in the least numbers of locations, the grid points
coefficients of displacements between the first and the last non-zero columns
that pointo
In the second program, the equilibrium equations are read back from
the auxiliary tape to the computer one at a time. Gauss elimination method
The solution is obtained by back substitution, starting with the last equation
in the matrix and the results are then stored on an auxiliary tape. The
maximum number of equations that can be solved depends on the storage capacity
-79-
-80-
of the computer and the maximum.band width (the band width is defined as the
number of terms between the first and the last non~zero columns in an
equation) of the equations"
In the third program the results are read back from the auxiliary
tape to the computer. The strains, the forces, and the moments are computed
Read Shell
Dimensions
& Properties
Number the Do
Grid points Elimination
Process
,---1
1 Strain I Compute
Generate FUNCTIONS Strains, Forces,
-· - _ _J
nn No and Moments
All Equations?
Equation - ----,
I Loading I
L GUBl~OUTINE _J
I
CJ
Print !-'
Back Results
Substitute
out
9-1-1971
P. V. Banavalkar
J. E. Parker
Recommended Citation
Gergely, Peter; Banavalkar, P. V.; and Parker, J. E., "The analysis and behavior of thin-steel hyperbolic paraboloid shells" (1971). Center
for Cold-Formed Steel Structures Library. Paper 19.
http://scholarsmine.mst.edu/ccfss-library/19
This Technical Report is brought to you for free and open access by the Wei-Wen Yu Center for Cold-Formed Steel Structures at Scholars' Mine. It has
been accepted for inclusion in Center for Cold-Formed Steel Structures Library by an authorized administrator of Scholars' Mine. For more
information, please contact weaverjr@mst.edu.
Department of StructuTal Engineering
School of Civil and Environmental Engineering
Cornell University
by
Peter Gergely
P. V. Banavalkar
and
J. E. Parker
ABSTRACT i
NOMENCLATURE 1
I. INTRODUCTION 7
1. HYPAR ROOFS 7
2. LITERATURE REVIEW 9
D. EDGE MEMBER 43
5. BOUNDARY CONDITIONS 54
6. SOLUTIONS OF EQUATIONS 56
7. STRESS ANALYSIS 57
A. DECK STRESSES 58
B. BEAM STRESSES 59
IV. GENERAL COI.1PARATIVE STUDY 61
1. INTRODUCTION 61
3. DECK PROPERTIES 94
A. SHEAR RIGIDITY 94
B. THICKNESS OF THE CORRUGATED DECK 95
C. NUMBER OF DECKS 96
4. BOUNDARY CONDITIONS 97
5. LOADING 100
A. EDGE MEilBER WEIGHT 100
VI. INSTABILITY ANALYSIS OF HYPARS 104
1. INTRODUCTION 104
2. INCREMENTAL MATRIX FOR DECK AND EDGE HHMBERS 107
3. CHECKING OF THE INCREMENTAL MATRICES 108
4. INSTABILITY OF HYPARS Ill
S. DETERMINATION OF THE BUCKLING LOAD 112
6. NUMERICAL RESULTS AND THEIR DISCUSSION 114
7. DECK BUCKLING BY THE ENERGY HETHOD 120
8. STABILITY OF ISOLATED EDGE MEMBERS 123
VII. EXPERIMENTAL INVESTIGATION 126
1. INTRODUCTION 126
2. FLAT SHEAR TESTS 127
3. SADDLE-SHAPED HYPAR TESTS 131
4. NEDIUM-SCALE INVERTED UMBRELLAS WITH
VERY STIFF EDGE BEAMS 134
5. INVERTED UMBRELLAS WITH VERY FLEXIBLE
EDGE BEAHS 135
6. DISCUSSION OF EXPERUIENTAL RESULTS 137
VIII. SUMMARY, CONCLUSIONS AND RECOMMENDATIONS 145
BIBLIOGRAPHY 152
APPENDIX A. DECK PROPERTIES 158
Page
APPENDIX B. STIFFNESS MATRICES 166
i
stresses, and the stability of hypars depends greatly on the
shear rigidity of the thin-steel deck. This property must be
evaluated experimentally for each combination of decking, con-
nections to edge members, and seam connections. Furthermore,
in the case of hypars the deck is warped and thus the shear
rigidity may be different from that of an equivalent flat dia-
phragm. 2) The deck may buckle due to the shear stresses. and
the buckling load must be evaluated for highly orthotropic
shells. 3) The design of thin-steel hypar structures is gen·
ii
buckling of hypars; c) Sixteen flat shear tests to determine
the shear rigidity of the decks used in the hypar tests; d)
Twelve saddle-s~aped hypar tests (5 ft by 5 ft in plan) with
various rise-to-span ratios to evaluate the effect of rise or
warping on the shear rigidity and to study other factors such
as partial loading and single versus double layered decks.
Photos of the various types of tests are shown in Figs. 7.1 to
7.5. The experimental program is described in detail in Chap-
ter VII.
Prior to the main test program, several small-scale (2 ft
by 2 ft) four-quadrant tests and medium-scale single-quadrant
tests were also conducted. These tests were however discon-
tinued because of the severe scaling effects in the case of
the small-scale models and the violation of the symmetry condi-
tions in the case of single-quadrant experiments where the
neighboring quadrants were missing. Nevertheless, these tests
produced useful qualitative information and experience with
manufacturing and testing thin-steel hypar structures.
The edge members of the umbrella-type specimens were made
of tuoular members since this afforded easy connection of the
warped surface to the straight edees. The decking consisted
of single or double layers of standard corrugated panels. One
layer was connected to the edge ~enbers with sheet metal screws
at various spacings. The seam connections between the panels
were also made by means of sheet metal screws. In the case of
shells with two layers, the top layer was connected to the bot-
tom layer in a similar manner.
iii
The medium-scale umbrella models were loaded using air bags
under each of the four quadrants. The saddle-shaped hypars
were loaded with sand, whereas the small-scale models were load-
ed through loading pads and suspended weights.
The following are the principal conclusions of the experi-
mental part of this investigation:
The effective shear stiffness of the cold-formed deck and
the rise (or curvature) of the structure are the most important
factors influencing the behavior of hypars. For low shear stiff-
nesses and for small rise-to-span ratios the deflections may be
large, the bending stresses tend to increase relative to the
membrane stresses, and the possibility of deck buckling increases.
As in the case of flat shear diaphragms, the shear stiffness
depends strongly on the seam and edge connections.
The increase in shear stiffness due to the addition of a
second layer of deck was found to be only about 1/3 if the
second layer was connected only to the first layer and not dir-
ectly to the edge members. Similarly, the deflections of a
double-layered shell are more than half of those of a correspond-
ing single shell. If the two layers are interconnected with
sheet metal screws (on an 8 in. grid in the present saddle-
shaped hypar tests), the deflections are further reduced by
about 10 to 20%, depending on the rise ratio.
A particular problem of certain types of hypar structures
is the deflection of unsupported outside corners (see Fig. 1-2,
page 212). The membrane shear cannot carry the load over such
flat corners and thus considerable bending and deflections may
develop. The tests showed that the bending stiffness of the
iv
edge members has a great effect on the corner deflections, in
fact, they indicate that the design of the edge members in hypars
with flat corners is usually governed by deflection limitations.
The measured bending strains in flat saddle shells (rise-
to-span ratio of 1.8). was much greater than the bending in hypars
with greater curvature (rise ratio of 1/3). The membrane theory
is insufficient for the design of flat hypar structures. How-
ever, the design of the connections (seam or edge) may be based
on the shear forces obtained from the simple membrane theory.
Several single and double layered saddle-shaped models
were tested under partial loading. Since such loads must be
carried mainly by bending of beam strips along the deformations
of single decks, relatively large deflections were noted. The
deflections under the 8 in. by 8 in. loaded area were about
three times greater in the single decks than on the double-
layered structures.
Since the effective shear rigidity of the deck is of para-
mount importance, the effect of curvature (warping) on it is
an important question. The effective shear rigidity of various
deck, edge member, and connection configurations are determined
by tests on flat diaphragms. The comparison of the measured
deflections for saddle hypars with various rise-span ratios and
the evaluation of the effective shear rigidities backwards from
the measured deflections indicated that the shear rigidity is
reduced by about 20% due to the warping effect.
The buckling of the deck is one of the design factors.
For small rise-span ratios and for low deck shear rigidities
the deck may buckle. As an example, a 12 ft by 12 ft model
v
having a single layer 24 gage corrugated sheet deck buckled
at a uniform load of 70 psf (see Fig. 6.14). This model had
relatively stiff edge members (3 in. dia. tubular sections).
The corner deflections remained linear with increasing load
beyond the buckling load.
The buckling load of double-layered structures is much
larger than that for single deck shells. A model, similar to
the above but with two layers of 28 gage standard corrugated
decks, did not buckle up to a load of 145 psf, when the test
was discontinued.
The major part of the ~~lytical investigation consisted
of two finite element approaches for the calculation of deflec-
tions, stresses, and instability. In addition, two simple
methods were developed for estimating the deck buckling load
and the buckling of the compression edge members, which would
suffice in preliminary designs.
Two types of finite elements were used: curved shallow
shell elements and flat elements. The details of the analysis
are described in Chapter III. Both approaches were verified
by comparisons with existing experimental and analytical results.
The stiffness of the eccentric edge members were properly
accounted for in the mathematical representation of the struc-
ture. The connection of the decks to the edge members may al-
low rotation about the axis of the edge members and movement
normal to the edges due to slip at the connections. These pos-
sibilities were also considered in the analysis.
The instability of the decks was studied with the help of
the incremental stiffness matrix approach. The effective stiff·
vi
ness of the system is reduced due to the in-plane forces in
the deck. The in-plane forces depend on the deflections of
the shell and to obtain the buckling load, the eigenvalues of
a large order system need to be evaluated. In the present
study the load incrementation method was used instead. The
effect of the in-plane forces was evaluated iteratively at
successive load increwents. The buckling load is obtained
from the nonlinear load-deflection curve, (Fig. 6-6, Page 276).
The comparison of the results of the flat element and
the curved element approaches reveals that both give good re-
sults for shells supported around the perimeter. However, the
flat element method gave better results in the neighborhood of
unsupported flat corners.
The analysis of the structures tested in this and in other
studies confirmed the conclusions of the experimental part of
the investigation. The stresses in most types of hypars are
low and the design is usually controlled by deflection or
buckling linitations.
The relative stiffness of the deck and the edge members
is an important factor. For stiff edge members the deck tends
to bend between opposite edges, whereas in the case of flexible
cantilevered edge members the shell partially supports the edge
members. Analysis of a structure including the weight of the
edge members indicated that this effect may have to be con-
sidered in the design of hypar structures.
The analysis of buckling of hypar decks showed that the
buckling load of double-layered shells is three to four times
vii
greater than that of single decks. The predicted buckling
loads compared well with experimental or previous analytical
evidence. The buckling load does not depend much on pre-buckling
deflections, however it depends on the axial stiffness of the
edge members.
The finite element analysis was also used to calculate
the deflection of an unsymmetrically loaded inverted umbrella
structure. The results, which compared well with experimental
data, showed that these deflections are about four times great-
er than those due to symmetric loading. This increase of de-
flections obviously depends on the type of structure; in this
case much of the flexibility was due to the bending of the
central column of the umbrella structure.
Since the instability analysis of hypars by the finite
element method involves considerable amount of computer capac-
ity and expense, approximate methods were developed for the
calculation of buckling loads. The buckling of the compression
edge members was studied by isolating them from the structure.
The instability of columns loaded by tangential axial forces
that remain parallel to the member during deflection was evalu-
ated. The results are tabulated in Fig. 6-13, page 284.
The buckling of hypar decks was also investigated by the
energy method (Section VI-7. The resulting equation has to be
minimized to get the critical load; this can easily be done
with the help of a computer. This approach is much simpler
than the finite element instability analysis and is preferable
in preliminary designs.
viii
A few buckling analyses of cold-formed hypar shells
showed that the critical load for double-layers is about three
to four times greater than a shell with a single deck.
The finite element analysis computer program will be made
available to designers by the American Iron and Steel Institute.
ix
-1-
GLOSSARY OF TERMS
ey Strain in y-direction.
n Number of interconnected decks.
e Twist rotation of a beam, radian.
e' Rate of change of twist rotation,
radians/inch.
e1. Angle of twist of a beam at the i-th end, radian.
elX
! Rate of change of angle of twist at the i-th
end, radians/inch.
Rotation about x-axis, radian.
Rate of change of rotation, radians/inch.
Rotation about y-axis, radian.
Eigenvalue for the critical buckling load.
\) Poisson's ratio.
\) \)
xy' yx Poisson's ratios in x and y-directions respec-
tively for the equivalent orthotropic plate.
0
x
Stress in x-direction, lb/inch 2 .
o
x '' o
y ,' t
x 'y' Stresses measured along x' and y' axes.
Stress in y-direction, lb/inch 2 .
Shearing stress, lbs/inch 2 .
Total potential energy of a system, inch-lb.
CHAPTEP I
Ii·JTP.O DU CTI ON
I. 2. LITERATURE REVIEW
isotropic hypar with moment-free rigid edges, with the edge mem-
bers of infinite axial rigidity but negligible bending rigidity
6
in planes tangential to the shell. Apeland and Popov reduced
these two equations to a single eight-order differential equa-
tion. Using Levy-type boundary conditions (with two opposite
edges knife edge supported) they tried to establish the effect
of edpe disturbances in the sa~e way as that for cylindrical
shells. Their important conclusion was that the effect of the
edge moment does not die off very rapidly in the case of hypar
shells.
The formulation in terMs of widdle-surface displace-
ment u-v-w, results in three coupled partial differential equa-
tions two second-order (u-v) and one fourth-order (w). Sal-
1
buckling. The study in Ref. 19, indicates that for a very flex-
ible edee member, the possibility of overall buckling- nrior to
the deck buckling does exist. The erroneous conclusion of ref.
23 was reached because of two reasons:
(i) A. non-compatible displacement field between the shell
1.4 ACKNOWLEDGMENTS
This investieation was sponsored by the American Iron
and Steel Institute. The financial assistance and the technical
guidance provided by the Institute is gratefully acknowledged.
We thank the Task Group on Shell Roof Structures, and especial-
ly its chairman, Dr. J. B. Scalzi, for their interest in this
project.
Special thanks are due to Dr. George Winter, principal
investigator of the project, for his great interest in the
investigation and for the many wise suggestions throughout the
work. His insight into structural behavior often steered the
study back from tangents and away from pitfalls.
The nonconservative stability analysis was carried
out by Dr. R. Huskat. His and Hr. A. Banerjee's help with the
experimental program was very valuable.
CI!f PTEP II
DECK PR0PE'RTIES
II.L INTRODUCTION
A material in which resistance to mechanical actions
is different in different directions, is called anisotropic.
Orthotropy is a special case of anisotropy, where the body
possesses elastic properties which are symmetric about three
mutually perpendicular planes.
Orthotropy can be further classified into two cate-
gories namely, natural and reometric. The natural orthotropy
is a result of the material property itself. A classical
example of natural orthotropy is timber, where the modulus of
elasticity, along the direction of its prain in tension, is
su1'stantially hieher th.an the corresponding modulus in the
direction perpendicular to it.
In the geometric type of orthotropy, the difference
in elastic properties in the perpendicular direction, as shown
in Fir. 2.1, is due to the reometrical confiruration of the
structural element even though it is made t~ of an isotropic
homogeneous ~aterial. Different types of decks belonging to
this category are shown in Fig. 2.2.
In both the cases, the definition of elastic con-
stants in two mutually perpendicular directions is required.
However, geometric orthotropy is of particular interest for
the hypar structure dealt Hi th in this 1•rork.
-18-
-19-
a
£
y
a + ily 2-lb
2-lc
TXY 0 0 (l-:-vxyvyx)Geff
The terms vxy and vyx represent the coupling effect o the
actions (stress or strain) applied in two perpendicular direc-
tions. From Fig. 2.2, it is obvious that these coefficients
(vxy and vyx) cannot be equal. For example, consider the
cellular deck shown in Fig. 2-2c, where a uniform stress
applied along the bottom plate of the deck in the x-direction
will produce a negligibly small strain in the y-direction, in
the hat portion. Whereas a uniform stress applied over the
entire cross-section along the y-direction will produce a
-21-
E
= E
_!. " 2-4
y
2-7
2-8
V' = l2 0
x
'e:: ' +
x
!2 0 y 1 e:: y ' + !2 T x ' y 'y x I f I 2~1ob
ing Nxy = Nyx (Section III.2 B), is given by t~e \j9~e 5 9ion
Exyt = G' = a. G nt 2-11
where n is the nurnher of interconnected decks, 1;'. ~, 'the thick-
ness of each deck and G is the shear modulus O:f t~( l'natel'tal.
Cl is the relative shear rigidity coefficient P:ivi~; 'the l'atio
of the shear stiffness of the actual orthotropiC ~oci sfStem
and the isotropic plates of thickness nt. In t~~ ?a~e of an
isotropic plate, the vertical load on a hypar i~ b?rtly ~arried
for both single and double decks with varying rise to span
ratios.
It was found that for the lower values of a (<0.12),
the center deflection of all-supported hypars was proportional
to the value of a. Both single and double deck hypars were
tested for three different rise to span ratios (1/8, 1/5, 1/3).
The theory developed in Chapter III correctly predicts the
effect of a and rise independently on the central defl~ction
III.I. INTRODUCTI0N
qi =u + v 3-1
= _a_
()/j..
(P+V) = 0 3-2
1
-31-
-32-
-6V
P. = - - 3-3
1 aL\. 1
K' •. = 3-5
1)
model.
The linear elastic stiffness analysis consists of
four important steps:
(1) The formulation of the element stiffness matrix.
(2) The formulation of the master stiffness matrix for
the entire structure by assembling individual elements.
(3) The solution of Eq. 3-4 for the given boundary con-
ditions and loading.
(4) The interpretation of the deflected shape and the
computation of stresses and forces.
Two alternative stiffness formulation methods are
studied here:
Method 'a': The use of rectangular curved elements. based on
shallow shell theory.
Method 'b': The actual curved shell surface is approximated by
the assemblage of flat rectangular elements.
III.2. ELEMENT STIFFNESS
The elements rectangular in plan are selected. These
elements are very simple to formulate and for the structure
under consideration, their limitation of application to the
rectilinear rectangular boundaries, is not considered to be of
any serious consequence.
As shown in Eq. 3-lc, the element stiffness matrix
can be derived from the strain energy U of an element, ex-
pressed in terms of an assumed displacement field.
A. DISPLACEMENT FUNCTIONS
The displacement fields assumed for the analysis are
-34-
as follows:
1
u = ab [(x-a)(y-b)u 1 - x(y-b)u 2 + xyu 3 - (x-a) yu 4 ] 3-6a
V = ao
1
((x-a)(y-h)vl - x(y-b)Vz + xyv 3 - (x-a) yv 4 J 3-6b
+ (3ax 2 -zx 3 )(3ay 2 -2y 3 )w 3+(a 3+2x 3 -3ax 2)(3by 2 -2y 3)w 4
- ax(x-a) 3 (b 3+2y 2 -3by 2 )ey 1 -a(x 3 -ax 2)(b 3+2y 3 -3by 2 )ey 2
- a(x 3 -ax 2)(3by 2 -zy 3)e y.3 -a(x-a) 2x(3by 2 -zy 3)e y 4
3
+ b(3ax 2 -2x 3)(y 3 -by 2 )ex 3+b(a 3+2x 3 -3ax 2)(y -by 2)ex 4
2 2 2 2
+ abxy(x-a) (y-b) exyl+abxy(x -ax)(y-b) exyZ
2 2 2 2
+ abxy(x -ax)(y -by)exy 3+abxy(x-a) (y -hy)exy 4 ] 3-6c
aw · e = - d'».T : e
ex = a-Y 3-7
• Y ai, xy
One may also note that the inplane rotation about the z-axis
is being omitted in this formulation.
The function for w, normal bending displacement, is
of the cubic order. The terMs corresponding to the degrees of
freedom w, ex and ey are obtained by the cross product of the
corresponding tcrws for the beam function in x and y-directions.
In order to represent the constant strain corresponding to the
. .
tw1st1ng a2w 1.
term a·xay' . e.' t h e terJTl ' xy ' ' t he a <ld.1 tiona.
. 1 d cgree
Z -- C xy
Arr 3-8
u x 3-9a
'
v,y 3-9b
2C
Yxy = u ,y + v
,x - -
AB
w 3-9c
-w 3-lOh
,yy
-38-
ZK xy = - 2w ,xy 3-lOc
Strain Eri.ergy
The strain energy for a typical element consists of
two parts: the mernbrane and the bending strain energy.
U = Ub + LT
m 3-11
3-12
Ny = Elt
T"
£'yt 0 £y 3-13
N 0 0 Exyt Yxy
xy
The method of col'lputation for the above mentioned elastic con-
stants is discussed in Chapter II.
Using the strain displacement (Eq. 3-9, 3-10) and the
stress-strain relationship (Eq. 3-13), the membrane strain
energy (Fi~:'· 3.2) of an element can be expressed as follows:
1 2 3
b a 2
1
1J
m= 2 0J 0J {Extuz,x + Eyt ·~ 1y + 2E 1 tu,xv'y
4 s 6
+ Exyt[u2~Y + v2
'x
+ 2u,yv,x]
7 8
c
- 4Exyt(AB) w [u, y + V,X)
9
2
+ 4 Exyt(~B) w2 } dxdy 3-14
-39-
p 0 0 -Zw 'xy
xy Dxii
-
The bending strain energy is r;iven by 38 ,
u . 1
warp = -2 Er 3-19
C. FLAT ELEf·.7.ENT
schematically as follows:
[ K]
m [K]~m u
v
[K]curved = ------- -------- 3-20
.,, [Y.]bb
[K]~,) Ill 1°•.T
+[K]bbm
u
[K]m 0
v
[K]flat = ------- -------- 3-21
0 [K]bb 1'.T
w = i: i:
m=l,3,S, ... n=l,3,5, •..
amn s in m; x. s in n JY 3- 2 2
where~
a mn 4
3-23
+ n 4DY)
b
where,
3-24
3-26a
3-26d
Using Eqs. 3-25 and 3-5, one can obtain the conventional beam
stiffness matrix.
However 1 the warping restraint is of practical im-
portance, particularly for thin-walled open sections. To in-
clude the warpinr effect~ the displace~ent field for e is
assumed to be of the same form as that of v and w58 .
1
3 [(a 3 +zx -3ax 2 )el+(3ax -zx )e2+ax(x-a) e~1
3 2 3 2
e =
a
+ a(x -ax 2 )e~ ]
3
2 3-27a
0, = ae 3-27b
x ax
beam.
1 0 0 0 +Z c -Y c 0
-u
7
1 v
0
. ~
s 0 0 0
1 .. y 0 0 0 w
s
p.] s = 1 0 0 0 ex 3-29
1 0 0 ey
0 1 0 ez
1 A
-46-
= rP-] s 0 J 3-30
0 [A.] s
F. ELASTIC SPRING8
In order to simplify the mathematical solution without
undue loss of gencrality 1 certain structural elements are ideal-
ized in the for~ of concentrated spring stiffnesses. For ex-
ample, the central column in the case of an umbrella shell, if
idealized as a physical roember with its end points, will not
only create an additional node point for the master stiffness
matrix but v ill also disrupt the regularly arranr:ed griil pattern
1
-a -
3-32
12EI lZEiz
5sY v S S GJ
= a L ' SZ = a ; T = a
the surface.
b) Flat Element
The middle surface of the shell is approximated by a
series of flat plates. The geometrical approximation of the
actual surface needs three important steps:
(1) Definition of Surface -
As pointed out in Chapter I, different hypar struc-
tures can be built with various combinations of the basic units
(Fi~. 1. 2). It is necessary to express the equation of the
generated surface with reference to the chosen global axis.
The general equation of a structure using the hypar units can
be expressed as,
c.
Z
= A.B.
_1_ (x-i1.) (y-y1.) 3-33
l 1
Figs. 3.lSa and 3.1-Sb show two structures and also give the
-49-
zx
OX' = 0
J1+Zf Rz
OY' = 0
1 y 3-34
-z
p;
-z
'R
oz I y 1
=
l+Z x2+z y2 l+Z x2+z y2 j+z2+z2
x y
where
zx = AB
£__ y
o
and x0 .
[T]eli, = 3-35
0 p,] 2
6x6
li,x ty 1z R.x 1y 0
[A] 1 = x my
1'l mz [A] 2 = mx my 0 3-36
n x ny nz 0 0 nz
II I. 4 LOADING
1·;
·1 ab/4 8 x1 - ~b 2 /24
ab/ 4 8
xz ab 2I 24
ab/4 8 -a 11 2I 2 4
x3
at/4 2
= 8 x4
= -ab /24 3-38
2
a b/24 8 xyl a 2 ~ 2 /144
2
a b/24 8 xy2 -a 2b 2/144
a 2b/24 8 xy3 a 2b 2/144
2
-a b/24 8 xy4 _
3 2~2/144
Boundary
Conditions u v w ex xy
Hinge 0 0 0 0
Knife-edge 0 Q 0
Fixed 0 0 f) 0 (I
"'·'
Free
Symmetric 0 0 0
tive displacement betueen the deck and the edge member, normal
to the boundary (Fig. 3.20b). This type of a connection can
result because of an oversized hole, loosely connected screws,
or due to tearing of the deck. Depending upon the continuity
achieved between t11e deck and the edge members, different
values of fixity coefficients are used. Tr and T8 represent
the torsional and the horizontal fixity coefficients, respect-
ively.
In the case of the moment-free deck to edge member
connection, Tp ~ 0. The edre roei:'ber stiffness matrix is modi-
fied by multiplying the columns and rows corresponding to the
twisting degree of freedoms (ex and axy) by TF.
The problem is furtter complicated by the eccentric
connections. As sho1m in Fig. 3.2la, even with a discontinuity
of the rotational degree of freedom, twisting can still be
introduced in the edge member because of eccentrically trans-
ferred vertical or horizontal load. This problem is not solved
satisfactorily. By the ~ethod of fixity coefficients, the
twisting action introduced by these eccentric forces is elimin-
ated. There is no w-0ment transfer when t~o elements are inter-
connected by means of hinges and this results in the local re-
lease of the member forces. This for~ulation does not include
the effects of these releases. The details for the incorpor-
ation of these local member releases are given in Pefs. 62 and
6 3.
II I. 6 snu:TIONS OF EQl1ATTONS
[K] [T]R [~] by [T]~. transforms the forces to the shear center
or the centroid of the beam.
Because of the mathematical idealization, certain
difficulties are encountered. A beam with an eccentricity in
the z-direction is shown in Fig. 3.23. The variation of tl:e
axial forces is shown in Fig. 3.23b and 3.23c. Since the forces
are balanced at point 0, the axial fotce also contributes to
the equilibrium of the moments at point 0. This results in the
inequality of the moments along the axis of the beam PQ. The
problem becomes particularly critical in the case of rapidly
changing axial force and a deck with strong bending rig~di ty
(e.g., concrete hypars). No suitable solution is found for
this problem as of this moment. In the absence of definite
guidelines, the deflected shape of the structure should be used
to decide the sign of the moment.
Experience shows that the bigger of the two reoments
Ct.\sp or MoQ) is always in conformity with the correct deflected
shape of the beam. The difficulty experienced in computing the
stresses of an eccentric edge mewber is one of the shortcomings
of using the nodal points only along the shell surface.
The results obtained by this stiffness analysis are
compared with experimental and the available solutions in the
literature in Chapter IV.
CtTAPTI'P IV
A GENERAL COf'PARATIVE STUDY
-61-
-62-
method 'a'. For Struc. 'l', the difference in the central de-
flection is about 0. 8~; whereas for Struc. '2', the d; fference
is about 1. 309ii. The central deflection for Struc. '2' is only
0.5% on the higher side of tho deflecti~n obtained by Pecknold
and Schnobrich.
The correlation obtained by methods 'a' and 'b' 1s
excellent for these two structures. However, it wust be
pointed out that both of these struct~res~ taken from refs. 17
and 20, are supported all-around. 'From the practical point of
view, these structures are only of academic interest. The
boundary conditions such as free edgest encountered in an UJll-
brella shell (Fig. 4.2), provides a more critical test for the
comparison of the di ff er en t P1ethods.
It ·was not practical to compare tnethods 'a' and 'b'
for all the exampl~s, therefore only a selected number of struc-
tures were chosen for coJll.parison (Struc. '6' and '9' were
'
used). Struc. '6' is a swall scale concrete model. In this
structure, the stiffening edge members are located eccentri-
cally~ on top of the shell. The idealized edge members are
considered eccentric only in the z-direction (see Fies. 4.2
and 4.5).
Struc. '9' is also an umbrella shell hypar with 28-G
double layered standard corrugated decks placed perpendicular
to each other. Here the edge wember is connected eccentri-
cally to the deck with the deck on top. In the case of Struc.
'6' there is full fixity between the edge member and the shell,
whereas in the case of Struc. '9', the connection between the
.-6 5-
check was also noted when working with ;the computer program formu-
lated by Park~r 19 •
In the case of Struc. ''6', the deflection profiles obtained
by rriethods.'a' and 'b' along the compression member are reason-
ably close. However, these two nethods give entirely different
. I , .
computational work.
The strain-displacement relationships for the curved
·element are dependent only on the twist curvature ~B irrespective
of the shape of the actual structure. To explain this further,
consider only the quadrant oabc of a structure of Type I, Fig.
4.1. One could build two cantilever hypars from this quadrant.
The first structure would have eclpes oa and oc fixed whereas
edges ab and be would be free. In the second structure, the
fixed ancl the free edges would he interchanged. If both these
structures are subjected to the same loadinr conditions, method
'a' would give identical deflections and absolute values of the
stresses.
The solutions bv, roethod 'a' for Strucs. 'l' and '2'
did not show any advantare of using a 16-term displacement
function for w-displacernent, ,.,hich ensures the slope compati-
bility normal to the boundaries of the adjoining element as
against the non-compatible 12-term polynomial used in °.ef. 19.
The solution obtained for Struc. '2 1 ~ith the inclu-
sion of complete rigid body modes 20 and that obtained in this
study , without the incl us ion, . did not show 111uch of a differ-
ence (Fir. 4.6). To study tte effects of inclusion of rigid
body modes further and also to evaluate the differences in the
solution usinr 16 or 12-tcrm polynomials for w displacement,
Struc. '15' was analyzed. The results are plotted in Figs. 4.37
and 4.38. It may be worthwhile to note that in this structure
the rise to span ratio is rather high for it to be considered
as a shallow shell (see Section V.2). The deflection profile
-70-
across the diagonal ob shows that there is practically no dif-
ference between the solution obtained by the use of a 12-term
polynomial for the normal displacement w, and the function used
by Banavalkar. The maximum difference of 2% is seen in the
corner deflection ob (0.090 inches by the present method and
0.092 inc~es in Ref. 19). Even the u-v displacements all over
the shell, obtained by the two methods were within 0.5% of each
other. The striking similarity in the results tends to confirm
the conclusion that both methods give the same results for the
uniformly loaded hypars. This view is also shared by Pecknold
and Schnobrich 21 . The comparison nay not be as accurate for
unsymmetrically loaded hypars where the 16-term displacement
function for the normal displacement w would possibly give bet-
ter results.
However, the comparison with results reported in Ref. 21
shows a difference both in the deflections and stresses (Figs.
4.37, 4.38). Though the deflection profile and the stress
variation are alike, the added flexibility of the curved ele-
ment with the inclusion of rigid body modes is apparent in Fig.
4.37, where the corner deflection is nearly 60% larger than the
one obtained in this study as well as by the method used in Ref.
19. Though the solution obtained in Ref. 21 used a 12 x 12 grid
size as against a 8 x 8 grid size used in this investigation, it
is not believed that the difference in results is due to refin-
ing of the grid size.
Analysis of the same structure by method 'b' using flat
elements, results in the corner deflection ob (0.123 inch)
being nearly 33% larger than that given by the method 'a'. As
-71-
pointed out el:;lrlier, the defiections by. wethod 'a' are given
normal to the surface whereas for the Pethod 'b' they are in
the global axes. However, this does not affect the corner de-
flection ob. !lore over' the deflection profiles along the com-
pression mewber oa and the tension member ab, are different for
the two methods. It is quite interesting to note that both of
these methods~ which give close results for edge-supported hypars
(see Table IV-3), could differ in the case of this structure
(Fig. 4.37). The inclusion of t~e rigid body modes in the
solution seerns to account for the correct behavior of the flat
corner but since no cowparative results - with experiments - are
presente d 21 , i.t
. . not poss1. b~ 1 e to comment on t h e va 1.1 d.1ty o f
1s
the method in Ref. 21.
the top of the hat, do not show good correlation.with the exper·
-75-
Even though part of the vertical load near the column is car-
ried by th~ concrete shell, in order to satisfy the static
equilibrium·for the vertical load it appears that the experi-
mentally measured forces in the compression rib are quite low.
The axial and the bending stresses are measured along
the diagonal ob at an angle of 45° with the x and y axes (Fig.
4~15). The measured axial stresses show excellent agreement
i..ti th the analytically calculated value of 72 psi. An important
point :to note is that the calculated and the experimental, values
are about 34'% higher than those given by the rnembrane theory at
a load of 4.0. 9 psi, the reasons for whicl:i. are not readily
apparent. The values of bending stresses are very low and arc
not compared here. The variation of the shearing force is
plotted all over the shell for both the structures. Though
there are minor differences in the shape of variation of shear-
ing forces, two important observations can be mad~. The values
of the shearing forces over a substantial portion of the shell,
are larger than those given by the membrane theory. The shear-
ing force near the colu~n is nearly twice as large as that
given by the membrane theory. Tlds sudden increase in the
. shearing force clearly indicates that the shell participates
in tr:nsrnitting a certain portion of the vertical load. The
same behavior is also noted in Struc. '5'.
Struc. '8' is .the same as Struc. 1
6' but it is sub·
jected to an unsymmetrical load (Fig. 4.20), where half .of the
structure is loaded uniformly. Only half the structu;re along
the line ·cf. €Fip. 4. 20) is analyzed, using 1.6 elements in each
-79-
·. ··.
In the analytical solution of Struc. '9', it-is noted
that the deflection oa is underestimated. Fir. 4.29 shows a
. .
typical connection between the tension Member ab and the com-
pression Member oa. Because of the eccentric connection be-
tween the deck and edge members, all the node points are along
the.top of the edge members ab and oa. The in-plane forces on
the ~ember a~ are transferred eccentrically to the member oa
at the noc,~ a, resulting i:i its up1..1arcl deflect ion as shown in
Fig. 4.29. In order to illustrate the effect of this eccentric
: . ·.·
. transfer of the in-plane forces 9 Strucs. '9' and '11' are
analyzed for the two boundary conditions V and VI (see Table
Ill - 2) •
·class
.. ~ . ' '
tion -of less than 20~. For very small 1J'lapnitudes of stresses
(such as less th:m, 2000 psi) the calcula.tion of the ~rror hased
'
on the measured stresses 1•.•ill ·be misleading.· T~e acquacy of
. .,. " ' .' ~
reaches as high as. 2440 psi (Struc. '11 ') which is. nearly 55%
· . ·,1
larger than that given by the J11ernb rane · theory. This appears
,,, I,
,.
inconsistent i~i th the expected behavior, since a part of the
load is also carried by the b~nding k~tion.
In order to unders'ta.ncl' t!1e ·difference. in behavior be- . I{'
· tween the single .layer and doubl'e layer d~cks, Fi).Js ... 4. 34 and
4. 35 show. the variation of the axial stresses and the ,verti-
cal shearing forces carried by the edpe mepber,. For Struc.
· '11'' (which has a 28-G single layer deck), bot1' the...~ompression
member oa and. the tension member be placed across . th.~ corruga-
tions (along the weak axis), carry hith axial loads as compared
·to the me~bers ab and oc, placed perpendicular.to the direc-
tion of the corrugations. This trend is also observed exneri-
,
mentally. Because of very 1011 in-plane stiffness across the
corrugations, the effective area of the deck resisting the in-
p 1 ane s h.. ear a 1 ong w1•t.1
t t.:e
l e d re mem,Jers
1-, oa .lb.
anL~ c is very srial l
· , , and therefore the entire shearinlS forces are resisted by the
-89-
IV.6 SUMMARY
The validity and the accuracy of the finite element
methods were assessed. Both approaches were found to converge
sati~factorily. A grid of 6 by 6 gave essentially the same
results as a grid of 8 by 8.
For hypars with fully supported edges, both the flat-
element and the curved-element methods yielded deflected shapes
that are identical with those given in the literature. Satis-
factory agreement was also found with experimental results
even when the effects of eccentric edge members were included.
However, the deflections of flat corners, such as those at the
outside corners of umbrella-type hypars, are underestimated by
I
the curved-element method. The flat element approach predicts
the experimental deflections and stresses of various types of
hypar structures with satisfactory accuracy.
CHAPTER V
V.l INTRODUCTION
. '
Based on the analysis of some selected structures
(Table IV-1) it is possible to show qualitatively the effects
of different parameters on the behavior of a ·hypar shefl.
Since. the number
'
of parameters affecting the behavior of the
shell is quite large and iheir interaction is very complex, _
attempts to show their effects on the structural behavior by
means Of forlfl.Ulae would involve extensive COJTlputational work.
During the following discussion so~e of the parameters which
were not investigated are ~entioned.
A. SHEAR RIGIDITY
According to the membrane theory, the normal 16ads
on the hypars are carried by the in-plane shearing force xy • N
of the hypar shells and therefore the effect of increasing theI I ' '
·,.
thickness is analogous to that of increasing the. value o,f a..
To substantiate this observation th.e comparison between the de-
flection and stresses for Struc. '11' (28-G single deck) and
Struc. '12' (24-G single deck) is given in Table V-1.
The variation of the in-plane shear rigidity, which
is directly proportional to the thicYness and the shear
rigidity factor a, also affects the manner in which the verti-
cal load is transferred to the supports by the membrane action.
Because of the high shear rigidities for the concrete struc-
tures ('6' and '7') and Struc. 'S' using the cellular deck with
the full effectiveness of the bottom plate, .the values of the
in-plane shearing forces show a substantial increase near the
supports (Figs. 4.12, 4.16, 4.19). The increase in the shear-
ing force indicates the participation of the deck in_ carrying
a part of the vertical load. As against thisJ Strucs. '9'-'13'
with low shear rigidity do not show any incr~.ase in the in-plane
shearing force (Nx.y) near the supporting columns . (Figs. 4. 28,
4.36). In other words, in these structures the entire vertical
load is primarily carried by the edge members.
C. HUMBER OF DECKS
As far as deflections and stresses in a hypar are con-
cerned increasing .the number of. decks has. the same effect on
the behavior of the shells as that of increasing the shear
rigidity factor a. and the thiclmess>. How~ve.r, this observation
does not app.ly for buckling (see .Ghapter VI)..
I • . .
As discussed in
Chapter II, the effectiveness of the deck in resisting the
-97-
loads depends upon the JTJanner in which two or more decks are
interconnected and connected to the supportinr edge merrhers.
However, it must be pointed out that.in order to avoid chatter
and get a better structural performance, it is desirable to
interconnect the decks all over the surface of the shell.
\'Then two decks are used, they are placed in a mut -
tially perpendicular manner and. this gives an equal bending
rigidity to the structure in both directions, thereby distri-
buting the applied loads more evenly to the supporting edge
wembers. The comparison. of the results for Struc. '11' using
a single deck·(28-G) and those for Struc. '13a' using the
double dccl~, all other constants being the same, shows that the
uniformity of the stiffness in Struc. '13a' has more even
distribution in the edge meinber_axial stresses (Table V-1).
Though the corner deflection shows practically no change, the
center deflection 5e for Struc.: '13a' is nearly half that of
Struc. '11'. The change in the bending stresses of the edge
member is very small but because of the increased membrane
action the bending stresses i~ the center of the quadrant are
reduced by nearly three times.
In practice, the use of a double deck with two decks
placed mutually perpendicular is more desirable than a single
orthotropic deck.
v. 4. BOUf'DARY CONDITIO~\rs
ties of edge members· and the manner in which they are connected
~b th~ deck are r~alistic from the practical point of view.
A. E'DGE MEMBER PROPERTIES.
The edge mei:iber properties Ab, Iy, .I.z, J and r b. for
available rolled sections show varia.tions over a wide range.
A sufficient nm111:-er of analyses could not be carried out to
formulate any definite rules by which the effect of the varia-
tion of· these· individual properties on the behavior of the
·shell can be assessed·. Except for the. concrete hypars, the
analysis was carried out for zero torsional fixity and there-
fore the influence of the torsional constants .J and r is not
clearly known:.
To get the general idea of the effect of the stiff-
ness of the edge members, one can compare the results of Struc.
'13' with very stiff edr,e members and Struc. '9' with very
flexibl~ edge members. The difference in the behavior of these
two extreme structures is obvious from the deflection profile
along the diagonal ob (Figs. 4.24 and 4.33). In the case of
Struc. '13' because of very high bending rigidity of the edge
members, the deflections along the periphery are quite small
and the deck bends freely between .the opposite supporting
edges. The simply supported plate bendin~ action is quite
dominant in ttis case. Because of the small bendinp, rigidities
of the edge members in Struc. '9' , it appears 'from the deflec-
tion profile along the diagonal ob that it is the deck that
supports the edge wembers near the free corner and therefore
. ..
the deck stiffnesses (both bending and mewhrane}'are quite
-99-
and the deck (Figs~ 4.30-4.32)• Though the results are very
-1'10-
V. 5. LOADI:~G
for th·e behavior of the shells. The higher the value of this
constant, the more dominant· is the membrane action. As dis-
-103-
VI. 1. INTRODUCTION
-104-
-105-
deformations (nearly one half the rise of .14. 4n), the struc-
ture did not show any tendency of overall buckling though the
deck buckled.
The present study was primaTily ~oncerncd WlLh d"rk
a2,,l'·J
[I'J] . .
l)
= atJ.. a6. 6- )
1 J
The coefficients of ti1e increJT1ental ·matrix [N], depend only
upon the geometrical parameters of an element, such as its
length. The incremental matrix is identical for both ortho-
tropic and isotropic cases.
For constant values of Nx' l';y and ?lxy the incremental
-108-: .
matrix for' a .sffal'low sh'ell h)tpa.r· elemf?nt and that for· flat
plate· e;I.ements. ar,e: identical. ~he OJ?.l.Y diffe.rence is the manner
in which ?the Jn-plane forces are determined (see Chapter I I I) .
'' ' :
is given by,
6-9
y = EI
-D-b 8 = A
bt 6-10
-110-
1
1 is the moment of inertia of "the eccentrically connected
the effective
.,, stiffness of an element will adjust the incre•
mental force dis.tribution. The accurnu la tion of :·these local··
? ; '
'
elemen,t instabilities will eventually lead to 'Duckling. The :
discuss.ion of this membrane and flexural behavior .for the shell
structures is given in detail in Ref~; SI, 66.
The incremental [N] matrix used for both curved and
'
flat elements is identical and can be represented by,
6-11
-1111 Tl11
[ll ]global = ---------· 6-12
11 21 11 22
pointed out in Chapter IV, the curved and flat elements, for
hypars with all edges supported give very close results for
the linear elastic analysis. It appears that the flat ele-
ments give very conserv;itive results for the bucklinr of single
decks. As pointed out in connection with Eq. 6-12 for the flat
elements, the transformation of the incremental matrix from
local to the global axes results in the modification of the
flexural as well as Jllembrane stiffnesses. Since the in-plane
membrane stiffness for a corrugated deck is very small in the
weak direction, a premature buckling could possibly be trig-
gered by the reduction in the rnerbrane stiffness. As against
this, in the curved element formulation because of the
assumptions used in the analysis, the mer ·tbrane stiffness is not
1
modified.
For a double deck or an isotropic deck, because of
high in-plane rigidity in both directions, hoth ~ethods pre-
dict the buckling loads in the same range. As pointed out in
Chapter IV, the bending action is very dominant in t!'le flat
portion of the shell. But in the buckling analysis, the curva-
ture of the hypar in the center of the quadrant may be more
critical. This fact occurs even more in the case of a single
deck and therefore the curved elenient, which accounts better
for this curvature effect, probably predicts a hi~her load.
-118-
b a
v = 21 ( ( [D Xw""X 2
·i •·
+ 2D
1
w 2
xx wyy + Dy wyy + 4Dxy wxy
2
0 0 "'
2
w2 +
+ 4 Geffh (C/AB) 2Nxy wx wy] dxdy 6-17
2 2 2 4 2
+ 2Dya b s + 4Dxya s + C'
o.2 = 4D y a 2b 2s 2 + 4D xya 2b 2
a = nrr/A, b = rr/B,
+ 4D
xy
(l+s
2
u1 + c/u 2 ] 6-19
where
u = a/b = nB/A and c = C1 b 4 = 4Geffh(BC/A) 2/rr 4
they axis (Fig. 6.12). The buckling load was not very sensi-
tive to n and therefore the prediction of the number of buckles
may not be accurate.
For structure No. 13 (double-layers, 28 Gage) the above
analysis resulted in a buckling load of 192 psf at n =2 and
at an angle of 35.S degrees. The finite element analysis gave
about 200 psf and t~e test structure did not buckle up to a
load of 145 psf when loading was discontinued.
The main advantage of this energy analysis of the buckling
load is that it is very fast as compared with the finite element
analysis. Simple supports are assumed and therefore the actual
buckling load may be somewhat higher if some bending or in-plane
fixity is present. The deflection of the edge members is not
considered, but the finite element analysis showed (Section VI.6)
that edge member deflections tlo not affect the deck buckline
load appreciably. The deck buckling load is influenced by the
area of the edge members. If the edge members are very slender
then deflections will control; on the other hand, the above
analysis should give conservative results for structures with
heavy edge members.
VI.7 STABILITY OF ISOLATED EDGE MEMBERS
The edge members of a hypar structure receive uniform axial
loads along their lengths from the deck. Some or all the edge
members are in compression. The loading remains axial during
the deformation of the edge menbers and therefore it constitutes
a non-conservative force field. To obtain an approximate value
of the buckling of the compression erl.ge membe-rs, the membrane
-124-
EI IV + S(L-x)y'' +
y
My = 0 6-20
CHAPTER VII
EXPERIMENTAL INVESTIGATION
VII.l INTRODUCTION
An extensive experimental investigation was carried out.
The purpose of the testing program was twofold: (1) to study
the behavior of light gage steel hypar roofs subjected to verti-
cal loads and (2) to provide a comparison with analytical ap-
proaches.
The following tests were carried out:
a) Sixteen flat shear tests - Properties and dimensions of
all the specimens are presented in Table VII-1 and a picture of
one is shown in Fig. 7.1. It was necessary to carry out these
tests to determine the shear rigidity G' of the decking used
for the hypar models. Twelve specimens were 6' x 6' in plan,
three were 5' x 5' in plan (all. referred to as "medium scale
tests:;), and one was l' x l' in plan ("small scale test").
b) Twelve saddle shaped hrpar specimens - All of them were
5 1 x 5' in plan with various rises (Table VII-2-VII-3). A
picture of one is shown in Fig. 7.2. The specimens are desig-
nated by three numbers: the first one indicating the rise/span
ratio, the second the number of layers of decking, and the
third whether it is an original or duplicate. For example,
for test no. 512' 1151!
indicates a 1/5 rise/span ratio, 11111
indicates one layer of decking, and "Z' indicates that it is
I
and the bottom layer was soldered to the edge members at every
point of contact. One full panel was used for each layer.
5. Loading Apparatus
The mediuru scale frames were placed in a horizontal plane
and load was applied by means of a 50 ton hydraulic jack.
Vertical support along the loaded edge was provided by rollers
on beams. Reactions were provided by a fixed wall beam to
which the frame was pinned. Steel bricks were placed on the
loaded edge to prevent out-of-plane warping.
The small scale frame was attached to a wooden frame with
metal guides to prevent the specimen from warping out-of-plane.
The entire set-up was placed in the vertical position in a
Tinius-Olsen hydraulic testing machine and load was applied by
the machine.
6. Deflection Measurement
For each test, deflections were measured by .001" dial
gages .. At first, .0001" gages were tried but it was discovered
that they are too sensitive because the readings obtained from
then were erratic. Sufficient accuracy was obtained with the
. 001 · gages. If possible the stem of the gage was placed in
contact with the specimen, otherwise a thread was used to at-
tach the stem of the gage to the specimen.
7. Determination of Shear Rigidity G'
G' was determined by the use of the expression given by
Luttre11 45 . The initial linear portion of the load-deflection
curve was used to obtain the slope. The customary testing pro-
. cedure was to initially load the specimen to a predetermined
value so as to seat all connections and then ~o conduct as many
-131-
The edge beam frames .with supports were the same for all
the tests. They consi$ted pf 3 11 O.D. x 1/4" wall thickness
cold rolled steel tubular members welded together to give the
correct rise/span ratio for each specimen. The diagonally op-
posite corners, two low and two high, along with the midpoints
of each side were rigidly supported in the vertical direction
by steel bricks (Fig. 7.2). rhis support system together with
the fact thgt the tuLes have a large bending rigidity was con-
sidered to offer continuous rigid support in the vertical
direction.
A tie bar, 2 1/2 11 x 1/ 4" in cross-section, was used to
limit the spreading of the tw~ low corners due to the vertical
load.
The decking was made up of 28_gage standard corrugated
steel sheets with 2' cover. Three sheets were used for each
layer, with the middle sheet cut to fit the 5' width. Single
layered decking was connected to the edge beams by #14 self-
tapping screws with aluminum-backed neoprene washers at 8"
spacing while the two seams were fastened together with #8 x
1/2" self-tapping screws at 2 2(3" spacing. For the double
layered decks, the top layer was fastened to the bottom layer,
around the perimeter only, with #8 x 1/2" self-tapping screws
at 8 11 spacing. The two seams of the top layer were fastened
together exactly the same as those of the bottom layer.
-133-
1. Introduction
'l "'~
For both of the models with one layer of decking., #14 self-
tapping screws at 8" spacing fastened the decking to the edge
beams and #8 x S/8 11 self-tapping screws at 2 2/3.i spacing fas-
tened the seams together. Three panels, each with 2' cover,
were used per quadrant. For the model with two layers of deck-
ing, the bottom layer was attached to the edge beams exactly as
described above. The top layer was fastened to the bottom
layer, around the perimeter only, with #8 x 1/2 11 self-tapping
screws at 8" spacing. The seams for the top layer were fas-
tened together exactly the same as those in the bottom layer
as d9scribed above.
For all three models, uniform normal air pressure was ap-
plied to the inverted umbrella shape by four rubberized canvas
bags placed between the floor and the hypar in the inverted
position. A water mano~eter was used to measure the pressure.
Vertical deflections were measured by level sightings on
meter sticks held at each location. Strains in the deckinp were
measured with SR-4 gages with ln gage length and those on the
edge beams with SR-4 gages with 1/2" gage length. The vertical
deflections, experinental axial stresses, and experimental
bending stresses at 40 psf normal pressure are presented in
Figs. 7.8-7.10.
VII.S INVERTED UMBRELLAS WITH VERY FLEXIBLE EDGE BEAMS
1. Introduction
Two models were tested with extremely flexible edge beams.
The purpose of these tests was to determine the mode of failure
as well as to check how closely the failure could be predicted
-136-
by theory.
2. Small Scale Model
One of the inverted umbrellas with very flexible edge
beams (Fig. 7.4) had overall plan dimensions of 2' x 2', center-
line to centerline of the perimeter edge beams, with a 3" rise
(1/4 rise/span ratio).
The edge beam frame was made up of 3/16'" O.D. x . 014 11
wall
thickness brass tubes brazed together. The decking consisted
of two layers of 2 mil corrugated steel sheets soldered together
at each point of contact around the perimeter. The decking was
soldered to the brass tubes at each point of contact.
Uniform vertical loading was simulated by 64 discrete
weights (16 per quadrant) hung from strings passing through very
small holes in the decking and attached to 1 11
square pads.
Vertical deflections were measured with .001" dial gages.
The dial gages were positioned such that the stems did not
touch the shell. To take a deflection reading, the stem of
the gage was pressed until contact with the deck was made;
avoiding the application of any force by the stem as much as
possible.
Experimental deflections at 40 psf load are presented in
Fig. 7.11.
Collapse of this model occurred at 73 psf. It was ini-
tiated by splitting of the decking along a row of holes through
which the weights were suspended. This splitting caused the
brass tubes to fail in bending very close to the central sup-
porting column.
-137-
dimensions of 5' x 5', G' is less than that for 6' x 6' test
no. 5 but slightly more than that for 6' x 6 1 test no. 6. Thus,
it appears that scatter in the test results masks any effect of
the small variation in size.
The spacing of seam connections and number of seams have no
discernable effect on G' based on the results presented in Table
VII-1. Also, it appears, at least for a limited range of thick-
nesses of panels, that the shear rigidity increases linearly with
the thickness. For example, compare the results for tests no.
5 and 7.
Tests numbered 10 - 16 had two layers of decking running
transverse to each other. In each case the top layer was con-
nected to the bottom layer along the edges, rather than being
.· .
connected directly to-the edge members. The effe~t of this ar-
rangement of connections is to make the top layer less effec-
tive in shear than the bottom layer, which is attached directly
to the edge members. The reason for this is that any shear
deformation which occurs in the bottom layer along the edge
members occurs also in connections for the top layer which in
turn reduces the effective shear stiffness of the top layer
below that obtained by connecting the top layer directly to the
edge beams. From Table VII-1, tests no. 10 and 11 versus 1 and
2; 12 and 13 versus 4, 5, and 6; and 15 versus 8 and 9 show that
the shear rigidity for the double layer tests is less than twice
that for the corresponding single layer tests. It appears that
connecting the second layer in this manner results in only a 33%
increase in G' over that of the singie layer, although as men-
tioned previously the relative positions of the screws appear to
-139-
deck with that due to a 200 lb. load on a double layered deck
reveals that a double layered deck is more than twice as rigid
as a single layered deck for carrying a concentrated load; the
reason being that a concentrated load on a single layer is car-
ried mainly by a few beam strips whereas on the double layer it
is spread out and thus carried by more beam strips in each layer.
For all the models tested, the maximum deflection due to a 100
lb. load on a double layered deck is approximately one-third
that for a single layered deck.
b) Inverted Umbrellas With Very Stiff Edge Beams
Figs. 7.lO(a)~ 7.ll(a), and 7.12(a) show the measured verti-
cal deflection at 40 psf normal pressure for three inverted um-
brellas tested at Cornell. It is seen that difficulty in obtain-
ing ~ymmetry was experienced. The lack of symmetry was probably
due to unequal air pressures in the rubberized canvas bags. The
bags were not entirely air tight so that unequal rates of leak-
age from the bags could have caused relatively large percentage
variations in the pressures.
c) Inverted Umbrellas with Very Flexible Edge Beams
Figs. 7.13 and 7.14(a) show the measured deflections at
40 psf for the 2 ft by 2 ft small-scale test and test No. 9,
respectively. Fig. 7.13 reveals that rotation about one diagonal
apparently occurred during testing of the small scale model. Fig.
7.14(a) shows that a reasonably good condition of symmetry was
obtained for the medium scale model. For both models, it is
apparent that the deflection at the free corners is much larger
than that at the center of each quadrant.
-142-
3. Hypar Stresses
a) Saddle Shape Supported all Around
The experimental stresses presented in Table VII-2 reveal
the difference in structural action between a very flat hypar·
(rise/ span = 1. 8) and one with a much larger curvature. (rise/
span= 1/3). ·The bending stresses in the models with a 1/8 rise/
span ratio are much greater than those in the models with a 1/3
rise/span ratio. The decrease in bending with incr~asing rise/
span rati.:o .is accompanied by an increase in membrane action as
seen in·the last column of Table VII-2. The experimental force
in the tie bar for the models with a 1/8 rise/span ratio is much
less than ·that given by membrane theory whereas the force in the
tie bar for the models with a 1/3 rise/span ratio is almost as
large as that given by the raembrane theory. Thus, it. is demon-
strated experimentally that for very flat hypars, shear stresses
are much smaller than predicted by the membrane theory and that
bending constitutes the major part of the structural action
whereas for deep hypars (rise/span ~ 1/3) membrane stresses pre-
dominate with values close to those given by membrane theory and
bending stresses are insignificant. The axial stresses in the
deck, for all models, are seen to(be insignificant.
Table VII-3 shows that a concentrated load produces signif-
ic·ant bending stresses, especially in the flat hypars, but very
little membrane action.
b) Inverted Umbrellas '1ith Very Stiff Edge Beams
Figs. 7.lO(b), 7.ll(b), and 7.12(b) show the axial and bend-
ing stresses obtained from strain measurements at 40 psf load
-143-
CHAPTER VI I I
BIBLIOCPAPh'Y
. 14. Croll, James G. A., and Scrivener, J.C., ' 1 Pdge Effects
in Hyperbolic Paraboloid Shells, 11 Journal of the Struc-
tural Division, ASCE; ~arch 1969.
17. Connor, J. , and Breb bia, Carlos A. , 1 'St if fnes s ~·~a tr ix for
Shallm·1 flectangular Shell Element l>" Journal of Engineer-
ing: ~ echanics Di vis ion 1 ASC[ / 0ctober 1967.
7
1968.
r .i.:?_;_c~~ :~r
nroperties of Orthogonally Stiffened Plates' I'
Jcur.11d.l 0£ Applied Pechanics, I"arch 1956.
49. Wright,. R. 1'1., et. al. 1 ''BEF ./\nalogy for Analysis of Box
Girders," Journal of Structural !:livision, ASCE 1 July 1968.
DECK PFOPEPTIFS
P. R. s z2t
l.l = l ds A-2
X I:t! 0
0
-Y-
s z2t
I T ds = Iy' A-3
0
P9..
= A-4
Ext
Equating equations A-2, A-3 and A-4, we get
IO
E
x = r' I:.: 1 A-5
y
The other membrane constants are given in Table A-I.
A-Ib Bending Rigidities
Bendinp rigidities for the x and y-directions are
given in Table A-I. The method of determining the constants is
given in f. 35. For the orthotropic plate, the twi~ting con-
stants Dxy and Dyx are not equal and hence the average values
o f tlese
1 . t.e
constants are use d 1n . 35 .
h ana 1.ys1s ~··
r inor d ev1a-
.
From Eq. A-5~ [xh' the elastic extensional constant, for the
hat only,
-160-
Elot2
Exh = Iy-,- A-7
1' • is the worncnt of inertia of the hat portion about the base
y
plate. By compatibility,
E t
p
2
= p1 ( 'x'h) c-1)
E t
A-8
1
The equivalent orthotropic plate is assumed to be of the thick-
ness t 1 •
E
x
= p,-9
( a + ~\
L t NJ
R,(l+( ~h)(~))
E t
1
For all practical purposes, the deno~inator of the above equa-
tion (A-9) is equal to unity.
A-10
EY and vyx - If the full cross-section in Fir. A
3
is effective
in resistinr the axial force along the y-axis then,
A-11
A
vyx = v2
to. A-13
''y
The shear rigidity for the closed deck can be considered equal
to that of the base plate which is usually cirectly connected
to the edge me~bers. Even a thin plate interconnecting the
free hat portions will increase the effective shear rigidity of
-161-
a . r1 A-14
~
where Ac is equal to the area enclosed by the cellular hat
portion.
However, the magnitude of D x tacitly assumes that
. y
the shape of the deck does not deform (dotted lines in Fig. A4) 47 •
The in~plane shear forces set up due to the twisting, produce
secondary bending moments in the individual plates. Besides,
the bending of the overhanging plate, outside the cellular
portion, further adds flexibility to the cellular section.
In order to account for this reduction in torsional
stiffness, an effective torsional rigidity of the cellular deck
is defined as the torsional rigidity of the ideal system (shown
by the dotted lines in Fig. A4 ), free from individual plate
flexural deformations. The work of the deformations due to the
torsion of an idealized system is equated with the work due
to torsion and secondary flexure. The modified twisting
rigidity can be given as,
-162-
4A 2
.D
yx
= { (15+2h f) c + A-15
tz
-x
'
(a)
p - Fig. Az
(b)
p z
Assumed llinges
~Midway Between
/ Ribs
\
Rotation in
an 1dealized
System z
/\ctrrnl
l<otat ion
TABLE A-I
PROPERTY: DECKS
CPEN CELLULAR ISOTPOPIC
Membrane
Io
Ex E E p
.._.
Iy '
A
s _r_
F.
y F r F
P..x
E
\) \) \) \)
xy
I Q, A
x
0
\)
yx Vy-1 5- \)
Ay
\)
y
1
El vE I0 ' vE vF.
y
!~-
1- \) yx \) xy " 1. 0 1-v 2 x-xy 1-v 2
E
xy
aG aG ~
l.J
Bending
~ 3
3
iE t 3 .t. tl Et
D
x l2S 12(1-v 2 ) 12c1-v 2 )
n
·'~·
y f:Iyd EI vrl
I_. Ft
3
2
V[t 3 £ 12(1-v )
'I
VE t 3
()
'-·1 12S
12(1-v 2 )
3
Et
'- -
xy GI 0 0
24(1+v)
GI
0
s 4Ac 2G Et
3
D -9.- ds 24(1+v)
yx
Rrcd Ut) £
-166-
APPENDIX B
TABLE B-I
<P 4 - '63a a
6b
b a
<P 5 6a 6b
b a
4>6 6a __
31)
_..,
a ;.,
4>7 Th 3i
<Pg
a
60 - b3a
a b
<Pg 61) 6a
<Pio a b
3fi . 6a
a
e.g. 4>1 = b3a Ext + - E
3b xyt
-167-
TABLE B-II
Coeff.
676 ab
2
11 b 2 11 a 1 11
-7, -c. 7a- 35 b:z- 25 so
156 b 54 a 144 1 72 1
-35 ~a 35 ;3
t)
2S an -ab 234 ab
27 a 12 1 36 1
35 b2 25 a 25 a
12 1 6 1
25b - 2.Sb"
lrb 2 1 3
352
a 25 25
540 54 a 144 1 72 1
25 81 ab
35 a3 35 b3 ab 25 ab
13 b 2 27 a 12 1 6 1 211.. ab2
35 a3. 35 ~2 25 a 25 a 6
---·--
27 b 12 1 6 1
35'""7
a
2511 250
Continued
-168-
Con'tinued
Coeff. D D D
x y xy
1 1 169 2b2
25 so ~a
13 b 2
78 ~ 12 1 6 1
- _.3_s_a_3_ _ _~_ 1
s b 2~·_ _,_2_s____
6
_ _1_4____ a_ 25 a
27 b 22 a 2 12 1 36 1
- 35 ~2 35 h3 25 -b IT b
13 b 2 11 a 2 1 3 143 2b2
- 70 a2 35 :lJ l 25 25 18 a
~---- ~-------------~--~---~
4 b 52 a 16 b 8 b
35~ 35b zsa Ea
2
11 h " 1 61
- 35 z,
a.
, 25 so
2 b ''
3 22 a 2 4
'35 2 105 b 75 b
a
4 b 18 a 16 b 8 b
- 35 ~ 35 b 25 a 25 a
--.-2
2 b 13 a 4 _z.
b
35 ~ - 155 b 75 b 75
3 b3 9 a _4_b____2 b Q. ab3
35 a 3 35 b '-2 5 a 25 a 2
3 b3 ---13-T -:-· b-..- - - - b_ _ _ _ _
1_3_2_b_3_
- 70 2 - 210 r,-·. - 7s - 150 IT a
a
?
11 a- b b 11 2 3
105 ~ 75 TI 6 a b
52 · ·b· 4 a3 16 a 8 a 52
6 26 - a 3b
35,' a 35 b3 25 b IT b 3
2
22 b 2 a 3· 4 4 2 2 a 3b 2
~27.___-_1_0_~.~a___-_3_s_.~~2~----7-;_,_a~----2~s_a_______--~9:__~--
6
Continued
-169-
Continued
Coeff. Dy
3
26 b 3 a 4 a 2 a
35
~~-~~-~~-;;.--~-
a 3s b3 25 b 25 b
11 b 2 3 a3 a a
--10-5 -a 70 b2 75 25
9 b 4 a 2 a 9 3
ab
35 a 25 b 25 b 2
a a 13 31' 2
75 150 IT a -
16 a 8 a
25 b 25 b
4 2
75 Cl 75 a
16 ab 8 b
225 225 a
4 ab 2
225 225 ab
1 1
225 ab 450
ab
4 2
225 ab 225 ab
C - F ( C )
2 - -'xyt AAxB13
TABLE B-III
7 3 7 __!_ b 2
el 20 a es 20 a eg 20 b el3 30
-
1 1 2
e2 ITO7 ab a6 40 ab 8 10 _!_ b2
20 el4 lflO ab
1 2 1 2 6 11 1 2
03 -IO a 87 30 a 120 ab
1 . 2 1 2 3
e4 120 a .b es -180 a b el2 20 b
!91'.t e 1;6x I
u.,
SYMMETRIC
....
t
..........
I
TABLE B-IV
CURVED ELEMENT STIFFNESS MATRIX [K]
ti -t_l- -GL_~ ...~~-e?>.. G.,, ~4
4' Yo;
APPEIJDIX C
DECK STP.E
can be deduced frorn the expressions given beloN. The two major
diffe~ences are: (a) c=o; (b) The local nodal displace~ents
C-1
C-2
71
y
= - [D 11v, xx + D N.
y 'YY
] C-7
J I 2
---'1-x
I
b
INx ,Ny
2 ,Nxy
-·----· -10.IM_____
x ' My
-
h
2 I
I
- p
APPENDIX D
TABLE D-I
Coeff. Nx Ny Coeff. Nx Ny
78 b ·7s~a 27 b 78 a
<I> 1 1 75 a
1 75 b
- - - - - - - - = -2 - - - -
<I> 13 1 75 a - 175 b
11 b 13 13 b 2 13
<1>2 175 a
350 a
_ _ _ _ _ _ _1_3_ _ _ _ 11 a 2
<1>14 - 350 a
9 b
350
11
a
a2
<I>) - 350 b - 17 5 b- <1>15 - 700 . 17 5 b
------- -------+--
11 b2 11 2 13 2 11 a2
2100 2100 a <1>16 - 4200 b 2100
- - - - - - - - - - - - - - - - - - - -3= - - - - - - -
78 b 27 a ~ 2 h 26 ab
------- 115 a 5 ns
'1'11 115 s25
~-----,t------------------
a
11 b 2 9 11 2 11 a2
<1>6 - m· a 700 a <l>1s - 211)0 b - 21QO
·~------=------------------
13 13 a 2 1 3 11 2
<1>7 - 350 b 350 o ¢19 1050 b 1575 a b
11 2 13 2 2 b3 3
<l>s 2100 b - 4200 a <l>zo - I7S a 175 ab
---------~2·=1~b~----~z-=1-a·~--~ 1 _3____1_3_a_2_b_
<l>g - 175 a - 175 b ¢21 1oso b - 3150
------ --+--------·---- ------
2
13 b 9 3 b3 3
<1>10 350- a 700 a <1>22 350 a - 700 ab
--------------
9 13 a2 1 3 13
<1>11 - 700 b - 350 5-- <1>23 - 1400 b 12600
------------~--~---t-----------=--------
13 2 13 2 3 ~' 13
cf>12 - 4200 b - 4200 a <1>24 - 350 tl·_:_ - 10 so· ab
Continued
-176-
Continued
Coeff. N
x
r~y Coeff. Nx N
y
1:-3 il-:-2 . 2 a
~) ... :
(,,,.,,.,,,. 1400 b 6300 a b <P ":
J.
115 ab 1751)
26 2 a3 13 2 1 3
4>26 525 ab I75o <f>33 3150 ah - lOSIJ a
11 2 \ 2 b3 2 3
<P21 - 1S7S ab - 1 a3
<P34 1575 a 1575 a b
13 3 a 1 . 3 1 3
4>zs lOSO ab 350 b ¢35 3150 ab 1050 a b
11 2 1 3 1 3 1 3
<Pzg 6300 ab 140') a <1>36 4200 ab 4200 a b
3 3 a3 1 3 1 3
<P30 700 ab 3SO o 4>37 - 1050 ab - 3150 a b
13 2 I 3
<P31 12600 ab 1400 a
TABLE D-11
e2
ab a 2b b2
so es 300 es 60
b 1-,2. a2b2
au
03 IO e6 300 e9 1800
N ab·
e.g. N(l,4) -.· xy
so
SYI1'2.TPIC
TABLE D-I I I
[~.!] VATRIX
1
-177-
·o · o o . e2 · .o e5 eY3
o e2 o ~~
5 o exy 3
SYt1lliTP.I C
0 . ·~
xy4
TABLE D-IV
(N] f"ATlUX
2
-178-.
-179-
APPENDIX E
INTRODUCTION
In the case of light ga?e hypars the orthotropic deck
is connected at discrete points to the supportinp edge merrber.
Because of this discrete connection, there is a certain amount
of non-compatibility between the edge members and the deck. The
stiffness matrix used in Chapter III was developed by the use
of direct co-ordinate transforMation. The co-ordinate trans-
forJilation assumes that the neutral plane of bendinr for the
original beam cross-section re~ains unaltered. ~ith the type
of connections used for the light-gage hypars, the assumption
made above represents the true behavior.
However, in the case of stiffeners which are rigidly
connected to a plate or a shell or a monolithically cast con-
crete beam, there is full corrpatibility between the strains at
the junction of th.e deck and th~ hearn. In this cornpati ble cas c.
a part of the deck also acts along with the stiffener (effective
width concept). The interaction of the deck and the stiffener
results in the adjustment of the neutral axis of the section.
To account for this change of neutral cixis, the bendinp nroµerty
of the stiffener will have to be rrodified hy arbitrarily assum-
ing the effective width of the deck actin~ along with the be~m
STIFfNESS rrAT!UX
A typical eccentric member is shown in Fig. 3.10.
It is assumed that the mer.iber is uniform in size and its local
axes i, ~and i, througt the shear center and the centroid, are
parallel to the global axes x, y and z. Let Uc be the average
.
axial deformation of the section measured at the centroid of
the section (C. G.) and Vs and 1
\:
s be the bending deformations
measured at the shear center (S.C.).
Assuming a rigid connection between the stiffener and
the reference structural node points, the displacements at the
shear center and the centroid can be expressed as follows:
uc . = u - z
c
(aw)
ax - Y.c (av)
ax E-la
= v - ZS e E-lb
"s
ws = w + y SS E-lc
where e= !;. The total strain energy of the beam ele:rnent can
be given as
a au c 2 a a2vr 2
ub =
EAb
-z- l (--) dx + E~y l ( . s) dx
0
ax 0 ~
Elz a a2vs . 2 1 . a a
-z- 0~dx· + !.2 Er ·a·' 2dx E-2
cf C-z) 2·
+ cl.x + G.J ! I
ax 0 0
-2Z c Cav)e'
ax + e 12 p 2 }dx E-5
Bleich 56 .
2
cr _Q_ a J] E-7
c ·rr 2E I pc
provement. ·
-184-
z
tJy = -ze
tJZ = Ye
z
y
0 0
\
a = 200"
,,
\\
\\
\\........ /\,.
/
/
\
\_ ST 12 I 39-95 \_ 12 WF 40
'/(I
\} ' 1,.., t e It I 0 (f I 8 ><.~I
SYMMETRIC
_____.,. ..
(; 0 0 0
. ~ --
12.f [ 0 l 12.EAbYcl -12 £ l z ls i'EAi,Ycl~ 'EI• "l. 5
A.~ o.3 0. 3 a...2. o..•
t
···r- - --
12El0y 12.E.I'ijYs !'Elcy:-c.EliYs
- - - - - .---1 -
o.~ a~ tA.i-: o.2..
12-E l;r Y.s/~~ ·- . i-(,, El;/f~7i
TABLE E-I
r Eli/:1a_~'
11
re' EL ii y,-'i~" fli-c;I I&-'
(
STIFFNESS MATRIX FOR A COMPATIBLE
+~c,:r/o...; o.~~-GJ/to
ECCF.NTRIC STIFFENER
+ 12£rt. /a.:!i: _, E
1
rt/a.'.1.
Li E I 0 y !- 4El'lf Y6
I
a_ CL
4 E. I<J Ys -; a. 1
+'1 El 1l~/ o..
+fg GJ~
+4EI'i::,/a_
-186-
6 6 a a 6 6 a a
5 5 Gl IO 10 Gl s S Gl 11) IO GI wl
a a a a
6
Gz - IO Gl IO Gz - S6 Gl - S6 G2 - IO Gl -
10C 2 9 xl
2 2 2
- rs 2r a a 1 2 1 2
rs a a ~)1
ro IO Gl - 30 a ~a G1 ey1
2 2 a a
- 10 1 2G
rs· a Gz - 10 Gl Gz 30 Gl - 30a 2 9xyl
6 6 a
SYMMETP.IC 5 S Gl 10 - a10 Gl \•T
'2
Gl = yc 6 a a
I S Gz IO Gl IO Gz 9x2
G2 =~
A 2 2 2 a2G
p IS a l~ 1 ey2
Note Common Factor a ---
r:ATRIX FOR 2 2c::
TABLE E-11
-
lNCP,m~mITAL rsa 12 e xy2
cm~PATIBLE ECCEHTJUC ST! FFET«TE P
--· ----
-187-
APPETDI X F
COJ'-,;Pl"TEr· PROGRJ'.}'S
(:!ethod 'a' and ~'.ethod 'b'). Since the input data for both
curved and flat eleBents is nearly the same, a general des-
cription of both stiffness and instability programs is given
here. All informatj.on given in this appendix, is pertinent to
Im~ 360/65 model availahle at the Cornell Universit~ nouble
precision is used in all programs.
5. Loading
The solution can be obtained for three different
loading cases or any combination of the three. The three
cases of loading are: a. uniformly distributed load over the
whole surface; b. weight of the edge rnewber; c. discretely
applied forces at any nodal points.
OUTPUT
INSTABILITY ANALYSIS
TABLE 11-1
SINGLE DECK
TABLE 11-2
DOUBLE DECK
2 1/2" x
1/4"
-192-
TABLE III-1
A 0 0 0
_
\/ 1-
---·--
~ 0 0 0
AbE:G, , _AbEl'c. ..... A.bE
0 0 0
_ AbE!c: Ab£1c,.
·~
0.. a.. 0.... i a... a.,
.. - .- - --.--~--- -
-l2E I'll. ~S ' £Ii. - t.lfl:1: )12.E.I&~S G.E.li!
0 0
o~
D 0 :z...
Q,1; o..'J. a..o a..?> 0...
·- - ~.
··-•<, ---· C---·-
12. EI ;if 12..EI~ l& I -ft( I':I .. ,,.EI3 -12EI.a "fs - 'E l/f
'l.
0 0 0 0
o..~ o..~ Q,. a.,3 o.. a ~
.. .,. . - 2- - '-·-----·--····-- .. -· -···-- .. ·---·- ----- -t'l.EI,.?:YJ
1u11 -io/'J
,.. -iaEI;r1;-'Elt~ t'lEl~~ -11.E..t11 -0 £Ii:,l.S
:z. """' El;f't'!i
it I 'l. El '1~ 0
11 o.?,
t-12-El ll )Id '}...
2. d- 0. 0...2. (),,
- ~
0.
+6¥a.
... ~ ....1: • >'
~----~ ----- t-····· -·----- ~~°".- ...
~AbE. .:Zc.
"'"E 'l:c.
c1.
'-AJ-"l'c.lc -Ai,l'l-c ro£I4 l'£I~Ys a. IA bE ""( -i, 1
I
0
I+ 4E l;i 0.... <3. o..l- i 0.1... ~.it:rff a.. I
... 0.
- ·1
IAbE l'c:;.
·-- -. --· -----·-r·-·----·---
I
a.,.
-A"ET:rj
I:> c.
I
i
ANGLJ: OF TWIST e) .:OE IJ1'& ~ tizlg.
fl-12Eii"l'0 I
2- I
0.,1..
a. I
+ GJ"/ex \ I
I
AbE t.Z- I
I
a.. '-Ai:,E'fc.'l,i
ti- 4eI~ j a..
I
I
().., i
...,. I
··r:< I
+ 4~1~ I
-194·
'U , v, w, 01<.., Q211 e~ tf) >4-(J I 'U. l- 'V,_ wi. ex :.z... e<f z. e.,, ,. e-..~,.
AbE A&,E le -AbEY'-
'
---
a.
0 0 0
-
0..
...
d.
---·---·~ .
0
··-
- "be:
Oi.
0 0 0 '-AbE l:, AbE 'IC,
~ 0.
0
!
~-f.[1 UEI11s 1-12. S IJt -t:z. £!1 Ys -'-E.Il
0 0 0 0 0 0 I
&.l ~lo a.• ~3 ~J.
s " -. I-··. ,, .. ~
~---~---
.._~
Yo/'.J.'o.'iI .:.·11u 1 ~~
IZ£La
4H1;/o.
0
M:'-la
!.
16
GJC
~~
-195-
TABLE II!-4
ROTATIONAL TP..AJJSFOP 1"ATION t~ATPIX
[A] =
CZ
-cyczcosS+CxsinS
p,]1= J x
C2 + CZ
z
Cy Cz sinS+C x cosS
-jc!+c~sinS '
j c~ + c;
0
r::z:-::2cos 0. ., 0
J1...x,.1.;z
[:l.]2=
0 0
cx L
y - y
BOUNDARY LOA.;
'RUCTURE GEOI 'ETRY DECK EDGE ~.~EMBEPS
CONDITION~ P!1
-
I • TYPE
A
Ir~CI!.
I
INCH
--
c
-- INCH
~NO. OF~TH~CKNESS
TYP DECKS ·-INCH
~-- . ---
I d
INCE4 /INCH (Y,
1:10BS/INCH 2
6
\) Bl B2 TYPE PSF
x
I xso xso xlO I 1 0.80 - - 1.0 28.! 0.40 - - I ioJ
--- --·-
'" 6.46 6.46 1. 304 1, I
0. 2.5 - " r..5 0.39 - - II 144
----- ·----·--··
~ 30 30 var- II 1 28-G 0.00047 var- 29.5 0. 30 3 1'0. D. 3"0.D. III 40
ies 0.0149 ies !:;' ~T~1 k ~II Thk
pipe pipe
-------·-
. L ii !) II
2 l< n Tl n ll
"
,. I! n
---···"' ,,
• IV 360 600 100 . IV 1 See 1. 532 1. 0 t:
14 WF 14 NF " 80
69 Fig. 43 68
4.3 .,
- -
• III 24 24 6 I 1 0.25 - f
2.77 0.15 +See Fig.
4.5 :+
IV 40
9
--- ·----~---
fi
r '' ' I
'· I fl
" 1:
' " f'.'" I 11
"
·-·--
t
•• r. I 1: ,.. 7• fl It i; ti ~ Ii ti *28
' VI II E
---- - ------ -·
BOUNDARY LOA
STRUCTURE CEOMETRY DECK EDGE FEr,'BERS CONDITION~ HJ'
-
~o. TYPE AM
Ii~Qi I~CH
c NO. OF,THICKNESS
INCH TYPE DECKS INCH
I .
yd
INCH,. I INCH
;
a
, Exi0 6
LBS/INCH 2 v Bl B2 TYPE PSF
11 II 72 72 14. II
·.
1 28G 0.00047 0.06" 29.5 0.3 3 ''C. D. 2-31!0.D. v 40
.. 40 0.0149 ~tlThk Jii"Thk VI
pipe pipe
-- ... r;
..
.
12 . ' 1; '
·•.
11 11 II
24G 0.000753 1; II II
"
ll
VI II
,, 0.0239 .
..
13 - II
.
""
1i
..
,. II
II 2 28G 0.00047 0.05
. ,,
" I!
" II II
- ..
•· !! I! ,, "
13. II ~-
0.060 ,. [40
15 ;.
12. 12. 5.216 I 1 0.25 - 1.0 a.so 0.39 +NO BEAHS+ IV 14 ..
92 92
x Dimensions are
·-·
in ems
-
+ gm/ cm 2 ..
only
..
1: alf. the structure was loaded
-
• Two lc·ading conditions are analyzed:
1 With 40 psf only 2 40 psf + weight of the edge JTlernbers.
For types of structures see Figs. 4.1 - 4.5.
-197-
-198-
TABLE ·1v-2
DECKS
= Isotropic deck of uniform thickness
I
II = Corrugated sine-form. Fig. 2-2a.
III = Corrugated trapezoidal. Fig. 2-2b.
IV = Cellular trapezoidal deck. Fig. 2-2c.
BOUNDARY CONDITIONS
= Edges
I x = + A and = + B are fixed.
y
II = Edges x = + A and y = + B Knife-edge supported.
III = TF = O for edge members along x = +A and y --+ B.
(i) w =o along x =+ A and y =+ B
TABLE IV-3
. ' ., . '
cm.1PARISON OF DElJLE·CTIONS BY
J'IETHODS 'a 1 and 1 b 1 ; ; FOR 5-TPUCTURES. 1 and 2
STRUCTtJRE 1 STRUCTURE 2
LOCATION Method 'a' Method 'b' · r.'~ethod 'a 1 Method 'b'
-2
10 ;.. 2 cms 10 ems .10 -3..H~ch · -3.
.10 · ~nch~
. .
s. s·o ·
, ,.
6
0.742 0.750 3.271 3.251
2.218 2.248 7.526 7.561
2.613 2.647 9.081 9.184
0 2.531 2.551 9.196 9. 322
TABLE IV-4
9 28-G
Double Deck 832 848 1460 1870 2292 2718
11 28-G
Single Deck 890 62 6820 10700 7710 10762
12 . . 24-G
Single Deck 69 5600 6200 6269
13 28-G
Double Deck 2780 22 6505 4510 9385 4532
', J• • • , ,· EXPERir'ENTAL
~· ., . '
STPUCTURE 12
STRUCTURE 13
0
a
0. 18 .. o.is 0.26 0.040
ob 0.31 0.31 0.55 0.15
oc 0.18 0.15 0.26 0.040
oc 0.56 0.54 0.92 0.26
-
d-i----·-'-e
1--~-----
c ·• ~-------1
TABLE V-1
Rise 'C' 14.40" 0.38 1.57 0.38 0.69 -5329 -5329 2747 2747
9 • 10 13.so" o.41 1.69 o.41 o.75 -5518 -ss18 21s2 2752
Shear 0.05 0.18 0.31 0.18 0.56 - 856 - 856 4~; 43~
13 • 13a Rigidity -11:110·- -o,,,...... rs----0-.-2. .s...--o-.
. . . 1. ....s--o-.,. .s---
. --.s. .s,.......1--.
. . a-s
. .1----.4
.. ....s-1--..4. .s. . .1-
..
£ actor a
11 • 12
Thick- 0.0149 0.13 0.29 0.16 0.91 -1129 - 802 456 1070
ness
0.0239 0.116 0.26 0.13 0.57 -1150 - 870 420 1093
No.
Single
11 a 13a of Deck 0.13 0.29 0.16 0.91 -1129 - 802 456 1070
Double
Deck Deck 0.15 0.28 0.15 0.48 - 881 - 881 481 481
EDGE MEMBER STRESSES PSI DECK
MAX. BENDING BENDING STRESS Shear force at
*Structure --~--------------~----~------------------------~---------
PS I Center for Nxy
oa oc ab be lbs/inch
18,357 18,357 17,144 17,144 1,870 52.75
9 8 10 19,077 19,077 18,179 18,179 2,130 54.96
•
6,648 6,648 4,228 ~,228 4,510 50.0
13 6 13a
5,225 4,634 3,990 4,480 10,700 36.96
__1_2___________4_._7_2_3~___3_._9_7_4____3......_8_01________4_,_3_8_5____~--6~2_0_0___________3_7_.7_5___~
__1_1__&
5,225 4,634 3,990 4,480 10,700 36.96
11 & 13a 5,899 5,899 1.095 4,095 3,780 Sl.28
For locations of deflections and stresses see Fig. 4-2.
* For the type of structure according to number see Table IV-1.
204
TABLE V-2
THE COPPJ\RISON OF AXIAL STRESSES IN THE EDGE r·m1BER AS GIVEN
BY f''E!'BPPNE THEORY AND BY ANALYSIS
~~EMBRA'l\IE
ANALYTICAL PSI % OF MEMBRANE THEORY
THEfi'P.Y
STRUCTURE STRESS
NO. a PSI oa oc ab be oa oc ab be
SINGLE DECK
.
5 1. 0 *19660 10700 9070 ~440 11040 54.4 48.7 50.6 56.1
----·.. ~·--
,_ ____
18640
---- ·-
11 0.06 1570 1129 802 456 1070 71. 8 51. 5 29. 68.2
---·- ------ ...
----------~ ~ -- ..
12 0.06 1570 1150 870 420 109 3 73.2 55.5 26.7 69.6
-- ·-
DOPBLE DECKS
- ·-
9 0.04 7100 5329 5329 2747 2747 75.Z 75.Z 38.8 38.8
13 0.05
---54.4 27.7
1570 856 856 435 435 54.40 27.7
~------ ---~----
13a 0.06 1570 881 881 481 481 56.2 56.2 31. 30 31. 3
-·
14 0. 06' 3420 1673 1673 1575 1575 49.0 49.0 46.0 46.0
--~ ....
-205-
-206-
TABLE VI-1
STIFFENED PLATE :BUCKLING PROBLEMS
DitvtENSIONS
LOADING a-inch b- inch t-inch y ·6 K % 'Error
-::
::
J
..
....
-
...
.:s /? -
'
2
108 l08
... ..
5/8 5· 0.10 :Li. :to
·-·'
....-
... 10-8" 108 5/8 - - ,.
. 4. 0
b <0.2
1 2 108 108 5/8 5 0.10 20 .·5
a/2
..
..
j 108 54 5/8 - - 4.0
~
b
2
~ 108 54 5/8 5 0.10 7.69
..
a/2
'---a~
... a
1
- 1 24-G
70.5 ·. 70. 5 - - 0'•·670 . 7.46
~b
l --- - -
0.0239
TABLE VII·l Fropert1es ~~d ~i~enslo~s ~: flat Shear Tests
Teat l>h»nstons Steel Edge No. of Seam Deel< ... id~e G'
-
No.
l,
(ft)
6x6
Dockins
26G s.c.•
He:::bors
6''xl~"x.1046"
Pa.~els
l
Connections Cor:.~cctio:is
.163
(l'.::.1,,l '.I 'J
33100
1 layor channels valloy and @ 4"
3;... 6 x 6 26G s.c. 6"x3/4"x.1046 11 3 screws @ 811 screws @ every .056 11400 '
N
1 layer channels 3rd valley and @8 11 •• ....
c:>
•
4 6x6 2SG s.c. 6 1'xl\ 11x.1046" 4 screws@ 8" screws @ every .011 12000
1 layer channels 3rd valley and @8°
16 1 x l
2 layers. brass tubes
·1;-
"1''4,
-209·
Test ln
Wlil.lX Stress at Center ln Strom~ Dir. Axial Force
!\o. inches Bending (psl) Axial (psi) in Tie J3ar
___ ~· Lc'lyer Top Lay;er Bot. !.ayer Top Layer _.;..(_lb_s.>_ _
321 0.86 12400 10000 -290 1810 1560
822 o.6S 13000 13000 ·820 740 1310
821I o.65 10300 7600 ·680 1750 1250
w,. 1 ;i~~
in St_LcSS_!tt_ Center in Stron& Dir. Axial Force in
i nc:lCs !knclin.~ (psi) Axial (psi) Tie Bar (1 bs)
:a1c 0.(,J ----·--
17000 4300 150
fll2C 0.67 17400 2300 130
0
-212-
'A
'/
/
)
·;
(I I
, /
1 1/ ,_..._
'rj Vl
y '-' s::
0
·~
,/' +.)
m
'{ I-<
~
.-i' O{J
1 I .,.;
/1 ! ··I 4-4
s::
/,.I u
0
/'. I-<
re
p..
I >-
::c
µ
s::
11'
$-<
11'
'+-!
'+-!
·~
c.
N
i
rl
00
•rl
µ...
-213-
0 x 0 x E D
x' x
s
.& ..
-
s
) El,.Exy
Dl,Dxy
.... ;-
y y ,Ey, Dy
a) Sinusoidal
Corrugated
b) Trapezoidal
I- l ·-~
f Q j \ j \ j \f c) Cellular
~--~--·-~
d) Stiffened
Panel
I I
[ ]
I
'
e) Box-type
I 11 11 I
l---!--.,~
I
'
Effective Width
Varies
z
Fig. 2-3 Effective Cross-Sectional Area
of a Hat for Axial Force
e/2 e/2
Effective Width
1
-- t
I• .. l
Fig. 2-4 Effective Width of Compression
Flange in Bending
-215-
Principal Directions
,...
0
.....
u
aS
u.. Flat
Shear
>.. Test
..... I
•r-1
"O
•r-1
DC,
G
A
....C\
N
..... &. I
c:::
J-4
0.04
m
C)
..c:
[/)
.. s l -0 II
·~
"'
0.02
c in inches
:~c
4 8 12 16 20 24 28
0
....
N
....
I
S'-0"
.. r-
..."
0.02 S'-0"
C in inches
4 8 12 16 20 24
0
Fig. 2-8 Shear Rigidity Factor 'a.' Vs Hypar Curvature
(Double Layer of Decking)
-218-
e ex
1
xl
"" 2 n
ul Uz
wl exyl
lw 2 8xy2
e e
~
X3
ey4 C
U4
ey 3c/
3
"'- U3
W4
1e
xy4
w3 8 xy3
V3
o~_Y_L
M
y
M
xyc
0
x x
1
-
Mc )Mx
x
, Nxy
J
-
U--'M
Nxy
! y
M
y
! '
xy
l ---
!s-s
a) a • b • 80 inches
1
I
I t = 0.05 inch
E • 29.Sx10 6 lbs/inch 2
h/2 s-s \) • 0.30
IB q = 1 psi
s-s No. of Elements = 64
------+--- x (Full Plate)
0
0.00406qa ..
h/2 B = D
i.·- - - '-------+---.
s-s
...
-~L~- J
'
I
I- a/2 -·----..-- ...
1
--
1
y
b) a = 96 inches
b = 144 inches
t = s.o inch
E • 3xl0 1 lbs/ ·
b/2 F.E. inch 2
s-s 'V = 0 .. 30.
q = 200 lbs/;ft2-
~o. of Elements
• 18 (Half
Plate)
. ..
b/2 ~ '
0.0064a ..
s-s oB • D • •
a 2 a 2
l y
I
s-s
I c) a • b = 70.50 inches
t • 0.0149 inch
s--s 28G Std. Corrugated
I
lo
b/2 I> r-, •>
- D.e.ck ·
E = Z9 .sx10• lbs/ inch 2
·- ~I A v • O. llJ
- - ~
-x q = 0 • 3·0.. psi
I'(o .. ·of El~ments
b/2 •Varies _(Quadrant)
; ; 5B s !k ~t~-
a/2 a/2
....i::: I
4 s::: N
0
....
.µ
N
0
I
u
t>
.--4
'M
t)
i:l Deflection of
6 Beam Strip
B
= 6.96"
A
8 Present Study
A Series Solution
p
e
0 C.G S.C.
Y- IJt- - __,,__,_ __
z z
(a) (b) (c)
- - - - - - - - - - s. c.
y,v
-1
x
z ,w
Fig. 3-7 An Arbitrary Cross-Section of a Beam
z 'w
z
y
8
z
y,v
----b-----···--·- ·---------~ x ,u
ax
Fig. 3-9 Typical Eccentric Edge Member
-Plate e
. .__.
C.G., C.G. s.c.
s.c.
~Bracket
(a) (b)
•o 4
.....s::
- 2
0
No. of Elements
9.0 ....
~ --· ~~ ~~8 . 6 4,-,- w-,~~-=~-=:•r-e_e_T_.:.._r_s_i_o_n_(_I_I_a_)__c
..c:
u
~
=
•l"'i
"--- ~Warping Restrained (Ilb)
'<I
6.0
--~ -5.:~ -- . Q
4.0 2 4 6
No. of Elements
Fig. 3- l 3a Effect of Restrained Warping on
Vertical Deflection 'oqt
8.0
(IIb)
6.0
(!Ia)
~
0
....
....
1.1-l
4.0
2.0
6
0
No. of Elements
Fig. 3.13b Convergence Characteristics for
t 0
Qt and 'a Q'
x,u
al y,v
I
I
Column
_j__
A =A =A =A =A
1 2 3 4
B =B =B =B =B
1 2 3 4
c1=c 2=C 3=c 4=C
x- 1 =x- 2=x 3=x- 4=c
- - - -
Y1=y2=y3=Y4=o
y
(a)
A A
A =A =A =A =A
1 2 3 4
B 1 =B =B =B =B
2 3 4
c1=C 3=C
C2=C 4=-C
x1=x 4 =A
.__,._._ _-+-_ _ z_ __, _...,.. x ' -2 -
x =x =-A
3
/R(A,B,O)
y• . Y1=Yz=B
y
y =y =-B
(b) 3 4
pt Q'
s• R'
y
(a) (b)
Pr--_ _ _ _Q"'t----- x 0
ct
-
x
-----..... x'
x'
s
I -
y
!,Z'
I (a)
y'
(b)
y
-
x
y \
z z
-- Deck - Deck
1
--~Edge
Member
(a) Moment-Free (b) Sliding-Connection
Connection
rig. 3-ZO Deck and Edge Member Connections
Shear ;Deck
-~ I p
'. \
I
u"' ~·I ·f-+llo.'-....--...M'--i
1··~
I <{,-----.....
'J
(a) (b)
1
llnlf Band Width
1_~ _ r
' !'I
><
"
.... ""
"' ''
"' " "'
"' '
'
' ''
'
'
x ''
'
0 ~ 1
x
-- •· ··---. - ..... y
z x z
(a) 1
M-p
p
xl -[ __ '" --·-- .-_]-)---?-p x 1
p x 2 ···-{--l.___ _.I- ---p x 2 l:M
0
= 0
MoQ,
(b) MoP + (Pxz-Pxl)Zs
oQ = 0 - M-
A
--A ---1
h
I ·-
I
s/
b
Fig. 4-1 Structure Type I
g
1
<
y - -·
r; T---·-·-r:i ·-·· . ,b
Bl
Deck
.... > . 07 7" r-.
8. 871"
0
A.onfl}.16"
c [1 Rz
2
LYi·il
Typ. Deck
Section
1 2 3 4 s 6 7 8
h
I I I I I I I I a
B -
-
c ·-
,__. . -- -- -~~ -
-
e
E - ·-
F ---
G ·-
II -
c 0
J! .. _ _ _ _ _ _ _ 1...._ 2 _....
h___
tl"(TYP
- ~Varies
1. . 1
i c
, -1 / 2 ,, 0 Strucs.
f
i '6' 6 '8'
B 1/4" Thk
Shell
<l A a b
A
Plan
Strucs.
'6' & '8'
2-1/2" square
column
Struc. '7 1
Section P-P 2.
lo a
1. 0 .
I Symmetry
~/
2.0
**2.55
3.0
Scale
1" = 10- 2 cm
I .~-Symmetry
I // 1.0
r 2.0
I
1+2.46
3.0
* Method ta•
*Ill Method 'b'
+ Ref. 17
Fig. 4-6 Deflection Profiles
(Structure '1')
0.035
E
v
....c:
0
'O
c: Scale
....0
~
1" = ().0025 cm
u •
Q)
,..... 0.030 N
'+-!
C>
Q
"""'•
,...
QS
...e
0
z
0.025
2x2 4x4 6x6 8x8
Grid Size
'~
Fig. 4-7 Convergeace Characteristics for Deflection 0 ' (Structure '1')
-2l6-
I r Symmetry
r-·/
I·9.20*
9.32** 10
Scale
l"=Sxl0- 3 "
Jo b
T·----.--------_,,_-_,..,
I Symmetry
l ..
oxro 3
r
. 9. 18+
9.28+ 10
* Method 'a'
**
+
Method 'b'
Ref. 20
Fig. 4-8 Deflection Profiles
(St rue tu re '2.')
-237-
d e
z.41 Scale
l"•l"
f g
-·lilr·- Experimental
Analysis
9.32
7.50
7.50
Scale
l ":r.7. SO ksi
Stress~..-..-___.~-~ 9.32
7.50
9.83
7.50
Rosette
No.
•
NI
e,,e
~
•
e Anaj.ysis
-·--- Experimental
I A I B c I D I E I F I G I H
1- h -- - ·- .. __________,. _____ -- - - - c
r- - ----- -- ------:i
2 --
L - =::J
~------:-~
3 . ---
4---
L_ -- .:sJ Scale
1···=1000
v-------:--~ lbs/"
5-
~- -------. --_---1
6-
c
~-----------]
.· :=:J
7- L:--------~~---------- . ==
_____ _j ______ _
r Membrane Shear
= 653 lbs/inch
8- a o
Fig. 4-12 Shear Force Nxy lbs/inch
(Structure '5')
-241·
·:'
0.016
0.02 0.018
a b
o.o
Scale
1"•0.02"
0 b
0.02
- ·-A-- Experimental
-·-e·- Method 'a'
~1ethod 'b'
324
300
"' '
''
''
Vi
.D
r-i
200 /Membrane Theory
.....s::
Q)
u
"' ' 'l
""
0
1-1..
"'
100
+ "'
'''
a b
Tension Member
Scale l" • 100 lbs
324.0
300
''
.D
!/)
" "" ,../ Membrane
Theory
- · - 4 · - Experimental
Analysis
r-i 2 00 e
.....s::
GJ
u
""
''
0
µ .. . -A......
100
~~
0
Compression Member a
-
0
c
Analysis
80 /.
Axial Stress ~
4
.
t+-1
aJ
#
II) 60
Po
#
..""'
s:::
·~ I
N
~
aS
0 I
14
40 M•mbrane
Theory
I A I B I c I D I E I F I G IH I
[ -- - -- ---
1-~ b
~ c
•
,-~
~---·-
::sJ
2--
p=:
3--·- ._
_____________
--.---~
____:J
_____
4_r:--- ----~:']
5-r:
Scale
:J 1"=25 lbs/"
6-L ---:i
7 -- _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ___.
Membrane Shear
= 13.48 lbs/" - ,
--· ----- - - - --~-- -
a o
Fig. 4-16 Shear Force Nxy lbs/inch
(Structure '6')
-245-
0 a
-i!r. ---
. -- ---· --·-.o.
.
0.009
015
0.025
a b
o.oso
0 0
Scale
l"=0.025"
0. 02 5
o.oso
-·-A-·-· Experimental
e Analysis
Fig. 4 .. 17 Deflection Profiles
(Structure 1 7'1
-246-
324
300
t/j
..0
.......
zoo
.....s:!
160
Q)
u
J..<
0
µ..
100
+
a Tension Member b
Scale
1"=100 lbs.
300
Ill
..0
......
~
zoo
•..-!
(I,)
u
'""
0
µ...
100
0
Compression Member a
IA ja le ID IE IF IG ltt I
-------]
=::: ~
--- -- -
_k:::
-
v:--::
z-~
1 -~
- --j
-
3-~-·---~
4-c- -::sJ
s-r::: ---- ::1 Scale
ltt•2S
lbs/V
6_p=:---. ------1
7_
-- =- -=-----]-
k--==-====
MembTane
r3·
- ___ ___ _/
Shear
48lb-/" .
_.,......_
a o
Fig. 4·19 Shear Force Nxy lbs/inch
(Structure • 7 •)
-248-
0.10
d +
"""" ........... -~
.
0.068
-~0.098
0.10
0
0.10
0.10
0
--itr·- Experimental
Analysis
'
..__ob
~od __ . ~·.-::::
---- ---·.~· ~
.;,_...-·
./· /.>\.
40 - : _...,,....... - · ..ei- · - Experimental
,.,,... . '
1..- 0d Analysis
I .
. ./ ob
I
.I
I " I
....
N·
fO
'
I
I
.I
0.10 0.20 0.30 0.40 0.50 0.60 0.70
Deflections in inches
Fig. 4-21 Load vs Deflections (Structure 1
81)
-250-
Grid Size
4x 4 6x6 8x8
1. 60
l/l
CIJ 1. 574
..c:
u
i::
·~
i:::: 1. so
·~
..0
'<'
i::
0
·~
+-'
u
(!)
rl
l.H l.'!>72
Cl)
Cl
1. 30
0 • 3B <1--.__...._
0. 51 Scale
0.56 l "•O. 50 11
1. 0
1. so 1. 57
-·A-· - Experimental
• Method 'b'
o.s - -a- -- Method 'a'
1. 0
1. s
Fig. 4~23 Deflection Profiles
(Struc. •9 '}
-252-
6
o.s
Scale
1"=0.50"
1. 0
1. 5
- ·-A·- Experimental
Structure '9'
0. s ---&--- Structure '10'
1. 0
]. so
Fig. 4·-24 Deflection Profiles
(Strucs. '9' & '10)
6
-253-
3.62 Axial
. ·--A- S.tress
s.~3 1t'•5 ks i
5.52
Bending
Stres.s
1 u•lO ksi
19.0
0
0
Total
Stress
10.0 1"==10 ksi
-·-A·-·- Experimental
a
Fig. 4-25 Stresses in Compression Member
(Structures '9' and '10')
2.1slt;---~~-e----'il____..~-..---::~,,_.,.::::~;:::r-
2.1s
s.o
s. 2 -- ---· --- Axial Stress
1"•5 ksi
a
a--------..------r-----..,...----r-b
Bending
10.0 Stress
1"==10 ksi
17.3 17. 14
18.20
(1
Total Stress
10.0 1"111:10 ksi
-·-Ii.· - Experimental
0
Structure '9'
--.;a.--. Structure '10'
(J
I A I B Ic I D I E I F I G IH I
1~-------~~~----~~~
b ~------------~c
z-.~
3-L~ ~
4 L ::: ::::.; ~ 9 Scale
1"-10
lbs-inch/
inch
s C_75--- ~
6-·~~
7~-~
8----~~~------~~~~~..;;;a..--~-.
a o
I A I B Ic ID IE I F I G IH I
,~-------------1
1- [ - l
b c
r=- - -:...:;::----------=:..::-::...:- - -1
z_L. ~
3_ p:___~~- -w -j Scale
1"=100 lbs/"
4-r::= ,-:~-
[-- -----
s -~
[--- ----™
6_ t::::=
----]
7-~ :J
8_ [
Membrane Shear= 49.50 lbs/"
______ ------ ---j
a o
Deck
(a)
M=P II e
(b)
l_ __-;;::__.,,,,ar- - - - --
t.. f - ·-- ··--
- ·'-·-==--
-.
0
--- ---- ..__ ..........
2
M9.
2E!
(c)
0 0.38
0.50
Scale
l"=0.50 11
1. 0
1. 50
0 b
-·-A-· - Experimental
0.50 B.C. V
---e--- B. C. VI
1. 0
a b
10.0
B~ding Stress
13.4 1.,•10 ksi
Bending Stress
a b
Shear Loads
1"•25 lbs
/
/ B.C. v
50.0
75.0
Shear Loads
~ig. 4-31 Effect of in-plane Fixity on Bending Stress and
Vertical Shearing Loads on the Tension Member
(Structure '9')
-260-
()
--.--- r·-
-----a---e--a--- -a 0 . 0 4 3
---:r--
8
0.133
0.20
6
0 c
. I I•
_.,_ ..........,__-fJ--w..._...a0.053
0.20
6 Scale
l"=0.20"
0.29
6
0 b
Scale
o.so
1. 0
6
--- _--
_.....__
0.29
0.29
o.so
1. 0 0.91 Scale
l"=0.50"
Structure 1
11'
0
.--r ---...----r----r----.--.......--....---...,·- b
-...........
~
~ __ .-40.23
· '-.~o 42 0.26
• •·-&-·-'A
A-· _r,;;..__..·
0.50
() Structure '12'
0. 31
0.31
0.50
0.56 - · - A · - Experimental
e Analysis
Structure '1:5'
667
+
c b
Scale
a 1"=667 psi
667
0 c
0 Max. Membrane Stress
= 1570 psi
667 Analysis
0 a
1s.o~~~-_+__.._ _
a
Q_ _• __ ._I
• ___
150.0
c b
+
o----- I I I
c
30 0. 0 .
Scale
l"slSO lbs
150.0
+
0
a
IA IB Ic ID IE IF IG IH I
[------ - - -------]
l-~
h
.. ~
[-- - ---- ····-- ----]
c
2-v::
------- ·- ---··
~
I ---------------J
3-v
[---
~
----------------
4-v ·- ---····--------------1
~
Scale
1 11 = 100
lbs
5-~ ~
6 - ~- - --- - - - -- --s;J
7-v----------SJ
c---~----
8-~ .
------J
. _-Membrane Shear = 48 lbs/"
:'>J
.,
C· 0
a--~~-----------------...-.-----------------------.--~
o.os
~.
~ 0.092"
0.10 0.090
0.123
0 .... b
o.os
--4- Method I a'
Method I bf
0.10
·-- --- -- Ref. 21
'* Ref. 19
I 0.123
6
j 0.144
Section y = -zb
10-0
Scale
1 11 = 10
lbs/"
20-0
Scale
11! = 1.0
Section lb· -incftl'
a
x = 2
+
e Method 'a'
-·-·-Ref. 21
1. 33
Single Deck
Double Deck
1. 0
N
•
0\
.....
•
0.67
AB
2
20 30 40 50 60 79
~1
0 a
----·r-= 4 .. .. ---..:.._
- ,
·-
• - - • - - ·
---........._.._, a
(a)
Deck
-·
--- ---
-. -
.4fo_ __....,......_== - -
. -·--
. -·
- ·-
-··- . \a
--
0
(b)
a
............
-..
--~ 'A.....
'"'lL.
a.so "'It. 0. 4 8
0.12
Scale
1"•0.50"
0.4 ·er-. -6--._.._
·-&. ._...,
' -. ·--....._ ... o. 72
o. s1r-----..___
1. 20
0..--~~~-.-~~~~...,.--~~~.....,~~~~-,----b
.f
I. 0
I
I
I
I. 50
I
I 0.48
Ir Sym.
0.81
~/ Scale
1. 25 l"=O. SO"
Scale
3.27
l"=S ksi
5.06
/
s.o / --·-&:·--Without Edge Member wt
6.7 e With Edge Member wt
Linear
Elastic
Analysis
cSp II
Value of De-
formation at
the end of p I
First Step
i::::
......
1 I
"Cj f c N
C'd 0
.....,
0 N
....:I I
d a b
Point 'l' Point '2' Point '3' Key Sketch
1. 0 2.0 1.0 1. 0
I tS in cm.
g b
0.20 kgm/cm 2
I
I
/
1. 0
/
/
\ I
\ , 1 0. so kgll/cm 2 Scale
2.0 \ I 1''•1 cm
v
Compression Diagonal 'gb'
2.0 Buckled
0 . --
2.0
4.0
Scale
Compression Diagonal 'ob' 1 11 =2 cm
a ,. c
I
I
/
/
/
/
./
2.0
- - ---Buckled
Scale
Tension Diagonal 'ac' l"=l cm
2
kgm/cm
1. 0
2. 0
a c
0.20 Scale
1 ~'=I. Sent
a.so 0. 30
/
I I \ Buckling
I I I Load
I I
20 2 I I
200
b
4a
~
. 150
Cfl
p.
s::
•!""!
1 I
N
...a
"tJ
Cd
0 100
2 °'
I
....:I
3
-(do-
c
0
2.0
o in inches
Fig. 6-6 Buckling of a 28G Double Layer All-Supported Hypar (Structure '13')
-277-
\
.\ 23 psf
\
. 0\\ /"-·- 46 psf_ _ _ _ _ . /
\ I
/ ---
1. 0
'" Compression Diagonal 'ob'
0. 50
' ' 46 psf
1. 0
' \
\
v
,,,.....--- --- __
69 psf
.,,,.,,.,,
Scale
1 ''•0. SO"
0.50
1. 0
23 psf
0.50
46 psf Scale
'
p..
45 I Curved Elements
.....=
""C:J
t1S
Note: For Key Sketch
0
...:l
see Fig. 6-6
I
N
......
'°I
1. 0 1.0 1.0
tS in inches
Fig. 6-9 Buckling of a 24G Single Deck All-Supported Hypar (Structure '12')
-280-
o - · - · ·····T------.------y-----,..----,.-----..- h
0.25 ~
\
\
\
o.so \
\
\.,.....
....-- -,
____
'-.._ ............ SQ. 5 psf . /
/
0.75
Compression Diagonal 'ob'
0
0.25
36 psf Scale
0.50 l"=0.25"
0.75
60
so
.
'+-< 40
fJl
p,
i::::
·~
'"CJ
~
0
~ 30 Scale
l"•0.10" Deflection
1"•10 psf Load
20
,Ii!~
~
• l t
I I
'I I
I
. {
.~
,\
J, ,~
~
,~ }..
l
I
I I •
;:I'
4 t
I
,•' ~ I
f
I I
I~ ~ ~
I
,'
I
I
•
~
~
' f • I~
'Ii
•
I
I
'
JI IIf
~
nz y
' I
YH
Case Eguation ~
I p = tfEI/(k1) 2 1.0
er
II p • 11 2EI/(kl) 2 2.0
er .;
Ill Per • El/(kl) 2 0.699
IV 2 2
(q 1) s T El/ (kl) 1.122
er
v (ql) • ~El/(kl) 2 0.492
er
VI (ql) 2
s ir EI/(kl)
2
0.436
er 2 2
Vll (ql)
er
= 1T EI /(kl) 0.284
Direction of Shear
r-------------'""-:--------------------1Top
Layer
..._~..+or-o-.._.-____....,..,.toro---:.-~r""'r""~~---'."P"'ll:~...,...~....,,.rBottom
~~--.-&..-..-..;--~r----11111&---11~---......-..-.._----~~Layer
Edge Beam
Direction of Shear
1---~-------------~---------------~Top
...,_----:i.,.._________..,..i---__ ir--__,._~tor--,_.-~--.
Layer
Bottom
l-.l!K----::1.._......'l...._---i...,.._.;...,._~..__..;Jte;..___.....__ _ _ __...,"'-t Layer
Edge Beam
,
/
MH)O
J:.iOO
-)000
)'jOQ
1I
- I iOO
Il I
Figure 7. 7 1
iI
Lo.-1d versus Ocflcction
j
:flat SJ1:-ar T"st l\o. 14
1,;:io /'o
·l
j
l°
v --- ---- - ~ --
0.2
26 in inches
o.3 o.4 o.s
-- L ··-·· --·--- -'----~•--------
-0. :o
-1
~ \~ '
I
I
~ I
'
i
I: ~
tc I
~
I ..---1.
r-0002 3380
'I
I' !
I
! c in inches I
I
24 28
l~ 3 12 16 20
!
L.i::.10
1
33.;.00
I
I
I
0.06
I N
IQ
~
I
- - - - - F l . ' . l t Shear Test
10.04 1352G
No. 15 No.
Hypar Tests No.
1
~
321, 322
5'
I
!
i
I
I
\;
f
I
Iro.02 5760
! c in inches
I 4 s 12 16 20 24
! I I I I I
-Z9Z·
.26
• ll +.59
'.+.44~
+ .22 .
• 15 • l l • 01 • 01• • 01• 0
' :·:-.~.'E=-=-
1. ls .55 .66
.1 5-t-
.48 ..
+ + +·
-. ()7 .59 • 6()
+ +
--
.n •.-,2 ._io .17 '
+- +
•r:r. .ss
• l ') + +
• 73 14!)6 • 77
.18 + ·I-
+.33
• 22
-----·-· !------- ~---
.lG .22 .15
-------1---
.11
I
.1s
----·----------.......1.18
-1060 (+1550)
---·--·-
+2860
--· - - - - - --- - - - - - - - - - - - - t,.:_:__ _ _ _ _ _ _ _ _ _ _ _ - - ,
l__
· _ _-· . --:strong dir..
• 26
.13 j • 07
• '..iS
+-
.33
+
.07
.22 29 • L8 .22
+ ·+ + +
• 11
.l'.) .29
+
.T3 .13
+ +
• :? <) .MJ .26
+ + +
rl~.15
2') l ..
·-·- .11 .04
·:· 53 0 ( l 5 s0)
--- ----- ------ ----- - - - -+3500
--.r.=-----------------
~·
; ;1r010)
l 770
-1010
+1100
(-1160)
-880
+1100(1160)
·:-230
t +350 ++2580
+3860
-10 - t0:~c (-770)
+ + 1150
-;-t,:.;oo t :-1 l3r
+5620
-1 o:;o( s20)
-:-in
-'J40 ~ (-380)
+70
strong dlr.
. -·-· ·--····.40
·t ····---~--,..----~----·.26--- --- --------~.....-----'-0
.37
,.,, . 33 . '
• ) :.>
I '. • T :.
i . . ' 1
• aI 59 . ' ..
t- ~- ~·i- :
!
.17 - • 07 •l8 I:
.29
+
f ' ! '
• l :i
·r~ ·~
I] •ot~
.l-
'.~ () r .l l
.• 113
• 15 • :U1 • 22 - .04
+ + +
• 27.
1 u
I .01
i . '
. ' ' . ' '
\I . . . . '
I · ···I · ·· I· - -1-- -+----·--!1- --.---·------~~-- -- _: ------~---.J
• I u'·' 0 .1.5 .U7 .07 • 07 • Ol~ • 15
' '
' l ' '
i .
-1320 ,,(-770) .' +5960
+ +z9oonl . . ++10150
; . I
+-1120.
+!~3 70 + i ! .
' i
' . I
i
.
t
' l ' 1
·:-(.;.j()
+2200· r( 1. 030) • 1·3330
' l
'1·2970
·:·20')0
·•·'.Jo,;o+ + +t-560
+6l;10
+ '.' 'j!l('.)20) .
·•·')(;t
l
I
I
'
.300
----.:.....135
I •
I
.026 -.067
+ +
' .
'
I
I
.
' · .o:rn . 1
.029 .01s
• 063
+ + .032 .
+ !- .057. +
fr •<W2
II .
-~-·-----: taz · ·---~~1t1-- ·------·::lie-~---'·· .315
- • • - - .. ;· -- --·· • ·- ; I , •. '
: ... .,' .
.t~ 7
+
.14 .:.· . . ·+2 . : : •i~
l' - ..
I
' - -i -
I I '
' '
.76
.60
J ..
' ' I !
'
.os .. i .79
r
I
. I' I '
' . '
. \
l
·: t' •
I ·I • .. _.. i • l; i
: I ;
•· ••I
'
I ; ! ..
.I
• I
' ?
, ! : •
; : ~ !
I . I .
'' I' I ' :
I
i ' ''
'. ·Hit~o +s10 !· .. ; I '
·i-2C4o-!-+2270 · :
; ! ~
...
! ' j
':
' ..
.. ' :
L: .
I ! '
: _, I
l
I
I J "
'. .'
I
I '
9
July2017
www.enhsa.net/archidoct
ISSN 2309-0103
PROTOTYPING STRUCTURES
IN ARCHITECTURE
European Observatory
enhsa
enhsa
european
european
networknetwork
of Doctoral Research
of heads
ofof
heads
schools
of schools
of architecture
of architecture
in Architecture
ISSN 2309-0103
77 www.enhsa.net/archidoct
Vol. 5 (1) / July 2017
//
Shell structure: Analysis of hyperbolic paraboloid in paper
Ana Laura Rocha Peña
Abstract
In contrast to the opinion of many people, who think that technology will eliminate paper, tons
of paper are used daily worldwide.The annual average paper consumption is 48 kg per person,
which is equivalent to approximately 347, 035, 970.30 tons. In the last few years the architect
Shigeru Ban has motivated research in the field of paper as building material. In this paper, we
evaluate the implementation of paper as building material in shell structures. At first, paper
does not seem to be strong enough to be used as structural material, however we propose
shaping it, in order to improve its structural behavior.We have simulated a hyperbolic paraboloid
(HYPAR) of paper in Abaqus software based on the finite element method (FEM), and analyzed
its structural behavior.The analysis results demonstrate the feasibility of using paper as building
material.
Keywords
Shell structure, Paper as building material, Hyperbolic paraboloid, Form optimization.
Note
This essay is the result of an extensive research conducted for the author’s Master Thesis
at the Polytechnic University of Catalonia-Barcelona Tech (UPC), completed in October
2016. Due to the complexity of the calculations of the HYPAR, we first calculated a plate
by using the Navier method. Then, we compared these results with those obtained from
a simulation model of the same plate in Abaqus. Likewise, we use this model as reference
to compare it with a HYPAR structure and made a comparative analysis between paper
and concrete, since concrete is a material well known and by now, widely studied.
ISSN 2309-0103
78 www.enhsa.net/archidoct
Vol. 5 (1) / July 2017
//
Shell structure: Analysis of hyperbolic paraboloid in paper
Ana Laura Rocha Peña
Introduction
Shell structures are constructed systems described by three-dimensional curved surfaces, in which
one dimension is significantly smaller compared to the other two. They are form-passive and resist
external loads, predominantly through membrane stresses. A shell transfers external loads to its
supports predominantly through forces acting in the plane of the shell surface, which are called
membrane stresses. The shell surface is normally stressed in compression, or in combined com-
pression and tension. A ‘thin’ shell has to be sufficiently ‘thick’ to carry these compressive stresses
without buckling. Shell structures can be constructed as a continuous surface or from discrete ele-
ments following that surface. Geometric shapes such as sections of spheres, hyperbolic paraboloids
and regular polyhedrals are also commonly used.
The challenge is to find geometries that can work entirely in compression under gravity loading.
These geometries are not limited only to masonry, but are often built of any material. However, for
traditional masonry structures, the dominant load is often due to the self-weight of the structure,
and the applied live loads, with smaller effect, due to wind or snow (Adriaenssens, et al., 2014).
The antecedents of shell structures are founded in Gothic architecture evolving from the heavy
brick vaults to the slender and ribbed vaulted alloys.The most direct precedent in time is the barrel
vaults. These consisted of several layers of fine brick, the first of which is placed with the help of
small wooden guides and is overlaid with plaster paste, constituting a collaborative formwork. This
sheet reproduces the shape of the inner curve of the vault, the following layers are superimposed.
One of the most important developments in shell structures was the application of reinforced con-
crete. At the end of the Second World War (1945), due to lack of steel, reinforced concrete favored
the development of molds with its ability to work in compression and traction, as well as to provide
monolithic construction.
The construction of reinforced concrete shells became a process of study, experimentation and
innovation, whereas respective contributions made by Eugène Freyssinet, Eduardo Torroja and Felix
Candela were decisive (Carceles Garralón, 2007) (Figure 1).
The first cultures to use paper as building material were the Chinese and Ancient Egyptian ones. At
the beginning, it was used in form of papier-màché, followed by the development of papyrus by the
2nd century B.C. By the 9th century A.D., the Japanese culture started to use paper elements in the
construction of sliding doors and walls, called shoji-fusuma.That was the first time, when paper was
utilized as an interior building component. France was the first to use paper in furniture production
in the 19th century, and later as wall covering, introducing for the first time in history paper as a
decorative element.
Paper products were being used in the production of aircraft and tank components in World War I.
When realized that aluminum had problems with expansion and shrinking, the substitute for alumi-
num sheeting on aircraft wings came in form of plaster-made molds for shaping and cellulose rein-
forced sheets of paper combined with starch or similar adhesives. By 1920’s, paper and cardboard
started to be used as electrical insulation in the United States. In the same period impregnation
experiments began with the introduction of cellulose fiber laminates in industry. First, phenolic-res-
in was used, until the development of melamine resins led to an increased popularity of paper and
ISSN 2309-0103
79 www.enhsa.net/archidoct
Vol. 5 (1) / July 2017
//
Shell structure: Analysis of hyperbolic paraboloid in paper
Ana Laura Rocha Peña
In following decades, several architects began to experiment with paper as structural material. The
first building principally constructed out of cardboard was ‘The 1944 House’ that was followed
by a period of slow development in the field. Several architects have influenced the progress of
cardboard applications in architecture, with the two most influential figures, Buckminster Fuller in
1950’s and Japanese architect Shigeru Ban, most recently (Sekulic, 2013) (Figure 2).
Geometry Definition
Among different shell structural configurations, a hyperbolic paraboloid was selected for the sim-
ulation. This is an anticlastic surface, whereas the center of curvature is located on opposite sides
of the surface. In this type of structure, normal loads are transferred to its surface by tangential
stresses (compression in the convex curve and traction in the concave curve).
z=kxysin(ω)
Another way to understand this surface is to consider it as generated by a main parabola P1 moving
parallel to itself along another main parabola P2. Thus the surface has two systems of paraboloid
generatrices (Oliva Quecedo, et al., 2011).
There are three significant factors, which determine the mechanical properties of paper: The prop-
erties of fibers, the interfiber bonding and the geometrical disposition of the fiber (Figure 3). It’s im-
portant to know how these factors influence the properties of paper. In laboratories, when sheets
are produced for experiments, even if these have the same composition as machine produced ones,
they do not have the same properties, because paper properties do not depend only on the com-
position, but also on the production process. Laboratory prepared paper is different than machine
prepared paper (Sekulic, 2013).
Table 1 contains a collection of directly measured values of the elastic stiffness parameters for a
few paper grades. Many values are missing, because of measurement difficulties caused by the small
thickness of paper.Various estimation schemes have been developed to avoid direct measurements.
The table demonstrates that the ZD (Z-direction) stiffness of paper is generally low compared to
the in-plane values. The negative value of the Poisson ratio Vxz for the paperboard shows that uni-
axial tensile loading in MD (machine direction) increases with thickness. In compression, at least, the
elastic moduli, perhaps even the Poisson ratios, are usually equal to the corresponding tensile values.
In general, the density of paper is between 300 and 900 kg/m³.The elastic modulus usually increases
ISSN 2309-0103
80 www.enhsa.net/archidoct
Vol. 5 (1) / July 2017
//
Shell structure: Analysis of hyperbolic paraboloid in paper
Ana Laura Rocha Peña
Figure 1.
Los Manantiales restaurant, Félix Candela
Figure 2.
Japan Pavilion, Expo 2000 Hannover, Germany by Shigeru Ban
ISSN 2309-0103
81 www.enhsa.net/archidoct
Vol. 5 (1) / July 2017
//
Shell structure: Analysis of hyperbolic paraboloid in paper
Ana Laura Rocha Peña
with density and ranges from 1000–9000 MPa, when the effect of anisotropy is removed by
averaging over MD and CD (Cross-machine direction) (Niskanen, 2011).
Simulation Process
Analysis Parameters
Following the geometry exploration process, a hyperbolic paraboloid was chosen for the
prototype design. For the simulation, an initial model was used starting from a quadrilateral
of 6000 x 6000 mm, with height of 5000 mm, and thickness varying from 60 to 10 mm.
We have simulated the hyperbolic paraboloid of paper in Abaqus software based on FEM
(Abaqus, 2004), in order to investigate its mechanical behavior (Figure 4). To this model,
two general analyses (linear and nonlinear) have been conducted. A static linear analysis
provides a first approximation of the structure’s behavior, by considering equilibrium of the
system without deformation. For obtaining precise results with regard to the response of
the structure, a second-order analysis is required that incorporates the effects of material
and geometry nonlinearity. The nonlinear analysis considers the properties of the material,
the surface loads and the boundary conditions. All elements are assigned with an elastic
modulus of 5420 MPa and a Poisson’s ratio of 0.38, selected from the table of measured
values of elastic stiffness (Niskanen, 2011). A uniform vertical load of 1 kN (0.001 N/m²) is
applied to the surface. As to the boundary conditions, two of the four nodes are pinned with
zero displacement.
The development of the simulation model provides information about the physical behavior
of the Hypar structure in paper. In all cases, the maximum deformation and the maximum
stress by Von Mises in the center of the Hypar have been registered. Probably the most
important properties of structural materials are their strength and stiffness. The limit stress
refers to the maximum strength value of the material. The type of paper chosen for the
analysis has a maximum strength value of 5.00 MPa. Currently no standard regulations exist
with regard to paper as building material. We have chosen a maximum deformation limit of
L/100, i.e. 8485/100= 85 mm.
To calculate the longest distance of the system, the following equation applies:
Lmax=L.√2
where L is a side of the quadrilateral that makes the Hypar.
Only maximum values of the stress and deformation in the center of the Hypar have been
considered in the analyses, and any respective results superior to the limits set were dis-
missed. For better comprehension, the analysis process is divided in three stages. The first
stage, static linear analysis, considers the equilibrium of the structure without deformation.
Then a second stage, static nonlinear analysis, takes into account effects of the deformed
geometry. And finally, a third stage, plastic nonlinear analysis, considers effects of nonlinearity
of the geometry and the material behavior.
ISSN 2309-0103
82 www.enhsa.net/archidoct
Vol. 5 (1) / July 2017
//
Shell structure: Analysis of hyperbolic paraboloid in paper
Ana Laura Rocha Peña
Figure 3.
Wood fibers, interfiber bonding and geometrical disposition of the fiber
Table 1.
Measured values of elastic stiffness parameters in tensile loading for some machine made papers (Niskanen, 2011)
Figure 4.
Simulation model, visualization of the Von Mises stress and deformation
ISSN 2309-0103
83 www.enhsa.net/archidoct
Vol. 5 (1) / July 2017
//
Shell structure: Analysis of hyperbolic paraboloid in paper
Ana Laura Rocha Peña
The initial model was ideal for obtaining a first approximation of the Hypar behavior. As
shown in Figure 5, the highest stresses develop at the ground supports areas and the
center of the Hypar on both sides (superior and inferior). This is due to the fact that
the shell transfers the external loads to its supports through forces acting in plane of its
surface. For the specific analysis, the resulting stress values in the center of the Hypar
have been registered.
Figure 6 summarizes the Von Mises stresses in the center of the Hypar obtained from the
static linear analysis. Based on the analysis conducted, the maximum stress increases with
thickness reduction, even above the material strength limit.
The specific Hypar structure works along one axis as an arc and along the other axis as
a suspended arc. While compression stresses tend to deform the membrane along one
axis, traction stresses along the other axis, tend to counter this deformation. For this rea-
son, the maximum deformation develops in the center of the Hypar as shown in Figure 7.
Figure 8 summarizes the maximum deformations obtained from the static linear analysis.
Two systems have higher deformations than the respective limit set.
The difference between linear and nonlinear analysis is the system’s stiffness. When a
structure deforms under an external load, it is the stiffness that changes due to the ge-
ometry or the material properties. This second stage of nonlinear analysis comprises a
static nonlinear analysis that considers the system’s deformation effects. In Figure 9, the
results of the simulation model are presented. The last two systems do not converge.
Figure 10 summarizes the Von Mises stresses in the center of the Hypar obtained from
the static nonlinear analysis. The last two models with 20 and 10 mm thickness do not
converge.
As shown in Figure 11 the maximum deformations develop in the center of the Hypar.
Two of the six systems do not converge.
Figure 12 summarizes the deformations obtained from the static nonlinear analysis. The
last two systems with 20 and 10 mm thickness do no converge. The other four systems
develop favorable results within the respective allowable limits set.
Finally, for obtaining most accurate results, mostly similar to reality, with regard to the
Hypar structural behavior, a plastic nonlinear analysis has been conducted. The analysis
at this stage considers both, the effects of nonlinearity with regard to the geometry and
the material. Figure 13 shows the results of the simulation models.
ISSN 2309-0103
84 www.enhsa.net/archidoct
Vol. 5 (1) / July 2017
//
Shell structure: Analysis of hyperbolic paraboloid in paper
Ana Laura Rocha Peña
Figure 5.
Results of the simulation
model - Von Mises stress in
static linear analysis
Figure 6.
Structural response -Von
Mises stress in the center of
the Hypar based on static
linear analysis
Figure 7.
Results of the simulation
model - Deformation in
static linear analysis
Figure 8.
Structural response - Defor-
mation in the center of the
Hypar based on static linear
analysis
ISSN 2309-0103
85 www.enhsa.net/archidoct
Vol. 5 (1) / July 2017
//
Shell structure: Analysis of hyperbolic paraboloid in paper
Ana Laura Rocha Peña
Figure 9.
Results of the simulation
model - Von Mises stress in
static nonlinear analysis
Figure 10.
Structural response - Von
Mises stress in the center of
the Hypar based on static
nonlinear analysis
Figure 11.
Results of the simulation
model - Deformation in
static nonlinear analysis
Figure 12.
Structural response - De-
formation in the center of
the Hypar based on static
nonlinear analysis
ISSN 2309-0103
86 www.enhsa.net/archidoct
Vol. 5 (1) / July 2017
//
Shell structure: Analysis of hyperbolic paraboloid in paper
Ana Laura Rocha Peña
Figure 13.
Results of the simulation
model - Von Mises stress in
plastic nonlinear analysis
Figure 14.
Structural response - Von
Mises stress in the center of
the Hypar based on plastic
nonlinear analysis
Figure 15.
Results of the simulation
model - Deformation in
plastic nonlinear analysis
Figure 16.
Structural response - De-
formation in the center of
the Hypar based on plastic
nonlinear analysis
ISSN 2309-0103
87 www.enhsa.net/archidoct
Vol. 5 (1) / July 2017
//
Figure 14 summarizes the Von Mises stresses in the center of the Hypar obtained from the plastic
Shell structure: Analysis of hyperbolic paraboloid in paper
Ana Laura Rocha Peña
nonlinear analysis. In this particular stage, all simulation systems show favorable results within the
allowable limits set.
The maximum deformation of the systems develops in the center of the Hypar, as shown in Figure
15.
Figure 16 summarizes the deformations obtained from the plastic nonlinear analysis. None of the
systems exceeds the allowable limits set.
Conclusions
Research activities in the field of paper as building material have increased in the last years. Paper is
an excellent material with regard to providing innovative, new ways of application in construction.
This material can be used as structural material for construction and as formwork of complex
structures, in both cases, offering opportunities for sustainably and economically sensitive designs.
Following analyses of a hyperbolic paraboloid based on FEM, the following can be concluded: In a
first stage of a static linear analysis, as well as in a second stage of a static nonlinear analysis con-
sidering deformation effects, the response values obtained are almost in the range of the allowable
limits set in the analysis, except for the models with 20 and 10 mm thickness that have passed these
limits. Finally, in a plastic nonlinear analysis, most similarly to reality, the results obtained are all fa-
vorable and within the allowable limits set. The developed stresses of the models obtained values
within the range of 0.56 to 4.79 MPa, all below the respective maximum strength of the material
of 5.00 MPa. All system maximum deformations are less than L/100, starting from 3 mm for 60 mm
surface thickness and reaching 73 mm for 10 mm thickness. Therefore, the results obtained from
these three stages of analyses, favor implementation of paper as building material in shell structures.
Future research will investigate the mechanical behavior of the different types of paper, ways for in-
crease of the strength of paper in combination with glued composites, of reduction of the humidity
of paper, and among others, the fiber of paper as material for the 3D-printer.
ISSN 2309-0103
88 www.enhsa.net/archidoct
Vol. 5 (1) / July 2017
//
Shell structure: Analysis of hyperbolic paraboloid in paper
Ana Laura Rocha Peña
References
ABAQUS, 2004. ABAQUS keywords: Reference Manual: Versión 6.5. West Lafayette:
ABAQUS Inc.
Adriaenssens, S., Block, P.,Veenendaal, D. and Williams, C., 2014. Shell Structures for
Architecture : Form-finding and Optimization. London : Routledge.
Carceles Garralón, F., 2007. El paraboloide hiperbólico comogenerador inagotable en las
estructuras laminares. E.U. de Arquitectura Técnica (UPM).
Niskanen, K., 2011. Mechanics of Paper Products. Berlin: De Gruyter.
Oliva Quecedo, J., Antolin Sanchez, P., Cámara Casado, A. and Goicolea Ruigómez,
J.M., 2011. Análisis estructural de algunas obras de Félix Candela mediante modelos de
Elementos Finitos. Hormigón y Acero. E.T.S.I. Caminos, Canales y Puertos (UPM).
Sekulic, B., 2013. Structural Cardboard: Feasibility Study of Cardboard as a Long-term
Structural Material in Architecture. Universitat Politècnica de Catalunya.
Introduction to Shell Structures
• The curved form may lead to different failure modes and often
unexpected behavior occurs
• The analytical formulas are very complex and complicated in
comparison with all the other structural forms
• Shell structures are very attractive light weight structures which are
especially suited to building as well as industrial applications.
Introduction to Design of Shell Structures
Range of application
8
Introduction to Design of Shell Structures
General
• Built structural shells
Introduction to Design of Shell Structures
General
• Built structural shells
Reinforced concrete
Steel
Aluminium alloys
Plastics
Glass
Timber
Introduction to Design of Shell Structures
Structural typologies
Circular cylinder/cone
Introduction to Design of Shell Structures
Structural typologies
Shells are the most difficult form of structure to analyse and the form with
the most complex behaviour. As a result, all but the simplest conditions
must be analysed using computers.
Introduction to Design of Shell Structures
Examples – Steel reticulated dome
US pavilion Expo 67 Montreal
Architect: Buckminster Fuller & Shoji Sadao
1 - Form-finding model
2 - Interior
3 - Mesh detail (steel bands resist shear)
Introduction to Design of Shell Structures
Examples – Timber-steel free form grid shell
Architect: Thomas Herzog
Engineer: Julius Natterer
Curved
shapes Bending stress state
Continuous
Membrane stress state
Plated
Shell structures
Reticulated (bar structures)
Conceptual design
Design for strength and buckling
Detailing
Introduction to Design of Shell Structures
Behavioural phenomenology of shells
• Behavior of a given structure (slender!) can be controlled by design if
the three characteristic ranges of load-deformation curve are correctly
defined
• Pre-critical range P
• Critical point (or range)
• Post-critical range
Pcr Post-critical
Critical point range
Pre-critical range
N xsup
,cr
N xinf
,cr
L
L
Introduction to Design of Shell Structures
Behavioural phenomenology of shells
• Instability phenomenon – Jump of Equilibrium or Snap Through Instability
• Affects shallow arches and shells, reticulated shells
u length
w Perfect bar (unloaded) P
(unloaded)
w
length
P
u Perfect
length w
cylinder Perfect
(unloaded) P plate
u
P
P Perfect plate
P Perfect P
bar Perfect cylindrical
Pcr Pcr shell
Pcr
A
Imperfect
Imperfect plate
bar Imperfect cylindrical
shell
w0
w w0
w
w0 w
2E t
2
cr , p
31 2 b
2E t
2 2
E b
cr ,c 2
2
31 b 4 r
Stable Unstable
component component
increase in sensitivity to
increase in critical load
geometrical imperfections
Introduction to Design of Shell Structures
Favourable and unfavourable effects of spatiality
• Curvature effect in axi-symetrical compression
increase in sensitivity to
increase in critical load
geometrical imperfections
Introduction to Design of Shell Structures
Coupled instabilities for plate and shell elements
W – weak interaction
M – moderate interaction
S – strong interaction
VS – very strong interaction
Introduction to Design of Shell Structures
Coupled instabilities for plate and shell elements
• Erosion of Theoretical Critical Buckling Load
Introduction to Design of Shell Structures
Instability phenomena: Influence of imperfections
• Agreement of theoretical and experimental values
bars
shells
Introduction to Design of Shell Structures
Instability phenomena: basic types and models
• Dynamic propagation of instability or progressive instability
• Domino effect
(double layer grids)
• Instability propagation
(single layer reticulated shells)
Introduction to Design of Shell Structures
Models and Methods of Analysis
• Pre-Critical, Critical and Post-Critical Analysis
a)
c)
Introduction to Design of Shell Structures
Methods of Analysis
• Basic Equations
• Simplified Linear Shell Theory
• The Love-Kirchhoff assumptions (simplified model)
• The shell thickness is negligibly small in comparison with the least radius of
curvature of the shell middle surface (shell is thin)
• Strains and displacements that arise within the shell are small (products of
deformations quantities that occur in the derivation of the theory may be
neglected, ensuring that the system is described by a set of geometrically
linear equations)
• Straight lines that are normal to the middle surface prior to deformation
remain straight and normal to the middle surface during deformation and
experience no change in length (analogue to hypothesis for beams – plane
sections before bending remain plane after bending)
• The direct stress actin in the direction normal to the shell middle surface is
negligible (not valid in the vicinity of concentrated transverse loads)
Introduction to Design of Shell Structures
Methods of Analysis
• Model of an axi-symmetrical Loaded Shell
Model of an axisymetrically
loaded shell
N
N r0
r1 N r1 cos Yr1r0 0
N
r0 N r1 N r1 cos Xr0r1 0
N N
z0
r1 r2
Introduction to Design of Shell Structures
Methods of Analysis
• Cylindrical Shells
• Bending Theory Axisymmetric Loading
dN x
a dx d 0
dx
dQx
a dx d N dx d Z a dx d 0
dx
dM x
a dx d Qx a dx d 0
dx
d 4w Eh Eh3
wZ D
;
D
dx 4
a 2 12 1 3
Introduction to Design of Shell Structures
Methods of Analysis
• Shells
• General Bending Theory
N x N x
a 0
x
N N x
a Q 0
x
Q N x N x
Q a 0
a x N q a 0 x
x
N N x M x 1 M
M x M a 0
a aQ 0 x x a
x
2 M x 2M x 2M x
1 M
2
M x M x N a qa 0
a aQx 0 x x2 x a 2
x
Introduction to Design of Shell Structures
Methods of Analysis
• Basic Equations
1. Equilibrium equations 5 8
(static)
2. Deformability 9 12
compatibility (geometric)
3. Physical aspect 6 ---
TOTAL 20 20
Introduction to Design of Shell Structures
Methods of Analysis
• Buckling of Cylindrical Shells in Compression
• General Case
Equilibrium equations for elastic buckling:
N x N yz
a 0
x
x, u
N y N x 2 v M xy M x
l a aN x 2 0
z, w x x x a
2w 2 M x M yx M y M xy
2 2 2
aN x 2 N y a 0
y, v
x x 2 x a 2 x
with solutions:
2a mx
u A sin n cos
l
mx
v B cos n sin
l
mx
w C sin n sin
l
v0
n0 u, w f x axial symetrical buckling
Introduction to Design of Shell Structures
Methods of Analysis
• Cylindrical Shells
• Membrane Theory Application for Wind Action
q 2
N cos
r0
N qr0 cos
Introduction to Design of Shell Structures
Methods of Analysis
• Simplified Design Formulae (Cylindrical shells)
• Two possible approaches
• Overall column buckling if l/r ratio is large
• Shell buckling which involves the cross section deformation and can
be, in general, either:
• Axisymmetric, when the displacement are constant around
circumferential section
• Asymmetric (chessboard shape), when waves are formed in
both axial circumferential directions
r=a
l
Introduction to Design of Shell Structures
Methods of Analysis
• Simplified Design Formulae
• Axial-symmetric buckling of cylindrical shell in compression
d 4v d 2w w
D 4
Nx 2
Eh 2
0
dx dx a
Eh3
D
12 1 2
Radial displacement:
m x
w A sin
l
N
Elastic critical axial stress ( cr cr )
h
cr
Eh m Eh
; 4 2
a 3 1 2 l a D
l
For =0.3 1.72 ah
m
• In case of axial-symmetrical buckling, the critical shear does not depend of cylinder length!
• If one of the cylinder ends is free (w0), cr drops to 38% compared to simple supported case.
• Cylinder is highly sensitive to tangential displacements at the boundaries. If v 0, critical
stress drops to 50%!
Introduction to Design of Shell Structures
Methods of Analysis
• Simplified Design Formulae
• Axial-symmetric buckling of cylindrical shell in compression
Post-elastic critical buckling
h EEt
cr
fy E tg 0
a 3 1 2
fp
Et tg l Et
1.72 ah
m E
0
Introduction to Design of Shell Structures
Methods of Analysis
• Simplified Design Formulae
• Axial-symmetric buckling of cylindrical shell in compression
• Post-critical buckling: stable and unstable components
Ncr m2 2 E l2
cr D 2
h hl 2 2 2
2 a D m
m 2 Eh h
2
Ncr D
l a2 m
2 cr 0.5 f y
n 1 1
2
f y M1
Graphical presentation of
2 cr 0.5 f y ECCS “knock-down” factor
n
1 0.41231.2
fy
Introduction to Design of Shell Structures
Methods of Analysis
• Simplified Design Formulae
• Buckling of cylindrical shells under external pressure
• Membrane (hoop) stress in practical range
external
pa
y ; x 0 uniform
h pressure
pa
y ; x 0.5
h
• Von Misses formula for critical pressure, cr
Eh 1 h2 2 2n 2 1 “hydrostatic”
cr n 1
a n2 1 32 12a 2 3
type pressure
2
a n
1 ; 3 1 Simplified formulae for long cylinders
1
l
1 Eh3 n 2 1
h2 2 2 1 1
2
Eh 1 pcr
12a 1
pcr 2 2 n 3 1 3 2
a n 3 12a 2 2 2
n min pcr
Introduction to Design of Shell Structures
Methods of Analysis
• Simplified Design Formulae
• Buckling of cylindrical shells under external pressure
• Principle of ECCS approach
pu 1
0 1
py 1 2
pu
1 2 ; 0.5
py
py
cr
p
h
Pcr E min
a
1.5 ECCS design strength for
0.855 a h
min unstiffened cylinder
for l/a0.5
1 2 l a under uniform pressure
Introduction to Design of Shell Structures
Methods of Analysis
• Simplified Design Formulae
• Buckling of cylindrical shells under external pressure
• Principle of ECCS approach
• Wind action is more complex than simply an external pressure
• It is needed to check the cylinder stability separately for:
• Wind radial pressure
• Wind axial effects
• Wind tangent effects
• Interaction of the three
• Approximately, wind critical pressure can be taken as 1.6 times critical
external pressure (Maderspach, Gaunt, Sword )
• ECCS Design Recommendations (No. 125/2008) offers also a solution
Introduction to Design of Shell Structures
Methods of Analysis
• Simplified Design Formulae
• Buckling of cylindrical shells under compression and external pressure
x p
x ,cr pcr
Interaction curve
Introduction to Design of Shell Structures
Methods of Analysis
• Simplified Design Formulae
• Buckling of cylindrical shells in bending
• (Flügge)
cr , x
M Nx
1.33 cr , x
0.99
M cr Eh 2a
1
2
Introduction to Design of Shell Structures
Methods of Analysis
• Simplified Design Formulae
• Buckling of cylindrical shells in torsion
• (Swerin and Flügge) long cylinders
32
E h
cr 34a
3 2 1 2
• Types of analysis:
• Global analysis
• Membrane theory analysis
• Linear elastic shell analysis
• Linear elastic bifurcation analysis
• Geometrically nonlinear elastic analysis
• Materially nonlinear analysis
• Geometrically and materially nonlinear analysis
• Geometrically nonlinear elastic analysis with imperfections included
• Geometrically and materially nonlinear analysis with imperfections
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Basis of design and modelling
• Shells shall be designed in acc. with EN1990 and, in particular, to satisfy the following
requirements:
• Overall equilibrium
• Equilibrium between actions and internal forces and moments
• Limitation of cracks due to cyclic plastification
• Limitation of cracks due to fatigue
• Types of analysis:
• Global analysis
-linear elastic material law
• Membrane theory analysis
- linear small deflection theory (undeformed
• Linear elastic shell analysis
geometry)
• Linear elastic bifurcation analysis
• Geometrically nonlinear elastic analysis
• Materially nonlinear analysis
• Geometrically and materially nonlinear analysis
• Geometrically nonlinear elastic analysis with imperfections included
• Geometrically and materially nonlinear analysis with imperfections
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Basis of design and modelling
• Shells shall be designed in acc. with EN1990 and, in particular, to satisfy the following
requirements:
• Overall equilibrium
• Equilibrium between actions and internal forces and moments
• Limitation of cracks due to cyclic plastification
• Limitation of cracks due to fatigue
• Types of analysis:
• Global analysis
• Membrane theory analysis LBA
• Linear elastic shell analysis - linear elastic material law
• Linear elastic bifurcation analysis - linear small deflection theory
• Geometrically nonlinear elastic analysis- imperfections of all kinds are ignored
- the basis of the critical buckling resistance
• Materially nonlinear analysis
evaluation
• Geometrically and materially nonlinear analysis
• Geometrically nonlinear elastic analysis with imperfections included
• Geometrically and materially nonlinear analysis with imperfections
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Basis of design and modelling
• Shells shall be designed in acc. with EN1990 and, in particular, to satisfy the following
requirements:
• Overall equilibrium
• Equilibrium between actions and internal forces and moments
• Limitation of cracks due to cyclic plastification
• Limitation of cracks due to fatigue
• Types of analysis:
• Global analysis
• Membrane theory analysis
GNA
• Linear elastic shell analysis -change in the geometry of the structure
• Linear elastic bifurcation analysis - the elastic buckling load of the perfect
• Geometrically nonlinear elastic analysis structure
• Materially nonlinear analysis
• Geometrically and materially nonlinear analysis
• Geometrically nonlinear elastic analysis with imperfections included
• Geometrically and materially nonlinear analysis with imperfections
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Basis of design and modelling
• Shells shall be designed in acc. with EN1990 and, in particular, to satisfy the following
requirements:
• Overall equilibrium
• Equilibrium between actions and internal forces and moments
• Limitation of cracks due to cyclic plastification
• Limitation of cracks due to fatigue
• Types of analysis:
• Global analysis
• Membrane theory analysis
• Linear elastic shell analysis
• Linear elastic bifurcation analysis MNA
• Geometrically nonlinear elastic analysis- gives the plastic limit load and the plastic
• Materially nonlinear analysis strain increment Δε
• Geometrically and materially nonlinear analysis
• Geometrically nonlinear elastic analysis with imperfections included
• Geometrically and materially nonlinear analysis with imperfections
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Basis of design and modelling
• Shells shall be designed in acc. with EN1990 and, in particular, to satisfy the following
requirements:
• Overall equilibrium
• Equilibrium between actions and internal forces and moments
• Limitation of cracks due to cyclic plastification
• Limitation of cracks due to fatigue
• Types of analysis:
• Global analysis
• Membrane theory analysis
• Linear elastic shell analysis
• Linear elastic bifurcation analysis
• Geometrically nonlinear elastic analysis
GMNA
• Materially nonlinear analysis - gives the geometrically nonlinear plastic limit
• Geometrically and materially nonlinearload
analysis
and the plastic strain increment
• Geometrically nonlinear elastic analysis with imperfections included
• Geometrically and materially nonlinear analysis with imperfections
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Basis of design and modelling
• Shells shall be designed in acc. with EN1990 and, in particular, to satisfy the following
requirements:
• Overall equilibrium
• Equilibrium between actions and internal forces and moments
• Limitation of cracks due to cyclic plastification
• Limitation of cracks due to fatigue
• Types of analysis:
• Global analysis
• Membrane theory analysis
• Linear elastic shell analysis
• Linear elastic bifurcation analysis
• Geometrically nonlinear elastic analysis
• Materially nonlinear analysis GNIA
• Geometrically and materially nonlinear analysis
- where compression or shear stresses
• Geometrically nonlinear elastic analysis with imperfections
dominate in the shell included
• Geometrically and materially nonlinear analysis
- elastic with imperfections
buckling loads of the "real" imperfect
structure
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Basis of design and modelling
• Shells shall be designed in acc. with EN1990 and, in particular, to satisfy the following
requirements:
• Overall equilibrium
• Equilibrium between actions and internal forces and moments
• Limitation of cracks due to cyclic plastification
• Limitation of cracks due to fatigue
• Types of analysis:
• Global analysis
• Membrane theory analysis
• Linear elastic shell analysis
• Linear elastic bifurcation analysis
• Geometrically nonlinear elastic analysis
• Materially nonlinear analysis
• Geometrically and materially nonlinearGMNIA
analysis
GNIA
• Geometrically nonlinear elastic analysis- gives
withthe
imperfections
elasto-plasticincluded
buckling loads for the
- where
• Geometrically and materially nonlinear"real" compression
analysis with or shear stresses
imperfections
imperfect structure
dominate in the shell
- elastic buckling loads of the "real" imperfect
structure
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Basis of design and modelling
• Shells shall be designed in acc. with EN1990 and, in particular,
Material to satisfy the following
Shell
Type of analysis Shell theory
requirements: law geometry
• Overall
Membrane equilibrium
theory of shells
membrane not
perfect
equilibrium applicable
• Equilibrium between actions and internal forces and moments
Linear elastic shell linear bending
• Limitation of
analysis (LA) cracks due to cyclic
and plastification
stretching
linear perfect
• Limitation
Linear of cracks due to fatigue
elastic bifurcation linear bending
linear perfect
analysis (LBA) and stretching
• Geometrically
Types of analysis:
non-linear
non-linearapproximatelinear treatments of perfect
certain parts of
• Global
elastic analysis (GNA)
analysis
the structure
Conditions of use:
•Materially
Membranenon-linear
theory analysis linear non-linear perfect
analysis (MNA) --linear
the boundary
elastic conditions
material laware appropriate for
• Linear elastic shell analysis
Geometrically and materially transfer
- linear
LBA smallof thedeflection
stresses in theory
the shell(undeformed
into support
• Linear elastic bifurcation
non-linear analysis (GMNA)
non-linear
analysis reactions
non-linear
without causinglaw
perfect
bending effects;
• Geometrically nonlinear elastic analysis geometry)
GNA - linear elastic material
Geometrically non-linear -MNAthe shell geometry varies smoothly in shape
-change- linear insmall
the deflection
geometry theory
of the structure
• elastic
Materially nonlinear
analysis with analysis non-linear linear imperfect
-(without
the
GMNA discontinuities);
- imperfections
elastic of
buckling all
loadkinds arethe
ofand
the ignored
perfectplastic
• Geometrically and materially nonlinearGMNIA
imperfections (GNIA) - gives
analysis
- the
the
loads
plastic
have
limit
a smooth
load
distribution (without
-
structure
- the
gives
GNIA basis
the of the critical
geometrically buckling
nonlinear resistance
plastic limit
• Geometrically nonlinear elastic analysislocally
Geometrically and materially strain
-load
gives increment
withconcentrated
imperfections
the
evaluation
Δε
or included
elasto-plastic point loads).
buckling loads for the
non-linear analysis with non-linear - whereandcompression
the plastic strain
non-linear increment
or shear stresses
imperfect
• Geometrically and materially nonlinear analysis with imperfections
"real" imperfect structure
imperfections (GMNIA) dominate in the shell
- elastic buckling loads of the "real" imperfect
structure
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Materials and geometry
• The rules in EN 1993-1-6 are not limited to steel shell structures
• The standard is valid for isotropic shells and shell segments made from any materials
that may be represented as ideal elastic-plastic
• For materials with no well defined yield point, 0.2% proof stress can be taken
• The material properties apply to temperatures not exceeding 150 ºC (otherwise see EN
13084-7, 2005)
• Where materials has a significant different stress strain curve, there are alternative
ways of representation of the material behaviour
• Bauschinger effect
• For austenitic steels (and aluminium alloys) at higher plastic strains, Rasmussen (2003)
curve is more appropriate than Ramberg-Osgood
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Materials and geometry
• The rules in EN 1993-1-6 are not limited to steel shell structures
• The standard is valid for isotropic shells and shell segments made from any materials
that may be represented as ideal elastic-plastic
• For materials with no well defined yield point, 0.2% proof stress can be taken
• The material properties apply to temperatures not exceeding 150 ºC (otherwise see EN
13084-7, 2005)
• Where materials has a significant different stress strain curve, there are alternative
ways of representation of the material behaviour
• Bauschinger effect
• For austenitic steels (and aluminium alloys) at higher plastic strains, Rasmussen (2003)
curve is more appropriate than Ramberg-Osgood
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Geometrical tolerances and imperfections
• Relevant tolerances due to the requirements of serviceability:
• out-of-roundness (deviation from circularity)
• eccentricities (deviations from a continuous middle surface in the direction
normal to the shell along junctions of plates)
• local dimples (local normal deviations from the nominal middle surface)
• Other forms of geometric imperfections:
• deviations from nominal thickness
• lack of evenness of supports
• Material imperfections:
• residual stresses caused by rolling, pressing, welding, straightening etc.
• inhomogeneities and anisotropies
• Wear and corrosion
• Non-uniformities of loading
• Residual stresses
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Geometrical tolerances and imperfections
• Relevant tolerances due to the requirements of serviceability:
• out-of-roundness (deviation from circularity)
• eccentricities (deviations from a continuous middle surface in the direction
normal to the shell along junctions of plates)
• local dimples (local normal deviations from the nominal middle surface)
• Other forms of geometric imperfections:
• deviations from nominal thickness
• lack of evenness of supports
• Material imperfections:
• residual stresses caused by rolling, pressing, welding, straightening etc.
• inhomogeneities and anisotropies
• Wear and corrosion
• Non-uniformities of loading
• Residual stresses
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Ultimate limit states in steel shells
• LS1: Plastic limit
• Identifies the strength of the structure when stability plays no significant
role.
• Covers two conditions:
• tensile rupture or compressive yield through the full thickness
• development of a plastic collapse mechanism involving bending
• The plastic limit load is also relevant to a buckling strength assessment
Rpl - the plastic limit load
Rcr - the elastic critical load
• The plastic limit load does not represent the real strength (even for
stocky structures): strain hardening of material, stabilizing or
destabilizing effects due to change in geometry
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Ultimate limit states in steel shells
• LS1: Plastic limit
• Types of analysis:
• MNA: often underestimates the strength very considerably
• LS4: Fatigue:
• Repeated cycles of increasing and decreasing stress lead to the
development of a fatigue crack
• Methods of analysis:
• expressions in Annex C (using stress concentration factors)
• elastic analysis (LA or GNA), using stress concentration factors
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Ultimate limit states in steel shells
• LS3: Buckling:
• Caused by loss of stability under compressive membrane or shear
membrane stresses in the shell wall, leading to inability to sustain any
increase in the stress resultants, possibly causing catastrophic failure
• Three approaches used in the assessment of buckling resistance:
• GMNIA analysis
• MNA/LBA analysis
• Buckling stresses
• The strength under LS3 depends strongly on the quality of construction
• For this purpose, three fabrication quality classes are set out
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Design concepts for the limit states design of shells
• The limit state verification should be carried out using one of the following:
• Stress design:
• primary In general,
may beprimary
replacedstress statesresultants
by stress control LS1, whereas
where
• secondary secondary stress states affect LS2 and LS3 and local
appropriate
• Local stresses govern LS4.
• Modelling:
• Represention by its middle surface
• Nominal radius of curvature, imperfections neglected (excepting LS3)
• Eccentricities and steps if they induce significant effects
• Eccentricity at junctions between shell segments
• Stringers, corrugations, holes, depending on the conditions
• Boundary conditions
Boundary Simple Description Normal Vertical Meridional
condition term displacements displacements rotation
code
radially restrained meridionally
Introduction
BC1r to Design of Shell Structuresw = 0
Clamped restrained rotation restrained u=0 βφ = 0
Design ofradially
Steel Structures:
restrained meridionally Strength and Stability of Shells
• Stress resultants
BC1f restrained rotation free w=0 u=0 βφ ≠ 0
and stresses in shells
• Stresses:radially restrained meridionally free
BC2r w=0 u≠0 βφ = 0
• Thererotation restrained
are eight stress resultants in the shell
• However, the shear stresses xn, θn due to the transverse shear forces qxn, qθn are
radially restrained meridionally free
BC2f Pinned insignificant
rotation freeand they may usually be neglected w = 0 in designu ≠ 0 βφ ≠ 0
• For most design purposes, the evaluation of the limit states may be made using
BC3
Free edgeonlyradially
the six stress
free resultants
meridionally in the shell wall
free rotation n , n , n , m , m , mxθ βφ ≠ 0
w ≠ 0 x θ xθ ux≠ 0 θ
free
• Where the structure is axisymmetric and subject only to axisymmetric loading and
NOTE: The circumferentialsupport,
displacement
only v isnclosely linked to the displacement w normal to the surface so separate
x, nθ, mx and mθ need be used
boundary conditions are not identified in paragraph (3) for these two parameters.
• Modelling:
• Represention by its middle surface
• Nominal radius of curvature, imperfections neglected (excepting LS3)
• Eccentricities and steps if they induce significant effects
• Eccentricity at junctions between shell segments
• Stringers, corrugations, holes, depending on the conditions
• Boundary conditions
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Plastic limit state (LS1)
• The plastic reference resistance Rpl
• Where it is not possible to undertake a materially non-linear analysis (MNA),
the plastic reference resistance Rpl may be conservatively estimated from
linear shell analysis (LA) conducted using the design values of the applied
combination of actions using the following procedure.
t f y ,k
Rpl
nx2, Ed nx, Ed n , Ed n2, Ed 3nx2 , Ed
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Buckling limit state (LS3)
• To find out the design buckling resistance
• Defined as a load factor R applied to the design values of the combination of
actions for the relevant load case
• Different approaches have been proposed, difficult to generalise
• In EN 1993-1-6, a considerable effort to produce general procedures
applicable to all geometries, all loading conditions and all material conditions
f y,k
3 x ,Rcr
ov f ov , ov,0 , ov , ov , ov
ov is the overall elastic imperfection factor,
ov is the plastic range factor,
ov is the interaction exponent and
ov,0 is the squash limit relative slenderness
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Buckling design by global numerical MNA/LBA analysis
• It is recommended for many applications
• It has the same basis as the traditional stress design buckling approach
• All relevant combinations of actions causing compressive membrane stresses
or shear membrane stresses in the shell wall shall be taken into account
• It involves the following steps, see left hand side figure
ov f ov , ov,0 , ov , ov , ov
ov is the overall elastic imperfection factor,
ov is the plastic range factor,
ov is the interaction exponent and
ov,0 is the squash limit relative slenderness
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Design by global numerical GMNIA analysis
• Developed to exploit the full power of modern numerical
C1: The analysis
maximum load factor on the
load-deformation-curve (limit load);
• Application is more complex than for frame or platedC2:structures
The bifurcation load factor, where
• Several sequence of analysis: this occurs during the loading path
before reaching the limit point of the
• LA followed by a LBA to evaluate elastic critical load-deformation-curve
buckling resistance
C3: The largest tolerable deformation,
• GMNA to identify the elastic-plastic buckling resistance of the
where this occurs perfect
during the loading
structure path before reaching the bifurcation
load or the limit load
• GMNIA with different imperfection modes (the lowest value is selected)
C4: The load factor at which the
• Check the precision of the GMNIA by comparison with test
equivalent oratother
stress the most highly
relevant data stressed point on the shell surface
reaches the design value of the
• Methodology yield stress
• Action combinations causing compressive membrane stresses or shear
membrane stresses
• Rk should be found from the imperfect elastic-plastic critical buckling
resistance RGMNIA, adjusted by the calibration factor kGMNIA.
• The design buckling resistance Rd should then be found using the partial
factor γM1.
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Design by global numerical GMNIA analysis
• Developed to exploit the full power of modern numericalC1: The analysis
maximum load factor on the
load-deformation-curve (limit load);
• Application is more complex than for frame or platedC2:structures
The bifurcation load factor, where
• Several sequence of analysis: this occurs during the loading path
before reaching the limit point of the
• LA followed by a LBA to evaluate elastic critical load-deformation-curve
buckling resistance
C3: The largest tolerable deformation,
• GMNA to identify the elastic-plastic buckling resistance of the
where this occurs perfect
during the loading
structure path before reaching the bifurcation
load or the limit load
• GMNIA with different imperfection modes (the lowest value is selected)
C4: The load factor at which the
• Check the precision of the GMNIA by comparison with test
equivalent oratother
stress the most highly
relevant data stressed point on the shell surface
reaches the design value of the
• Methodology yield stress
• Action combinations causing compressive membrane stresses or shear
membrane stresses
• Rk should be found from the imperfect elastic-plastic critical buckling
resistance RGMNIA, adjusted by the calibration factor kGMNIA.
• The design buckling resistance Rd should
A conservative then be found
assessment using
of RGMNIA maythe partial
be obtained
factor γM1. using a GNIA analysis and C4 criterion to determine
the lowest load factor R
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• GMNIA analysis
• Allowances for imperfections:
• geometric imperfections: pre-deformations, out of-roundness, irregularities at
and near welds, thickness deviation, etc.
• material imperfections: residual stresses, inhomogeneities, anisotropies
• EN 1993-1-6 requires that imperfections are explicitly modelled numerically,
not just treated as small perturbations in geometry
• They are introduced by means of equivalent geometric imperfections in the
form of initial shape deviations perpendicular to the middle surface of the
perfect shell
• The form of the imperfections with the most unfavorable effect should be
considered (the most unfavorable effect on the buckling resistance RGMNIA
of the shell); if practicable, they must reflect the constructional detailing and
the boundary conditions
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• GMNIA analysis
• The analysis should be carried out for a sufficient number of different
imperfection patterns, and the worst case (lowest value of RGMNIA) should be
identified.
• The eigen-mode-affine pattern should be used (the critical buckling mode
associated with the elastic critical buckling resistance Rcr based on an LBA
analysis of the perfect shell)
• The amplitude of the imperfection form - dependent on the fabrication
tolerance quality class
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• Imperfections
• The maximum deviation of the geometry of the equivalent imperfection from
the perfect shape ∆w0,eff = max (∆w0,eff,1; ∆w0,eff,2), where:
• ni = 25 is a multiplier to achieve
an appropriate tolerance level
• t is the local shell wall thickness
• lg is all relevant gauge lengths
(see Dimple tolerances)
Introduction to Design of Shell Structures
Design of Steel Structures: Strength and Stability of Shells
• GMNIA validation
• For each calculated value of the buckling resistance RGMNIA, the ratio of the
imperfect to perfect resistance (RGMNIA / RGMNA) should be determined and
compared with values of found using the procedures of 8.5 and Annex D.
• The reliability of the numerically determined critical buckling resistance RGMNIA
should be checked by one of the following methods:
• by using the same program to calculate values RGMNIA, check for other shell buckling
cases for which characteristic buckling resistance values Rk,known,check are known;
• by comparison of calculated values (RGMNIA,check) against test results
(Rtest,known,check).
A = +/- sx − 1,50
P
sx x
fyt
• Combinations of meridional
(axial) compression,
circumferential (hoop)
compression and shear
Introduction to Design of Shell Structures
Finite Element Application
• Principles, simplified and advanced models, concentration of stresses,
stiffening
• Linear
• Multi-linear
• Continuous
• Powell,
• Ramberg-Osgood
Introduction to Design of Shell Structures
Finite Element Application
• Load cases
• The load cases shall be determined from the combination of operational
modes or other design situations, such as specific assembly, erection or
maintenance conditions, with the external conditions.
• The design load cases used (IEC 61400-1:2005) to verify the structural integrity
of a wind turbine shall be calculated by combining:
• normal design situations and appropriate normal or extreme external
conditions;
• fault design situations and appropriate external conditions;
• transportation, installation and maintenance design situations and
appropriate external conditions.
Introduction to Design of Shell Structures
Finite Element Application
• Load cases and design situations
• Load cases and assumptions for global analysis
• IEC 61400-1:2005
• Dead loads - Self-weight: tower head; tower body, installation etc.
• Wind action (EN1991-1-4)
• Seismic loads
• Temperature
• Ice
• Design situations
• Power production
• Power production plus occurrence of fault or loss of electrical network
connection
• Start up
• Normal shut down
• Emergency shut down
• Parked (standstill or idling)
• Parked plus fault conditions
• Transport, assembly, maintenance and repair
Introduction to Design of Shell Structures
Finite Element Application
• Wind action using different load situation:
• Global analysis to find-out the internal forces
• Load of wind on tower (hub)
• Simplified distribution (a2)
• Surface distribution (a1)
• Load from turbine’s machinery
• Concentrated force and moment
a2
Introduction to Design of Shell Structures
Finite Element Application
• Solver technique
• Pre-critical analysis and point results (Newton-Raphson)
• Post-critical analysis with deformation–to-failure (displacement
control, arc-length, modal analysis)
• The Static Riks step (based on arc-length solver) is able to find solution
during unstable loading response, when Static General step (based on N-R
solver) stops at limit load.
Plim
Introduction to Design of Shell Structures
Finite Element Application
• Results: Stability of steel shell structure General Report – Herbert Schmidt
(JCSR 55(2000) 159-181)
Eigenvalue results pure bending
Eigenvalue results axial compression different discretization levels
different discretization levels
• Verification procedures
• Analytical determination of moment capacity (LA) →MRd
• Characteristic buckling resistance (LBA + MNA)
• The plastic reference resistance → Rpl
• The elastic critical buckling resistance → Rcr
• The overall buckling reduction factor Rk ov Rpl
• Characteristic buckling resistance (GMNIA)
• Calibration factor Rk kGMNIA RGMNIA
Introduction to Design of Shell Structures
Finite Element Application and Case Study
• Study Case
• Expressions for buckling stress design
• Unstiffened cylindrical shells
• Critical meridional buckling stresses
l r l 20000
119
r t rt 2150 13,14
t 12 N
x , Rcr 0, 605 E cx 0, 605 2,1105 0,943 669
r 2150 mm 2
r 2150 0, 2 t
164 cx cx , N 1 1 2
t 13,14 cxb r
0, 2 12
cx , N 1 1 2 119 0,9434 0, 6
1 2150
Introduction to Design of Shell Structures
Finite Element Application and Case Study
• Study Case
• Expressions for buckling stress design (hand calculation)
• Unstiffened cylindrical shells
• Meridional buckling parameters
0, 62 0, 62
x 0, 25999
w
1,44 1,44
10,51
1 1,91 k 1 1,91
t 13,14
1 r 1 2150
wk t 13,14 10,51
Q t 16 13,14
x 0 0, 20
0, 60 p 0,806 x 0, 728 0,806
1, 0
0, 728 0, 2 N
1 0, 60 0, 477 x , Rk 0, 477 355 169
0,806 0, 2 mm 2
Introduction to Design of Shell Structures
Finite Element Application and Case Study
• Study Case
• Expressions for buckling stress design (hand calculation)
• Unstiffened cylindrical shells
• Meridional stresses
Fx 1770 103 N
x , Ed
N
10,92
2 rt 2 2150 12 mm 2
M 17702 106 N
x , Ed 2
M
102
r t 21502 12 mm2
Introduction to Design of Shell Structures
Finite Element Application and Case Study
• Study Case
• Expressions for buckling stress design (hand calculation)
• Unstiffened cylindrical shells
• Critical circumferential buckling stresses
l r l 20000
119
r t rt 2150 13,14
c t 1 12 N
, Rcr 0,92 E 0,92 2,1105 9, 06
r 119 2150 mm 2
r
1, 63 267,32
t
Introduction to Design of Shell Structures
Finite Element Application and Case Study
• Study Case
• Expressions for buckling stress design (hand calculation)
• Unstiffened cylindrical shells
• Circumferential buckling parameters
0,5 (Class C )
0 0, 40
0, 60 p 1,12 6, 26 1,12
1, 0
0,5 N
0, 0128 , Rk 0, 0128 355 4,53
2 6, 26
2
mm 2
r 2150 N
, Ed pn 1,5 1,102 103 0, 296
t 12 mm 2
Introduction to Design of Shell Structures
Finite Element Application and Case Study
• Study Case – Complete Model
• Imperfection amplitude
lgx 4 rt 4 4300 12 908,63 mm
w0,eff ,1 lg U n1
3.5
2.5
Load factor R
1.5
1
GMNA
0.5 MNA
GMNIA
Reactions Axial Shear Bending
0 at force force moment
0 1 2 3 4 5 6
[kN] [kN] [kNm]
Displacem ent [m ]
• GMNA Results
Bottom segment
2.5
2
Load factor
1.5
0.5
GMNA
0
0 50 100 150 200 250 300
Displacement [mm]
Introduction to Design of Shell Structures
Finite Element Application and Case Study
• Study Case – Refined door opening segment
• Numerical model
Section Axial force Shear force Bending moment
[kN] [kN] [kNm]
Upper section 1586 409 13055
Bottom section 2015 666 23897
• LA Rpl = (126/355)=2,82
• Transformation of typical wind pressure load distribution
c r 1 2150
kw 0, 46 1 0,1 0, 46 1 0,1 0,514
t 119 13,14
l r l 20000
119
r t rt 2150 13,14
N
qeq kw qmax 0,5411430 1,5 1102
m2
Introduction to Design of Shell Structures
Finite Element Application and Case Study
• Study Case – Refined door opening segment
• LBA Results • GMNIA results
• Imperfection afine first buckling mode
• Amplitude of imperfection 23mm and 17mm
corresponding to normal and high tolerance
Relevant segment
2
kGMNIA
1.8
1.6
1.4
1.2
Section Rcr
Load factor
0.4
Rotation [rad]
Normal (~23 mm) 1,72
DESIGN AND ANALYSIS OF SHELL S1RUCTURES
SOLID MECHANICS AND ITS APPLICATIONS
Volume 16
Series Editor:
G.M.L. GLADWELL
Solid Mechanics Division, Faculty of Engineering
University of Waterloo
Waterloo, Ontario, Canada N2L 3GJ
M.FARSHAD
EMPA,
Switzerland
~·
''
Springer-Science+Business Media, B.V.
Library of Congress Cataloging-in-Publication Data
Farshad. Mehd 1.
Design and analysis of shell structures l'fehdi Farshad.
p. cm. -- 1So11d mechanics and its applications ; 161
Inc 1udes index.
ISBN 978-90-481-4200-2 ISBN 978-94-017-1227-9 (eBook)
DOI 10.1007/978-94-017-1227-9
1. Shells IEngineer1ngl--Des1gn and construction. 2. Structural
analysis <Engineering> I. T1tle. II. Title: Shell structures.
III. Series.
TA660.S5F42 1992
624.1'7762--dc20 92-18175
ISBN 978-90-481-4200-2
Preface xi
Appendixes 375
Subject Index
Preface
Shell Structures present immense structural and architectural potential in various fields of
civil, mechanical, architectural, aeronautical, and marine engineering. Examples of shell
structures in civil and architectural engineering are: varieties of concrete shell roofs, liquid
retaining structures and water tanks, concrete silos, cooling towers, containment shells of
nuclear power plants, and concrete arch dams. In mechanical engineering, shell forms are
used in piping systems, curved panels, and in pressure vessel technology. Aircrafts,
spacecrafts, missiles, ships, and submarines are examples of shells used in aeronautical and
marine engineering. Shells are found in various biological forms such as the eye and the
skull, plants, and animal shapes. Thus, another application of shell engineering would be the
field of Biomechanics.
Shell structures developed since ancient times and now are being increasingly used in various
industries. Shells are used in the covering of large spans, liquid retaining installations, silos,
and containment structures. They are also used in the construction of light-weight vehicles,
pressure vessels, and space structures. Advent of such materials as ferro-cement, fiber-
reinforced concrete, composite materials, and reinforced polymers have all enhanced the
domain of shell technology. With the development of new prefab1ication schemes as well as
the need for recycling of materials, the potential of shell applications has further increased. In
addition to mechanical advantages, such as durability, high strength and stability, shell
structures enjoy the unique position of having extremely high aesthetic value in various
architectural designs.
In spite of all these features and potential applications, many engineers and architects are
unacquainted with shells as well as the aspects of shell behavior and design. The purpose of
this book is to familiarize the engineering and architectural students, as well as practicing
engineers and architects, with the behavior and design aspects of shell stmctures. The goal of
this book is to present three aspects: the physical behavior, the stmctural analysis, and the
xi
xii
design of shells in a simple, integrated, and yet concise fashion. Thus, the book contains three
major aspects of shell engineering. These are: (1) physical understanding of shell behavior,
(2) use of applied shell theories, (3) development of design methodologies together with shell
design examples.
To achieve these goals, simplified shell theories have been discussed in this book and have
been immediately applied to actual problems. In this sense, the book bridges the gap between
the elaborate theoretical treatments of shells, on the one hand, and, the practical aspects of the
analysis and design of shells, on the other hand. Being aware of a wide variety of existing
numerical routines for shells analysis, we have, nevertheless, made use of simple analytical
schemes of shell analysis so that the designer can understand the analysis procedure and to
perform parametric studies. The theoretical tools required for rational analysis of shells are
kept at a modest level so that engineering and architectural students, as well as practicing
engineers and architects, can grasp the fundamentals of shell behavior and, at the same time,
understand the related theory and be able to apply it to actual design problems. To achieve a
physical understanding of complex shell behavior, quantitative presentations are
supplemented by qualitative discussions so that the reader can grasp a "physical feeling" of
shell behavior. To make the book useful as a reference manual, a number of analysis and
detailed design examples are also worked out in various chapters.
The actual design of shells, involves the use of appropriate codes of practice. Thus, while
making use of some existing codes on shells, in order to provide a text that could be used in
various countries, we have attempted to present the designs apart from the existing codes. In
some cases, the common guidelines provided by several standards, including ACI, BS, DIN,
and IS, have been used.
This book can be used as a text book, and I or a reference book in undergraduate as well as
graduate university courses in the fields of civil, mechanical, architectural, aeronautical, and
materials engineering. It can also be used as a reference and design-analysis manual for the
practicing engineers and architects. To make the book useful to design engineers and
architects, the text is supplemented by a number of appendices containing tables of shell
analysis and design charts and tables. Metric system is used throughout this book.
The material of this book have been developed through many years of teaching at the
Universities of Shiraz and Tehran, University of Toronto, and the Swiss Federal Institute of
Technology (ETHZ) as well as through research and practical design experience by the
author. Thus, in the development of this text, various viewpoints and experiences have been
extremely constructive.
The author would like to thank the Swiss Federal Laboratories for Materials Testing and
Research (EMPA, Dtibendorf) and, in particular, Professor F. Eggimann and Professor U.
Meier and Hr. H. Fritz for supporting this project. Special thanks are due to Hr. P. Fltieler
who has given great encouragement and support in bringing this book to its present
publication. The author would also like to thank professor Gladwell for his useful comments
on the manuscript. This book is dedicated to my family (Gowhar, Anahita, and Mazda) who
have shown great patience during the long period of manuscript preparation.
M. Farshad
Switzerland
FORM AND CONSTRUCTION WITH HYPERBOLIC PARABOLOIDAL SHELLS
IN PLASTICS
UNIVERSITY OF SURREY
by
June 1970
ProQuest Number: 10803749
uest
ProQuest 10803749
ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, Ml 4 8 1 0 6 - 1346
S U M M A R Y
2
A C K N O W L E D G E M E N T S
•3
CONTENTS
List of Figures, 8
Foreward, 12
Introduction, 15
Reinforcement, 17
Resins, 20
Polyesters, 20
Fillers and Pigments, 20
Moulding Methods, 21
Contact Moulding, 21
Hand Lay Up, 21
Spray Up Moulding, 22
Introduction, 32
Definition of Surface, 44
Transformation of Surface, 44
Organisation of Surface, 47
Magnitude of Surface, 47
Model Studies of Surfaces, 51
Introduction, 117
Assembly of Model, l6 l
Concrete Foundation, l6 l
Steel Supports, l6 l
0. The Shell, 164
Steel Ring, 165
6
Loading Arrangement, l68
Test Programme, 173
Dead Load, 173
Snow Load, 174
Concentrated Loads, 175
Pressure and Suction due
to Wind, 175
Thermal Loads, Settlements
Loads and Dynamic Loads, 175
7
LIST OF FIGURES
Part One
8
Part Two
I
2.4 A flat double layer grid fromed from twenty HP units,
2.16-2.19 The macro, infra and micro modular units with reference
to a particular complex geometrical configuration,
9
Part Three
10
3-38 Apparent Young's modulus for the model material
expressed as a function of time,
<e
3-39 Computer output for the electrical resistance
strain gauge rosettes,
13
PART ONE GLASS FIBRE REINFORCED PLASTICS AS A STRUCTURAL MATERIAL
- their properties
Introduction
Figure 1.1 An approximate comparison of stiffness to weight ratio, and strength to weight ratio; for various materials.'
1.1
Tensile Compressive
Direction Shear
Moulding Method of stress strength Young's
% Glass Specific Tensile Poisson's Y oung%s Compressive Poisson's
and reinforcement:
by weight Gravity degrees perpendicular modulus Strength Ratio modulus Strength Ratio
contact moulding
p.s.i. p.s.i. p.s.i. p.s.i. p.s.i.
(kg/sq.cm.) (kg/sq.cm) (kg/sq.cm) (kg/sq.cm.)
2 02 .mat ^ 30.7 1.45 0,45,90 9 ,00 0 80 0,00 0 11,000 0 .3 2 900,000 15,000 0.42
(2 oz.per sq.ft.) (632 ) ( 56,240) (773) ( 63 ,270 ) ( 1,054)
Figure Mechanical properties are for short term loading tests. The values were determined experimentally at 0.1 hour, in room temperature, and
should serve for guidance only
1.2
16
C H A P T E R O N E MATERIALS AND MOULDING METHODS
R E I N F O R C E M E N T
Rovings
Woven Cloth
d
'19
1.3
and undirectional weave. Cloth is one of the most expensive
type of reinforcement, but where consistency of high performance
and structural efficiency in terms of strength to weight is
required the high cost can be justified.
RESI NS
Polyesters
Contact Moulding
1. Hand Lay Up
Directional Properties
Shop Practice
Factor of Safety
F.S. = 2 minimum
F.S. = 4 minimum
Cyclic loads
F.S. = 4 minimum
F.S. = 6 minimum
F.S. = 10 minimum
Behaviour of Laminates
26
Furthermore it is assumed that the material will return to its ,
original shape when the load is removed.
Thus:
(7 = E (e) or E = - . (l.l)
e
where (7-;='.stress, E = modulus of elasticity and e = strain.
Final Remarks
27
213,000 3,520
(15,000) (600]
7,100
(500)
142,000 5,680
(10 ,0 0 0 ) (400)
- Ib/sq.in(kg/sq.cm)
£ 4,260
H (300)
cr
in
cn
JSC
71,000 ass fib 0|i dig. 'T ' 2,840
polyester r<?sins
(5,000) .E (200)
cr
(/i
JQ
. 1.420
U) (100)
U)
st ress
8 5,2 00
(6,000)
G RP-roving
7 1 ,0 00
(5,000)
5 6 ,8 00
(4,000)
- l b/ sq. in( kg/ sq. cm)
4 2 ,6 0 0 GRP- cloth
(3,000)
u G R P-chopped strands
14,200
(1,000 )
stress
0 2 3 4
strain %
1.4
a
On the other hand the incorporation of time and temperature
in the stress-strain relationship will result in a very complex
analytical treatment. For practical design problems this
laborious mathematical treatment is seldom justifiable. In
practice the problem is simply solved by increasing the factor
of "ignorance" viz. factor of safety.
REFERENCES
31
Introduction
3^
2.1
2.2
'35
in Dreux for testing purposes, a two sided cantilever structure,
made up of four HP segments.
(During the war only the Czech engineer Konrad Hruban built
"umbrellas" with HP shells for an industrial roof at Nove Mesto in 1943).
38
2.3
2.4
39
2.5
2.6
40
An assembly of 20 such HP units is shown in Figure 2.4 all the points
marked B lie on a flat plane. This system was applied by Chaperot
for a market cover in Lezoux and a garage in d'Arcueil France.
(83,958 ft.2).
4l
C H A P T E R T W O RESEARCH INTO THE MORPHOLOGY OF
HYPERBOLIC PARABOLOIDS
The first topic - Form - deals with the "idea", i.e. the geometrical
morphological aspects of HP surfaces.
42
It is possible to deal here only with general systems and
concepts of structural forms. In the final design process, Form and
Construction cannot be treated independently of each other. It
is the task of the creative designer to synthesize these two elements
to form one inseparable whole.
FORM IN HYPERBOLIC PARABOLOIDS
Definition of Surface
Transformation of Surface
2. By varying the angle between the director plane one can get a HP
surface with two different principal parabolas (Figure 2.9)•
45
2 .1 1
*46
Organisation of Surfaces
Magnitude of Surface
p o i nt of
the un i t
b i l a t e r a l + r o t a t i o n a l sy m m e t r y b i l a t e r a l * t r a n s l a t o r y s y m m et r y d ilatatio n
!
i a x i s of r o t a t i o n of t r a n s l a t i o n
2.12
subt ra c tio n
2 .1 3 '48
or gani sat i on of h y p e r b o l i c - p c t r a b o l o i d s ( h y p a r s )
t h e " r e p e t i t i v e e l e me nt a r y " unit ( wi t h a h i gher de gr ee of s y m m e t r y )
2.14
linear arrangement
s patial a rra n g e m e n t
49
m a c r o - m o d u la r u n it
a x is o f r e v o l u t io n
parall e]s_____
g e n e r a t r ix =
m e r id ia n
2.l6 2.17
m i c r o - m o d u la r u n it
in fra -m o d u ia r unit
axis of re v o lu tio n
meridian
2.18 2.19
'50
3» The micro modular unit-the 'structural pattern1 of which the
repetitive element is composed. For example, the micro modular unit
in Figure 2.16 is a honeycomb sandwich construction as shown in
Figure 2.19-
52
A-A B-B
2.20
‘53
A-A 8-B
2.21
2.22
'55
2.23
56
2.24
31
A
2.25
58
.
A-A B-B
2.26
'59
2.27
'6o
A-A B -B
2.28
6l
I
A- -A
q,
I
A-A 8-8
2.29
62
B
A-A B-B
2.30
63
2.31
6k
A-A B-B
2.32
■65
66
A-A B-B
2.34
67
B
2.35
'68
2.36
69
2.37
'70
2.38
71
2.39
72
A
A-A B-B
2.40
73
2.41
74
2.42
75
2.43
76
ii-./V:' it- . . *«
‘77
B A 0 c B
#J>.; k?p
-..' ' ' <^ y - ’X
,.Vv
-
■^'vv »
> 7 f
§
2.45
•78
iI l f i n n K r ^
wsh'ifW
A
b'
'80
6'
b'
C
a
-8i
b
C A
2.49
82
c
J k
2.50
83
d'
C1 A'
2.51
84
D
2.52
85
b‘
A'
2.53
66
;MS
> //» ■
2.54
87
c
b1
d’
C A C B A D C
88
c
d‘
2.56
89
V
m m ’v
2.57
90
b'
A’ A'
b'
A B A D C B
A"
91
'92
-
V *
§3
b'.d
A.C o’
2.61
9^
c
b.c
2.62
95
2.63
96
'98
2.66
:?-r
iXh.:
•99
3.cr
^O q
c
2.68
101
CONSTRUCTION WITH HYPERBOLIC PARABOLOIDS
Construction Principles
1 is always greater than b and h (Figures 2.71; 2.72; 2.73; 2.74 and
2.75)• In a surface construction 1 and b exceed h (Figures 2 .7 6 ;
2.77 and 2.78).
102
anticlastic shell (Figure 2.76);
103
2.69 2.70
2.71 2.72
2.73 2.74
104
2.75 2.76
2.77 2.78
2.79 2.80
'105
Possible Application Of Hyperbolic Paraboloid Shells
Figure 2.82 Undulated vault roof system with slanted arched ribs
having an undulated edge at its perimeter and supported on exterior
inclined buttresses. Macro modular unit - a hyperbolic vault composed
of 4+5 principal HP segments. Infra modular unit - a compound element
106
composed of three different segments. 1. One principal HP segment
has an "orange peel" like plan and two slanted hyperbolic edges,
(see also Figure 2.30). 2. The other principal HP segment has a
"bottle neck" like plan of which the two opposite long edges are
slanted and hyperbolic. The two short edges are also slanted and
hyperbolic. 3« The auxiliary HP segment is similar to the one
in Figure 2.81.
Figure 2.86 Scalloped dome roof system with its crown pulled
down having an undulated edge at its perimeter, and supported on
inclined exterior butresses. Macro modular unit - partly synclastic
2.86
109
and partly anticlastic paraboloid of revolution composed of 12
principal segments, (see also Figure 2.43). Infra modular unit - a
compound unit composed of two different segments. 1. The principal
HP segment is over an isosceles triangular plan of which two
adjacent edges arch parabolically upwards. The third edge is slanted
and hyperbolic. 2. The auxiliary HP segment is similar to the one
in Figure 2.81.
110
2.87
2.89
ill
tinit - a synclastic paraboloid of revolution composed of 12
principal HP segments (see also Figure 2.43) intersected by a
•torus' (with a parabolic cross-section) composed of 12 principal
HP segments. Infra modular unit - a compound element composed of
two principal and two auxiliary segments. 1. One principal HP
segment that makes up the dome is identical to the one in
Figure 2.85® 2. The other principal HP segment, that makes up the
'torus', is over a trapezoid plan of which two opposite edges are
parabolic and arch upwards, the other two opposite edges are slanted
and hyperbolic (see also Figure 2.44). 3 & 4. One auxiliary
segment between dome and 'torus' is identical to the one in
Figure 2.81 and the other being similar to it.
Final Remarks
An attempt was made to illustrate with the many drawings and models
that HP surfaces present an ideal media, for the designer who welcomes
and searches for a geometric discipline.
2.91
113-
REFERENCES
On Structural Forms
7. Heger F.J. Chambers R.E. and Dietz A.G.H., "On the Use of Plastics
and Other Composite Materials for Shell Roof Structures", Proceedings
World Conference on Shell Structures Oct. 1962 San Francisco, California.
U-
13. Kuhn Walter, "Plastische und Raumliche Symmetrien", iysstellung
im Kunstgewerbe H.useum, Zurich, 1 9 6 8 .
17« Hart Franz, "Kunst and Technik der Wdlbung", Verlag Georg D.W.
Callwey, Miinchen, 1 9 6 3 .
18. Torroja E . , "Logik der Form", Verlag Georg D.W. Callwey, Munchen, 1961.
115
P A R T T H R E E DESIGN AND CONSTRUCTION OF A SCALLOPED DOME IN
GLASS FIBRE REINFORCED PLASTICS
l l 6
Introduction
The designer must first speculate intuitively and then test the
hypothesis by an experiment.
A e $ c r i'hc
An attempt is made here to introduce a method of research that
will assist in the achievement of optimum design, without the
disadvantage of facing the complex and lengthy processes of structural
analysis with the highly mathematical treatment usually associated
with such work.
117
SECTION ONE ARCHITECTURAL DESIGN
3.1
’
119
The rigidity of the thin free edges between the inclined supports
can be substantially increased by the inter-penetration of an
auxiliary shell segments which form canopies as shown in Figure 3-3*
This modification of the principal triangular segment is achieved by
cutting away a lower slice of the principal segment by means of an
inclined plane passing through the points ACF as illustrated in
Figure 3»^- This lower slice is then rotated through l80° and
connected again to the principal segment. A gapless fit is possible
after rotation because the surface has an axis of symmetry in its
axis of rotation.
120
C H A P T E R O N E GEOMETRY OF THE SHELL SEGMENT
Definition of Shape
f
^ r- „
0.5 ? and —2 = 1.
3.2 3.3
JC"
3.4
122
express all the dimensions in terms of r:
r cos
a
2 ’
r sm a
2 *
a= ^ = 13.768° or | = 6.884° ,
hence:
tan ^ = 0 .1 1 9 8 6 ,
r cos - = 0.99279 r 1
r sin - = O . H 986 r
123
Arbitrary angle between them. Since f / f ; / 90°--
1 eL
Thus an oblique HP surface is obtained.
z = h 4 - f2 4 • (3 .D
a b
By inspecting Figure 3-^t it'can be seen that the directrix plane must
lie in the GOZ plane. The location of point G on the principal
parabola is not known. The straight line generator 0G is at right
angles to the Z axis, and thusat the point G, Z = 0.
2 2
XG y
° = fl ~2 “ f2 "~2 ’
a b
2
xr
0 = fl "a - f2 *
a
Solving for x
G
£2
fl
124
Substituting numerical values
*Q = 0.99279 r K § : “ 9 ?6 :9 9 2 7 9 " V) ’
0.488 r
+
tan w
- _ b
- _ 0-11986
0 ^488 _ .
0.2456 , ,
G
| - 1 3 .8 °
2 2 2 2
0 .5 *_ . 3~ 3_
a b 0 .9 9 2 9 r 0 .1 1 9 8 6 r ,
£
x
------- 5 x - £ cos —
a ; y = x tan
a
—
* cos 0_ b 2 ’ J 2
2
125
E
2
z — cos2 § [o.5 0 3 6 - 8.3434 (0 . 1 1986 2)] ,
2
I
= 0.377 ■ — .
Note that the space curve CO is a parabola with a lower rise than
the principal parabola (l) (Figure 3-4 ).
5. Check:
z = 0.377 r .
By inspection
Thus:
z = (f^ - f ) - (a - x) tan p . (3 .1 .2 )
126
Substituting equation (3.1.2) in (3-1) gives the general formula
for the curve AFC in its horizontal projection on the XY plane,
■which reads:
0 .3 7 6 5 r - (0 .9 9 2 8 r - x) tan p = - (0 .5 0 3 6 x 2 - 8.3431 y2 ) .
0 .3 7 6 5 r - 0 .9 9 2 8 r tan p + x tan p = ^ 0 .5 3 6 x 2 ,
Solving for x:
x2 - 0 .7 2 2 8 r x = 0.030024 r 2 ,
= 0.3614 r + \j 0 .1 3 0 6 + 0.030024 r ,
= (0.3614 ± 0.4007) r ,
= 0 .7 6 2 2 r.
127
Substituting for x_ in the general equation (3.1.2):
F
z = 0 .3 7 6 5 r - [ 0 .9 9 2 8 (0.3640)] r + 0.364 x ,
= 0.0151 r + 0.364 x ,
= 0.2925 r .
= 0.5036 m 2 r2 - 8.341 y 2
Solving for y ;
2 1
y 0 .5 0 3 6 m 2 - 0.364 m - [ 0 .3 7 6 5 - 0 .9 9 2 8 (0.364)]
8 .3 4 3 1
2
y 0.364 m - 0 .0 1 5 1 2 )
With this equation the projection of the curve AFC on the XY plane can
be plotted as shown in Figure 3-6 .
Having found the XY co-ordinates for the space curve AFC its z
ordinates can be calculated by using equation (3-1.2) .
Substituting numerical values:
z = O .3 7 6 5 r - [0 .9 9 2 8 (0.3640)] r + m 0.3640 .
With this equation the curve AFC projected on the ZY plane can be
plotted as shown in Figure 3*6 . •
z = z ,
F ’
X X
(3-1.4)
y y
Thus:
z z + z.
F
X X + X.
F
z + 0 .2 9 2 6 r = ~ [x + 0 .7 6 2 2 r2 ( 0.5036) - 8.341 y2 ] ,
129
- _ 0 .5 0 3 6 x2 + 0 .5 0 3 6 2 (0 .7 6 2 2 x r) + 0 .7 6 2 2 2 (0 .5 0 3 6 r2 )
z =
8.4315' - 0 .2 9 2 6 r ,
0 .5 0 3 6 x 2 « 8.341 y 2
Z = + 0.7677 X - --------------------- (3 .1 .5 )
z = x sin p + z p cos p (3 .1 .6 )
y = y
Hence:
X cos p + z sm
(3 .1.6 .1)
z = -x s in
m p z cos
P
Since: Sin p = Sin 20 = 0.342 ,
Cos p = Cos 20° = 0.940 ,
130
which can also be written as:
-2
0.342 x + 0.940 z = 0.4430 ^-6 - ------- 8--- ^-- 3
p P r r
-2
z
^58l_JS + 0.7216 5
P "
!
8.3431 y2
- 0.2626 Z fi .
P
The equation of the hyperbolic paraboloid surface in
X Y Z co-ordinates is:
P P P
r -2 -2 -2
z _ x y
- 0.0589 ------ - 0.4450 P- + 3 .3 4 3 1 - P - +
r r r
ft z
0 .3 2 3 8 P + 1 .2 0 2 6 Z p - 0 .3 7 9 6 x = 0 . (3.1.7)
(3 .1 .8 )
p ’
X z
p p
- 0 .3 2 3 8 -- ----- - 1 .2 0 2 6 z + 0 .3 7 9 6 x = 0 (3.1.9)
P r
131
10. Location of Point B (reversal)
(1) Translation:
B = ° ’
x = 0 .2 3 0 6 r .
B
zB = [0 .5 0 3 6 (0 .2 3 0 6 2) + 0.7677 (0 .2 3 0 6 )] r,
= (0 .0 2 6 8 + 0 . 1 7 7 0 ) r ,
0 .2 0 3 8 r .
Using equation(3.1.6.l)
£ = 0 .2 3 0 6 (0.946) r + 0 .2 0 3 8 (0.342) r ,
pB
= (0 .2 1 6 8 + 0 .0 7 0 0 ) r ,
= 0 .2 8 6 8 r .
z = - 0 .2 3 0 6 (0.342) r + 0 .2 0 3 8 (0.940) r ,
PB
= (-O.O789 + 0 .1 9 1 6 ) r ,
= 0.1127 r .
132
(3) Rotation 180°:
x = 0.2865 r ,
z = - 0.1127 r .
P
Substituting these values in equation (3.1.6):
xD = [ 0 .2 8 6 5 (0.940) + 0 . 1 1 2 7 (0.342) 1 r ,
B ■*
= (0.2693 + 0.03854) r ,
0.30784 r .
z = [0 .2 8 6 5 (0.342)-(0.1127) 0.940 ] r ,
B 1 J
= ( 0.09798 - 0.1059 ) r ,
= - 0 .0 0 7 9 6 r .
(1 ) Translation:
z = [ 0 .5 0 3 6 (0 .9 9 2 8 2 ) r2 - 8.3431 y2 ] ,
2
= - 8.341 ^ + 0.4964 r .
(2) Rotation 20°
Since z = z + 0.2926 r ?
y = y •
2
z 8.341 - + 0 .2 0 3 8 r .
r
z„ = - 0.3640 X - 8 .8 7 5 6 — P + 0.2168 r .
P P
The equation of the space curve in X Y Z co-ordinates according
to equation (3.1.8) becomes:
P P P
2,
z = + 0.364 x„ + 8 .8 7 5 6 — P- 0 .2 1 6 8 r. (3 . 1 .1 0 )
P P
12. Check:
Substituting in equation (3.1.10) the ordinate of point B (rev.) which
is known. The equation of the space curve AB (rev) C in X Y Z
co-ordinates using equation (3.1.10) and (3.1.6.1) gives:
-2
+ 8 .8 7 5 6 - - 0 .2 1 6 8 r ,
134
■\ z [ 0.940 - 0.364 (0.342)] = X [ 0 .3 4 2 + 0.364 (0.940)] .+
2
+ 8 .8 7 5 6 ~ - 0 .2 1 6 8 r ,
2
z (0.8155) = x (0.684) + 8 .8 7 5 6 - - 0 .2 1 6 8 r ,
r
2
z ■= x 0.83875+ 1 0 . 8 8 3 6 ^ - 0 .2 6 5 8 r . (3.1.11)
If x 0.30784 r ,
z = 0 .2 5 8 2 r - 0 .2 6 5 8 r = - 0 .0 0 7 6 r ,
z =- 0 .0 0 7 9 6 r .
Figure 3.5.(5) •
Since
\ = + 0.0839 r ,
x^ = 0.2306 r .
Hence
z = - 1 .1 9 1 8 x + 1 .1 9 1 8 (0.2306)r + 0.0839 r. (3 . 1 . 1 3 )
135
“2
O .8 3 8 7 5 x + IO .8 8 3 6 ^ - O .2 6 5 8 r =
- 1.1918 x + 1 .1 9 1 8 (0 .2 3 0 6 ) r + 0 .0 8 3 9 r ,
136
0 .9 9 2 8 r
-ox
0 .2 9 2 6 r
0.3 7 7 Or
sr—
0.119 9 r
0 .2 3 0 6 r
■0.0839r
0.11 2 7 r
0 .2 0 3 8 r
0.119 9r
0.11 2 7r
_ L1
0 .2 3 0 6 r
------------- 0 .3 0 7 8 r
(rev.)
X
0.1199 r
Y=50
A,C
->X
(rev.)
t* Xi = 0. 230G r — H
0.119 9 r
3.5
137
------- a= 0.99 28r
0*0167 r-
0*0000 -
-X
00000-
-OOMOr- - f - d=r/io —|—
0*3765 r
0*3660 r-
-f- b — f—
3.6
138
S E C T I O N T W O ENGINEERING ANALYSIS
In our case this simply means, building first a model with a shell
thickness as thin as possible and loading it. From the observed and
measured deformations one can determine the critical buckling mode.
An unstable shell might have to be stiffended or by changing the
curvature of the shell or increasing its thickness.
(a) It brings shell analysis within the reach of those who are
unfamiliar with the techniques of advanced mathematics.
‘1 3 9
Three - Pinned Arch Analysis
1. Vertical Loading
2 2
Only the live load (lb/ft or kg/m ) of plan area is considered, its
intensity is denoted by (q). The self weight of the segments, being
small, are included in q. With a roof load of q lb/ft , the load per
unit length of arch varies linearly from q^ at the outer ring of
radius r, to zero at the centre (Figure 3-7») •
2 TX r
qr = q ,
Similarly
_ 2 Tl g
96 = — 9
2
RA - RB = -^r 9
M
2 2|
Tl r q3 r
o n n n
3
11 r q
n
3
II r q
3n
140
(-IE
N,
3-7
r=975em
■4o—
A= 12 0 c m 2
A = 132cm 2 A= 90cm 2
I 2 6 .5 1 = 5271cm 4 1= 374cm 4
48.5 I t = 1.8 cm
n.x.
I T
20 t= 0 .6 c m
_L 95cm
A-A
3-i
The bending moment at any point with abscissa ^ is:
_ j A ( g) _ jt A ( a r +v . Jt£a ( h ) ,
P n n 3 n 3
•j 'j
TCr q _ Ttl q
n3 n3 ’
( r3" S3 ) ,
_ 3 e3
3n ( i T A r3- > •
sS P - 2
- +
2 n
gq (I )
K 2 ' '
K2
r
+ TX 5 q
n
3. Arch Stresses
Horizontal thrust:
The thrust H is obtained by equating to zero the moments about 0 , i.e,
IM = 0 ,
o 1
M H (f ) .
o o
142
Hence:
M
o
H
f ’
-r r2 q
71
3 (0.3765) n
2
TCr q
0.8853
n
M M k - H (f - z) ,
f (b.b.r
M (1 + ) - 0.8853 - rn ^ 0.3765 r (1 - ) *
Tt r3 q
(1 ) - 0.8853 (0.3756) (1 - — - )
n
Tlr"1 q 1
n 3
TC r3 q
M (3 -2 .1)
n M
l zi 3
4. Axial Forces
N (+) = compression,
N (-) = tension,
where:
d z
tan ip d = 2 (0.3765 2 ) = 0.753 2 ,
cos ip
I1 + tan m w
1 + 0 .3 6 7 £
2 ’
tan tp 0.753 r
sin ip =
1 + tan *p
il + 0.567 ~
r
so that:
2
11 £ q 0.753 r
N = +
n
1 + 0.567
U r 2 c
+ 0.8853
n
1 + 0.567 ~
r
2
TC r q 0.8853 - 0.7530 r
n
1 + 0.567
144
which can also be written as:
2
ft r q
N
n
(3 -2 .2)
N
5. Shearing Forces
S s= S Cos q) - H sin q) ,
» £ (b.b.)
so that
1 + 0.567
Tl r2 q 0.7530 r
- O .8 8 5 3
n
1 + 0.567
-
+ —
s
w 2
Tl r g
n
1 + O .5 6 7
Tl r2 q
(3-2.3)
n S ’
A 1 .0 0 - 1 .0 0 0 0 .0 0 0 0 0 1 .3 0 8 7 - 0 .2 6 6 2 5
o
o
i
l 1 -0 .9 0 0 .8 l - 0 .7 2 9 0 .0 2 6 9 9 1 .1 8 7 0 - 0 .1 7 3 9
2* - 0 .8 0 0 .6 4 - 0 .5 1 2 0 .0 4 2 6 6 1 .0 8 8 8 - 0 .0 9 1 4 1
3' - 0 .7 0 0 .4 9 - 0 .3 4 3 0 .0 4 8 9 9 1 .0 1 2 0 - 0 .0 2 0 7
4* - 0 .6 0 0 .3 6 - 0 .2 1 6 0 .0 4 7 9 9 0 .9 5 5 1 + 0 .0 3 6 4 2
5' -0 .5 0 0 .2 5 - 0 .1 2 5 0 .0 4 1 6 6 0 .9 1 6 9 + 0 .0 7 7 9 5
6' -0 .4 0 0 .1 6 -0 .0 6 4 0 .0 3 1 9 9 0 .8 9 4 1 5 + 0 .1 0 2 1 5
7' - 0 .3 0 0 .0 9 - 0 .0 2 7 0 .0 2 0 9 9 0 .8 6 1 6 + 0 .1 0 7 2 7
DO
«
0 .0 4 - 0 .0 0 8 0 .0 1 0 6 6 0 .8 7 2 0 + 0 .0 9 2 3 0
o
8'
O
•
9' ‘ 0 .0 1 -0 .0 0 1 + 0 .0 5 6 5 3
o
o
0 .0 0 2 9 9 0 .8 8 3 5
1
0 o .o o 0 .0 0 0 .0 0 0 0 .0 0 0 0 0 0 .8 8 5 3 0 .0 0 0 0 0
1 + 0 .1 0 0 .0 1 0 .0 0 1 0 .0 0 2 9 9 0 .8 8 3 5 - 0 .0 5 6 5 3
2 + 0 .2 0 0 .0 4 0 .0 0 8 0 .0 1 0 6 6 0 .8 7 2 0 * - 0 .0 9 2 3 0
3 +0 .3 0 0 .0 9 0 .0 2 7 0 .0 2 0 9 9 0 .8 6 1 6 - 0 .1 0 7 2 7
4 + 0 .4 0 0 .1 6 0 .0 6 4 0 .0 3 1 9 9 0 .8 9 4 1 5 - 0 .1 0 2 1 5
5 +0 .5 0 0 .2 5 0 .1 2 5 0 .0 4 1 6 6 0 .9 1 6 9 - 0 .0 7 7 9 5
6 +0 .6 0 0 .3 6 0 .2 1 6 0 .0 4 7 9 9 0 .9 5 5 1 - 0 .0 3 6 4 2
7 +0 .7 0 0 .4 9 0 .3 4 3 0 .0 4 8 9 9 1 .0 1 2 0 + 0 .0 2 0 7
8 +0 .8 0 ' 0 .6 4 0 .5 1 2 0 .0 4 2 6 6 1 .0 8 8 8 + 0 .0 9 1 4 1
9- + 0 .9 0 0 .8 1 0 .7 2 9 0 .0 2 6 9 9 1 .1 8 7 0 + 0 .1 7 3 9
B + 1 .0 0 1 .0 0 1 .0 0 0 1 .0 0 0 0 0 1 .3 0 8 7 + 0 .2 6 6 2 5
3-9
'1 4 6
Approximate Design of Cross-Section Required
The arch segment, even under its own dead weight, will produce
a state of stress defined as bending. That is to say, the bending
stresses vary linearly from a maximum compressive value at the top
fibre to a maximum tensile value at the bottom fibre. The axis
passing through the points where there is no bending stress is
called the neutral axis.
1 I
- M h2
2 “ ~T~
M (A - A) = ^ ,'53° kg/ cm #
_ axial thrust_____________
(section) cross-section of area of segment .
hence:
149
C H A P T E R T W O EXPERIMENTAL MODEL ANALYSIS
150
Relationships Between Model and Prototype Quantities
1) 0* = stress in shell
2) a,b = linear dimensions of shell
5) V = Poisson's ratio
6) y any displacement
7) p = a concentrated load
8) q = a uniformly distributed load
Deflection: y = y S, (3*4.1)
;p Jm ’
E
Forces: F = S ■— F . (3*4.2)
p E m
m
1 E
Moments: M = S M (3*4.3)
p E m
E
Stresses: (j* = rr" . (3.4.4)
P m
™
a a
Dimensions: , (3.4.5)
D D
m p
am = ap , (3.4.6)
t t
m p
E
<1m “ q
p
—
E
(3 -4 .7 )
P
152
Since the prototype scalloped dome had a clear span of
20 m. (6 5 .6 ft.) a scale factor of 5 was established, thus S =
Dimension of Model
The length 200 cm. (6 .5 6 ft.) and width 49 cm. (1.6 ft.) of
the HP triangular segment was determined by the scale factor.
153
CHAPTER THREE CONSTRUCTION OF MODEL
Mould Construction
The master (mould) for the production of large size GRP shells,
by contact moulding (hand lay up technique) is generally made with
plaster applied over a timber framework. This technique is far from
satisfactory, as the construction of the mould is complicated and
expensive. To replace this inadequate method the author has developed
a special technique known as cocooning which provides the profile
required for highly warped HP surfaces at low cost. In this technique
a sprayable quick-drying vinyl latex is applied to form initially a
"cobweb" which can bridge sizeable gaps up to 60 cm. (l.9 7 ft) wide
over any reticulated network as shown in Figure 3»l6. After completing
the spraying process a continuous seamless leather-like skin exists
over the sprayed area. During the drying process, the latex of
the "cobweb" shrinks, forming a prestressed membrane with "minimum”
surface (Figure 3'17)» GRP laiminates can be produced from the skin
covering the area without any difficulty to form a single element or,
alternatively,a more robust GRP mould can be produced first.
1. The Box
The three sides and base illustrated in Figure 3-10 are made of
4 " plywood sheets. (The "box" for the elements of the prototype
structure, will not be made in plywood as a lattice timber framework
construction is more practical and economical).
The calculations set out in Chapter One were used to determine the
exact contour lines for the panels. Figure 3»6 shows the dimensionless
values for the various points on the curves. To obtain a numerical
value the values merely had to be multipled by (r) the radius of the
scalloped dome model - 200 cm. (6 .5 6 ft).
15^
3.10 3.11
3.12
Figure 3.10 shows the panels before assembly. Note the
parallel line grid on the triangular base plate. These lines are
the horizontal projections of the HP straight line generators;
and will be referred to again later.
Nylon cords were found most suitable. Figure 3-13 shows the
prestressing device - a lever - and the fixing scheme for the cord.
156
The fastening device consisted of a l/4 in. steel bolt & nut
and a fibre washer. The bolt had a fine hole just below its head.
While prestressing, the cord could pass freely through the bolt.
When the desired prestressing level was reached the nut was
tightened; friction between panel and washer held the cord in its
prestressed position.
3.16
158
Two identical formworks were built in order to speed up production.
Standard compressed-air spraying equipment was used and it
consisted basically of a spray gun, a pressure feed floor container
and an air compressor.
Moulding Technique
The vinyl-latex skin of the mould was ready for use after it had
dried thoroughly for about 48 hours. The first operation in producing
a GRP laminate was, of course, to wax polish and apply a release agent
on the mould surface. A floor wax and a release agent were used.
Two laminates were made each morning and left to cure at room
temperature. The segments were carefully removed from the mould on
the following day (Figure 3»20), and washed and cleaned with soap and
water. A portable diamond cutter, driven by compressed air, was used
to trim the edges of the flanges to a height of 3/4 in. Next small
holes were drilled, with a template, into all the flanges.
159
3.19 3.20
160
Finally, the auxiliary shell segment, that formed the projected
canopy was cut out from the principal HP segment, turned over l80°
and laminated to the same segment, A lr in. wide strip of chopped
strand mat impregnated with polyester resin formed a continuous joint
between the two shell segments.
Assembly of Model
1. Concrete Foundation
2. Steel Supports
l6l
------------------ i« i ------------------
-t- SIO t-
4
I
3-21
t / f > b e ll
3-22
162
3.23 3.24
3-25 3.26
3.27 3.28
163
(3) prevent high stress concentrations in the thin shell, at the
points of support; and (4) express the lightness of the shell structure
ideally it should appear to "soar11 upwards.
3. The Shell
After the GRP strips had cured on the falsework, the segments were
again turned over, bolts taken out, and the projected vertical flanges'
were cut off with the diamond cutter (Figure 3-28). (For experimental
reasons it was decided to cut off the flanges, despite the obvious
fact that their omission would reduce the stiffness of the dome).
For easy handling segments were first jointed in pairs then in
fours and finally in two groups of eight plus one group of ten. The
three completed sets were put on their steel supports, and the
remaining joints were laminated in situ from inside.
4. Steel Ring
165
3-30
166
'167
Loading Arrangement
si/
3.34
169
3.35
3.36
170
171
The loading platform was entirely made of timber. It consisted
basically of three triangular shaped braced trusses, made of 2 in.
by 4 in. joists, fastened together at their ends with \ in. plywood
gusset plates. A hydraulic jack was located at each of the
three corners.
Test Programme
1. Dead Load
The dead weight of one completed segment of the testing model is;
2
2 kg. and its projected plan area is 0.48 m . This corresponds to
The dead weight of prototype and model are related by the scale
3
factor, i.e: ¥ = ¥ (S ), (see page )-where the dead weiqht of
p m 7
¥m = 2 kg (2 6 ) = 32 kg.
The dead weight of the shell varies per unit area of sloping
shell but in order to simplify the calculation it was assumed capable
of representation as a uniform load per unit area of horizontal shell
projection. In a GRP shell where the dead weight per unit area is only
about l/ 6 of the live load, this simplication is not likely to cause
any appreciable error in the test experiments.
Since model and prototype are made of the same material this simplifies to
173
In order to simulate the stress condition in the prototype, the
dead weight of the model must first be subtracted from the dead
weight of the prototype.
Having 1,248 loading points on the dome (not considering the 104 points
on the small cantilevering portion of the canopies), the resulting
load on each point is:
211 kg _ _
17548 " °-169 k9‘
Since each loading hook is connected to four points, the load per
hook on the model which will simulate the dead weight of the prototype
will have to be:
2. Snow Load
load intensity of 110 kg/sq.m. Since the area of the model was .
1 2 .5 7 sq.m., the total load on the model was 1 ,3^3 kg., distributed
174
on 1,248 points, gives 1 .1 0 8 kg. per point, or 4.432 kg. per
loading hook.
Two loading cases were investigated: (l) A full span load i.e.
a uniform load over all the 26 segments; and (2) A half span load,
i.e. a uniform load over 13 segments.
3. Concentrated Loads
1. Measurements of Deflections
2. Measurements of Strains
176
backing of the gauges was bonded to the GRP surface with MDurofixn ;
the bond proved satisfactory throughout all the tests. Figure 3*34
shows the position of the 64 gauges on the shell segment. The
symmetry in the structure made the bonding of gauges elsewhere un
necessary. Gauges were only mounted on the outer surface of the
shell, hence only membrane strains could be measured. The shell
segment with the necessary ndummy gauges" is shown in Figure 3-43-
The gauges were connected to an electronic data logger, (Figure 3-37)»
which punched the strain readings on paper tape. The principal strains,
their directions, as well as the principal stresses were computed by
an ICL 1905F computer. A typical computer-printed output is shown in
Figure 3-39*
177
Mechanical Properties of the Model Material
1. Young's Modulus
For long term loading, the values for E will change with time
and temperature (see also page 25 )• It is practically impossible
to determine the time-temperature dependent E value (apparent modulus)
for a GRP laminate analytically. Figure 3-42 shows how the values
for the apparent modulus were evaluated. A strain gauge extensometer
was clamped to a \ in. specimen. The specimen was subjected to a
continuous constant stress of 2 ,5 0 0 lb/in (23 % of the short term
ultimate stress) for a period of 1 0 ,0 0 0 mins ( 6 days, 22 hours and
30 mins) and at a room temperature of 6 6 ° F . (19° C.). The various
values for the apparent modulus are shown in Figure 3-38. It was
surprising to find that the short term E value was 1.15 x 10^ ,
whereas the long term E value (after 10,000 mins at 6 6 ° F.) was only
0.88 x 106 ; a change of 23% • The time dependent E value was of
course considered in computing the principal stresses in the shell.
2. Poisson's Ratio
178
0.6x10®
tO3 ID ay 6 Days
3.38
TApr NUt' B F P 2 5
N>J 2 E R0 TIME L OG
TWO-GAUGE ROSETTES
43 - 90 ROSETTES
3-39
179
3 .4o 3.41
3.42 3.43
l8o
the tensile Poisson's ratio = 0.32; and the compressive Poisson's
ratio = 0.42 (Figure 1.2 ) . The value of Poisson's ratio for the
specific model material, was not determined experimentally because
it is difficult to compress a thin laminate without introducing
bending.
The average, short term ultimate stress, for the model material
2
was 11,000 lb/in (773*3 kg/sq.cm) .
l 8 l
Evaluation of Test Results
1. Dead Load
182
is explained by the fact that although the load on the entire dome
was uniform, the load on a meridional rib was not uniform.
It was most interesting to see that the low rise scalloped dome
behaved in a linear manner, in that doubling the load also doubled
the vertical deflections, and the stresses in the shell (Figures 3»51;
3 .5 2 and 3-53).
It can be seen that the design criteria for a GRP structure is not
so much the danger of exceeding the maximum allowable working stress
of the material, but rather the excessive and continuously increasing
deflections that will take place over the life span of the structure,
as the value for E is reduced with time. Straight free edges in an
GRP structure should, therefore, be avoided.
rib across the dome, under a uniform load over half the dome of
24 lb/sq.ft. at 6 3 °F. A maximum vertical deflection of 0.97 in.
(2.47 mm) was measured at % span in the model .. : .
which equals 4.83 in. (12.35 cm) in prototype. Figure 3-48 shows
the stress distribution, after 10 min. and 10,000 min of continuous
loading.
A comparison between the fully and half loaded dome shows that
the vertical deflection of the rib at £ span after 10,000 min.
increased by 22%, in the latter case, whereas the stresses at the
same point only increased by 2.7%*
4. Concentrated Loads
184
A-A
B -B
3.
4-4
A-A
B -B
3.45
185
Figure 3*^6 Stress distribution in the HP shell segment.
Load condition: 3*^7 lb/sq.ft. (17 kg/sq.m) * dead weight of prototype.
Duration of load: Top - 10 mins. Bottom - 10,000 mins. (approximately 7 days).
Temperature: 68 °F (20°C).
Information given: direction of principal stresses, stress in lb/sq.in.
(llb/sq.in. = 0 .0 7 0 3 kg/sq.cm.) (+) tensile (-) compressive.
186
.A
Cl
187
•2017.7
188
Figure 3-49 Stress distribution in the HP shell segment.
Load condition: 29 lb/sq.ft. (l4l.5 kg/sq.m) over half span.
Duration of load: Top - 10 mins. Bottom - 10,000 mins. (approximately 7 days).
Temperature: 68 °F (20°C).
Information given: direction of principal stresses, stress in lb/sq.in.
189
Figure 3«50 Stress distribution in the HP shell segment.
Load condition: 29 lb/sq.ft. (l4l.5 kg/sq.m) over half span.
Duration of load: Top - 10 mins. Bottom - 10,000 mins. (approximately 7 days).
Temperature: 68°F (20°C).
Information given: direction of principal stresses, stress in lb/sq.in.
(llb/sq.in. = 0 .0 7 0 3 kg/sq.cm) (+) tensile (-) compressive.
190
6. Plastic Flow
7. Failure Condition
The unit on which the guages were mounted and two adjacent segments
were loaded with a uniform load up to 48 lb/sq.ft. (234.24 kg/sq.m) when
signs of instability occurred. A maximum deflection of 1.57 in. (39*87 mm)
2
and a maximum stress of 3 1 0 ^ lb/in (2 1 8 kg/sq.cm) at t; span
was measured.
The "weak" area of the scalloped dome was found to be along the
rib at about f- the span from the support. Here the moment of inertia of
the cross-section was not sufficient, due to its shallow undulation,
to resist the bending moment and the axial thrust caused by the live
load. The downward vertical deflection at this point increased
continuously due to creep. It is considered that a live load of
48 lb/sq.ft. would eventually induce the rib to collapse inwards.
Final Remarks
The various tests demonstrated the great stiffness and the high
resistance against buckling of the scalloped dome made of HP segments.
It was possible to establish a relationship between the free span
of the dome and the necessary minimum shell thickness. An average shell
thickness (t) to the span ration (l) of 1 :1 5 0 0 was thought to be
possible. (t/l of an average chicken egg is 1:100). In a dome with
2
a diameter of 20 m. the dead weight of the shell will be only 4.33 lb/ft
(21 kg/sq.m.). Despite the very thin shell the scalloped dome
is not susceptible to the phenomenon of Msnap-throughM buckling, and
due to its plastic flow it has a high reserve of strength. The
structure also demonstrated its efficiency from the material side,
by the fact that a stress level of 3 0 0 0 lb/in (210.9 kg/sq.cm)
compression in the skin was reached, without the occurrence of
buckling. If we assume that the ultimate safe working stress in a
chopped strand polyester laminate to be 2 ,5 0 0 lb/in (1 7 5 *7 5 kg/sq.cm)
then we have here a structure in which the material is put' 100%
to work.
192
d e fle c tio n c u rv e s for p o in t4 (q u a r te r s p a n ) u n d e r UD.L.
24.0
91.7
ve 4
66.8 18*C
13.7
I 0300
.5 (7.62]
m
a>
■C
u ,21.5‘ C
C 0200
8.6 71*F
i
in
c
o
17 kg/sq. ti 20’C
3.471b. ’sq.ft. ad we ght
1 Day 6 Days
67*F
(0.17.) 1,000
63’ F
41.! kc /s 21.5'C
ght
ID a y 6 Days
3-52
193
s tre s s curves for ele c trica l resistance strain gauge No.25( 1/6 span)underUD.L.
1.500
(105.4:
1,400
117.1 kg 's l a 19.5* C
c u rve 3, VrO ttr scFt tr 3 67*P
1,100
(77.3;
91.7 Kg Is q- rr a 17*C
—
1.000 cc ve 4— « r6 Hr 'st |H t: a 63*F -
66.8 kg Is q n a 18*C
cL ve 3. 13./ 1b. 's it t
. a 65*F • _ _
..
Ib/sq.in. (kg/sq.cm.)
4- Ii.O v
cu ve 2. 8.6 lb' s<If t. a 71*F
-
17 k y/i 4-
c u ve 1. 3.47 lb. S( !•* t. = d ad we Ight a t 18 ■F
stress
3.5 3
3.54
\9k
S E C T I 0 N T H R E E COST
and
It should also be remembered that the material cost per sq. ft.
of laminate is directly related to the free span of the structure.
The ratio, cost per sq. ft. of covered area to the free span of the
dome, varies linearly for moderate spans and thus increasing or
decreasing the magnitude of the dome will also affect the cost of the
material per sq. ft. in the same ratio. Consequently with a relatively
expensive material as GRP, the question of dimensions gains particular
economic significance.
3- Manufacture.
196
REFERENCES
1. INTRODUCTION
The contemporary development of building mechanics is oriented towards
mathematics, thus offering a wide range of possibilities in the research of usually highly
complicated spatial-surface systems. The fact that the fundamental character of mechanics
cannot be ignored makes such an approach to the abstract analysis necessary even today.
In this, the properties of materials are not neglected at all, and they should be accorded
with the building mechanics that is mathematically oriented. The mathematical analysis of
these calculations is often complex and not adequate regarding the total time available for
the design of structures it is needed for. The approximation, that is the use of computers,
reduces long-lasting mathematical calculations and enables constructors to devote more
time to the design and construction. Analytical methods are supplemented by the tests on
differently sized models, which sometimes simplify significant assumptions of the
mathematical methods. The determination of forces and stresses acquired in the models
observing and measuring should not be underestimated, yet not overestimated, as well.
This paper represents a review of hyperbolic paraboloid (hereinafter referred to as
HP) shells as very frequently used roof structures and points out the possibility of the
mathematical analysis application in the case of small deformations.
2. GEOMETRIC CHARACTERISTICS OF HP
The ruled surface of HP is achieved by the movement of the line AB (ruling line)
along the two straight, mutually non-parallel lines AD and BC that do not intersect in the
space (Fig. 1).
x 2 = 2a 2 z ,
Analysis of Hyperbolic Paraboloids at Small Deformations 629
x = aρ cosh v, y = bρ sinh v, z = u 2 .
3. STRUCTURAL DEVELOPMENT OF HP
As spatial roof structures, HPs appeared in 1932 in France. At that time, they were
considered extremely daring for the contemporary theoretical approach, research level
and building techniques. Although HP is a spatial surface system with double curvature,
its realization is relatively simple, as all the roof-boarding elements may be placed in the
direction of the straight ruling lines.
Felix Candela, worshiper and author of attractive thin shell structures, particularly of
this kind (Fig. 2), greatly contributed to the affirmation of HP.
HP offers unlimited possibilities to architects and constructors in designing and
constructing. This warped surface may be applied over any foundation shape (rectangular,
triangular, circular, ellipsoid, and so on), and any building which is representative either
by its contents or by its position. A harmonious and daring structure of a unique form
enables creative expression (Fig. 3).
A HP surface may be continuous and homogenous over the whole foundation span, or it
630 LJ. VELIMIROVIĆ, G. RADIVOJEVIĆ, D. KOSTIĆ
may be assembled of multiplied parts that are copied by plane or axial symmetry. The
prefabricated construction with these shells is easily applicable, because the surface, even if
it is continuous over the whole foundation span, can be separated into integral prefabricated
elements arranged as strips or spatial rectangles that are compounded into a whole girder.
4. BEARING CAPACITY OF HP
Double-curved surfaces usually have satisfactory bearing capacities, while in HP it is
even greater, as the convex curvature stiffens in a way the concave curvature (Fig. 4).
Compressive stresses appear along the line a, and tensile strains are formed along the line
b. Having such stress conditions, HP is, righteously, regarded as a membrane, and the
calculation of the system bearing capacity is performed according to the membrane theory
with small deformations. In HPs, normal
forces along the ruling lines have constant
values, implying that there is no need for
the effect of the shell lateral forces on its
edges (edge beams), where constant,
longitudinally distributed shearing forces Compressive
are formed. stress
The shell principal stresses are created
alongside the vertical sections, which make
the angle of 450 with the ruling lines.
Placing edge elements or ribs that accept
the compressive or tensile stresses for the
Tensile
designed surface geometry most often strain
solves the reception of shearing forces on
the edges. Edge elements may be avoided
in highly curved shells, for spans that are
shorter than 30 m. In principle, care should Fig. 4. Membrane stresses in HP
Analysis of Hyperbolic Paraboloids at Small Deformations 631
be taken of the edge elements load, that is, their dead weight and its impacts. This is
particularly important in cases where they are asymmetrically loaded.
Many possibilities for the calculation of various forms of HP shells are given in the
literature. It is necessary to check the bending moments in the HP form that has a negative
curve in each point, or in the shells in which the relation of the span and structural height
cannot justify the rigid shell concept. Due to the multitude of forms, there are no rules that
could be introduced as generally applicable. Numerous theoretical papers that are
included in the references [1-8] offer clearly defined and tested recommendations for the
calculation of these shell forms. One of such recommendations is that the influence of the
bending stress may be neglected in the shells in which the ratio of the height and the
length of sides H/a is greater than 0,2 (9). In most cases, HP shells have great safety
against bulging due to their curvatures. In shallow shells, the size of elastic and plastic
deflection of console parts should be checked.
-------------- Tie
Fig. 5.
The areas around the HP shell edges - edge elements - are the parts with greatest axial
forces that are not in balance with the shearing forces. In order to prevent deformation, it
is necessary to place diagonal elements (ties or compression members, Fig. 5). The edge
632 LJ. VELIMIROVIĆ, G. RADIVOJEVIĆ, D. KOSTIĆ
elements are in many cases dimensioned not only for the axial forces but also for the
bending moments. Satisfactory results are achieved for the prestressing of edge elements.
is the vector equation of a regular surface S, and the surface S is included in the family of
surfaces St (S = S0), expressed by the equation
S t : r(u, v, t) = r(u, v) + tz(u, v)
defined in the points belonging to the surface S, which is the field of infinitesimal
deformations. The surfaces St, t ∈ R, t → 0 are the infinitesimal deformations of the
surface S if the difference in the linear element squares of these surfaces is an
infinitesimal value of a higher order compared to t, t → 0, i.e.
dst2 − ds2 = o(t).
This means that the curve arc length variation on the surface is o, δs = 0, in ID, that is,
the arc length of the curve on the surface is stationary in ID. The angles between the
curves on the surface are also not changed, as well as other elements that depend on the
coefficients of the first fundamental form.
The surface is rigid if it allows only for trivial ID fields. The deformation field is
trivial if it has the form of
z = a × r + b, a, b - constant vectors.
A necessary and sufficient precondition for the surfaces St to represent ID of S is that
the following is valid:
dr ⋅ d z = 0 (1)
where S : r = r(u, v) , z = z(u, v) is the ID field, and − denotes a scalar product, and x
denotes vector product.
This equation is equivalent to the three partial equations:
ru ⋅ z u = 0, ru ⋅ z v + rv ⋅ z u = 0, rv ⋅ z v = 0 .
There is a unique field y(u, v) for the ID field z(u, v) of the surface, so that:
z u = y × ru , z v = y × rv ,
i.e.
Analysis of Hyperbolic Paraboloids at Small Deformations 633
d z = y × dr . (2)
The rotation field for which the previous relation is valid is the vector field y(u, v) . As
the result of the ID surface, all its elements are subject to the rotation with the rotation
vector y(u, v) .
The field s (u, v) , determined by the equation
s = z − y×r
where the functions α(u,v), β(u,v), γ(u,v) satisfy the system of partial differential
equations:
α v − γ u = Γ111 γ − 2Γ112 α − Γ122β
α u − β v = Γ11
2
γ − 2Γ12
2
⋅ α − Γ 222β (3)
b11 γ − 2 ⋅ b12 α − b 22 β = 0
where Γ ijk are Cristoffel's symbols of the surface r = r(u, v) , and bij are the coefficients of
the second fundamental form.
The solution of this system of partial equations determines the functions α, β, γ. The
fields y and z are determined in the following way:
Being that
dy = y u du + y v dv = (αru + βrv )du + ( γru − αrv )dv ,
is the total differential of the vector function y , by integrating we get the field y(u, v)
that is determined in a unilaterally connected surface S. With such determined field y ,
the ID z field should be further defined. Namely, we have that:
dz = y × dr = ( y × ru )du + ( y × rv )dv . (4)
As
( y × ru ) v = ( y × rv ) u ,
the right side of the equation (4) is the total differential, so the field z(u, v) is determined
by integration.
We shall examine ID of the surface z = xy, HP. The vector equation of this surface is:
r = r(x, y) = ( x, y, xy ), (u = x, v = y) (5)
or:
r = x e1 + ye 2 + xye3 ,
634 LJ. VELIMIROVIĆ, G. RADIVOJEVIĆ, D. KOSTIĆ
By integrating, we get
Y1 = ∫ Ψ(y)dy = λ(y), Y2 = ∫ ϕ(x )dx = μ(x ), Y3 = ∫ xϕ(x )dx + ∫ yΨ (y)dy .
where µ(x), λ(y) are arbitrary functions. It can be proved that the bending field is trivial,
i.e. that the HP is a rigid surface with regard to infinitesimal deformations.
6. CONCLUSION
The HP is a ruled surface by which a space based on any desired form can be covered.
It is suitable for the systems of roof structures either as a whole roof or as its part. The HP
is a thin shell of a great bearing capacity. In spite of its double curvature, its execution is
simple due to the possibility of laying the roof boarding in the direction of straight ruling
lines. Such structures can also be built by the monolithic assembly of precast units.
The mathematical analysis determined the field of infinitesimal deformations on a HP
surface and showed that this type of shell may reasonably be treated as a membrane, i.e.
that this surface is rigid.
REFERENCES
1. Tedesko A.: Shell at Denver-Hyperbolic Paraboloidal Structure of Wide Span, Journal of the American
Concrete Institute (1960)
2. Candela F.: Stress Analysis for any Hyperbolic Paraboloids Architectural Record (1958)
3. Candela F.: General Formulas for Membrane Stresses in Hyperbolic-Paraboloidical Shells, Journal of
the American Concrete Institute (1960)
4. Herrenstin: Hyperbolic Paraboloid Umbrella Shells under Vertical Loads, Journal of the American
Concrete Institute
5. Bongard: Zur Theorie und Berechung von Schalentragwerken in Form gleichseitger Hyperbolischer
Paraboloide, Bautechnik-Archiv H.
6. Bennet: Hyperbolic Paraboloides, Architect and Building News, London (1961)
7. Ramaswamy Rao: The Membrane Theory Applied to Hyperbolic Paraboloid Shells, Indian Concrete
Journal (1961)
8. Soare M.: Contributi la teoria incovoierii paraboloidului hiperbolic i sa Studii si cercetari de mecanica
aplicata (1962)
9. Parne A.: Elementary Analysis of Hyperbolic Paraboloid Shells, IASS-Bulletin Nr. 4
10. Rile H.: Prostorne Krovne Konstrukcije, Građevinska Knjiga Beograd (1977)
11. Sekulović M.: Teorija Konstrukcija, Savremeni Problemi Nelinearne Analize, Građevinska knjiga
Beograd (1992)
12. Velimirović Lj.: Beskonačno male deformacije toroidnih rotacionih površi, Magistarska teza,
Univerzitet u Nišu (1991)
13. I.Ivanova-Karatopraklieva, I.Kh. Sabitov, Surface deformation. I, J.Math. Sci., New York 70, 2(1994)
1685-1716
14. I. Ivanova-Karatopraklieva, I.Kh. Sabitov, Bending of surfaces II, J.Math. Sci., New York 74, 3(1995)
997-1043
636 LJ. VELIMIROVIĆ, G. RADIVOJEVIĆ, D. KOSTIĆ
By
ACIQ\TOWLEDGENIENTS
·---.- . ....----
~----·
CON'I1EJ:TTS
-----· •-"'""_.._
ABSTRACTo 2
3
CHAPTER l.
1.1 INTRODUCTIOr. 8
1 .. 2 Geometry and Des~rj.ptic.n of the Hyperbolic Paraboloid Shell 10
lo 3 Review of Existing 'l;heories. 11
1"3~1 Membrane Theory. 11
lo3o2 Bending Theory.
1. 3. 3 Approximate S·~lu-Cionse 19
Review of Literature on Finite Element. 21
1.4.1 lfatrix Formulation of the Finite Element Theory. 21
l.4o2 Pro1Jerties of Elements. 23
l.4n3 Application of Finite Element to the Analysis of Shell
Structure. 24
Review of Experimental VJor:k 25
The Elastic Properties and the Analysis of Tinber Shells 27
Scope of Research. 28
CHAPTER 2.
•l'HEORY 1.
CHAPTER 3,
THEORY II.
Element PropeTties. 51
Structural Idealisation. 51
Suitable Elements for Idealising Shell Surfaces. 53
Element Geometry. 54
General Theory of Derivation of Shell Element Stiffness 55
3.2.1 Natural and Cartesian Definiton of Displacement and
Forces; Elastic Properties. 55
3.2~2 Natural and Cartesian Stiffnesses. 58
3 .. 3 Displacement Functions. 59
3.3.1 Requirements of the Displacement Function. 61
3o3o2 Displacement Function Previously Derived. 62
1. Non-Conforming Functions. Bending Element 62
2. Confor~ing Displacement ~~notion 63
3~ 3., 3 Natural Modes of Defoi'I!lation. 64
3~4 Derivation of Stiffness Matrix for Triangular Element
in Bending. 67
3.. 4.l Element Coordinate Systems. 67
3.4.2 Derivation of the equation of the Surface. 70
3.4.3 Symmetric Functions. 71
3,,4~4 Antisymrretric Functions. 72
3. 5 Computational Molecle. 74
6
CH.APTEB. 4.
COMPUTFJi PR'JGRIU!l.
CHAP_rER 5.
TESTS ON THIBER .P.Jif.D ITYBAK I:IODEijS. 98
CHAP'l1EH 6 ~
1.1 INTRODUCTION
/
/
Q-"('
/
' '-/Jlf
/ 0...
/
"
/ . / IJ'4 c
'::J
""
/
/
where
11
1.2
1 = -Z + Xz + Yx
k k
'~=Zk
.
-~
~x~y 2 2
where it has been assumed that Z = -q.
This equation has a general solution of the form.
4 = -l qkxy + f(y) + g(x) 1.5
2
The stresses in the shell are then given by
c~nstant throughout
the shell
·~2
Nx = rl.:.£{.tl , arbitrary 1.6
;) .2
'-'X
the edges are restrained by shear forces, then the shell should be
capable of bearing loads. It is thus essential to have edge beams
around the shell to provide the restraint.
where v4 and
u etc
w = F = 1.13
the second reduces to a single eighth order equation
Dv
8
0 + 4Etk
2
w"· •11 = -q 1.14
an equation which reminds of a similar equation in the theory of
cylindrical shell.
2. Flugges Equation
q = .4. qo
" ,.
~
r,: I ..,()
Sinyy 1.16
n; L
11:.1
where y = nn/b
which when ap~lied to the eighth order differential equation 1.14
yields the characteristic equation
( p2 - l) 4 - 1- p2 0 1. 1 7
p4
6
from which 4 rea.l a.11d 4 :hnagina:cy roots are obtained - Loof (l9 l)
pl-2 ± a1
p3m4 = ± cr.2 l.18
that the values o.f o:2 arc sma~.l and can therefore be neglected.
He thus arrived at a sixth order equation 1 requiring three boundary
conditions on opposite edges and has shovm t:hat no generality is
lost. He also presented a series of approximate formulae on the
hyperbolic paraboloid shell supported on flexible edges. These
formulae have been investigated by Batchelor (l9 6 3) who showed that
results based on these formulae failed to be correlated by experimental
resultso By making assumptions similar to that of Loof, Bleich and
Salvadori(l959) have reduced the equation 1.15 to one similar to
that of a plate supported on an elastic foundation~ In a similar
manner., to BJ.eich ar.d Salvadori 1 Vreedenburgh(l95B/59) has reduced
equation 1.14 to a fourth order one similar to that of the elastically
supported bear:i..
where gi(y) are the assumed\ functions which satisfy the boundary
conditions in the 3r-direction and f5_(x) are the functions to be
determined, The initial set of diffe~ential equation is thus reduced
to' the form
1!1i(0) = ? 1.20
where Fi is a set of ordir.ar;r differential operator and p are the
external reactions. The Kantarovi-~ch process is now used to minimise
the residual in the form
1.21
21
were carried out by Batchelor (l 963 ) o The shell was Bftx 8.Jt:x 3/41'1
with a rise of 141'1. The corners were all supported and tie rods
were provided. Arrangements were also made to vary the edge beam
stiffness. The model was of a reinforced concrete mortar. The
result of the investigation was very interesting. We observe that
the extensional or membrane theory failed to predict the behaviour
of the model, neither could the bending theory. A systematic
comparison was made between the shell behaviour and that of a plate
of the same dimension but with fixed and simply sup~orted boundary
conditions. It was observed that experimental results lie within
the solution obtained by treating the shell as a clanped or simply
supported flat plate. The result alsu showed a great susceptibility
towards the rigidity of the tie rod. With rigid tie rods, corrella-
tion with clanped plate solution was within 30°/09 whereas with
flexible tie the behaviour was nearer to the simply supported plate.
Batchelor concluded that the shell behaved more like a flat plate
since the deflections in the middle region of the shell was reasonably
predicted by regarding the structure as a flat plate.
A few test however have been carried out using fixed boundary
conditions by Ch&tty-( 1961 ), Brebbia( 1966 ). The deflections and the
stresses were observed to have been reasonably predicted by the
bending theory of shell. No tests however have been reported on
hyperpolic paraboloid shell on gable type or the simply supported
kind of boundary conditions although the possibility always exists
in theory.
27
THEORY I
-----
sional continuum.
a g_ 0
oz O'XX: Wx
a~ oy cryy
0 ~ 0 O"ZZ wy
== 0
I
'T
a~ 01' ~ <Jxy
<Jyz
c) d 0 wz
ai ~-~ ()x 6ZX
A~. !)./
I t'~
,L~'41. --
/ ~
/
Fi&. fl~.
34
Equation 2.9 is now cast into the more usuful form suitable for
engineering application in the form
§tv = Rtr 2.10
where S is the stress resultant replacing ~ and ":!... is the strain
vector replacing €_, R is the external load vector replacing both w
and ~' while r represents the external displacement replacing ~1. To
express the principle of v::irtual forces, 2.10 now assumes the form
-t t
v = R r
S 2.11
where the bar below v has been dropped to indicate that it is the
actual compatible strain, R is now arbitrary. If in particular w9
form a statically equivalent system expressed by
S = bR 2.12
and Eubstitute 2ol2 in 2.11 we obtain
Rf"btv = Rtr 2.13
and since R is arbitrary
r = btv 2.14
which is the mathematic expression for the unit load theorem. Suppose
however that S = Se a self equilibriating stress system, 2.11 now
assumes the form
Setv = 0 2.15
which is a formal expression of the compatibility condition. Equa-
tions 2.14 and 2.15 form the basis of the force method or analysis.
37
.J
'
If the stress strain law for the structural member is now given
by a relation of the form
v = rs + H 2.23
where f is the flexibility of the member and
His a column of initial strains due,for instance,
to self straining induced by incorrect fit, manu-
facturing error etc.
Substituting for v given by 2.23 in the compatibility equation 2.21
we obtain
b 1 t[rs + H] =o 2.24
Now substitute for S from 2.20 we obtain
bl t[fboR + f'b 1 X + H] = 0 2~25
Expanding and rearranging equation 2o25 we obtain
x =-
(b• tfb 1 )-l[bl tfboR + b' tu] 2.26
Substituting for X in equation 2.20 we obtain
s
boR - b 1 (bl tfb 1 )-lbt tfboR -· b• (bt tfb1 )-lbt tH
=
= [bo -b 1 (b1 tfb 1 )-lb1 tfbo]R - b1 (b 1 tfb1 )- 1bt tH 2.27
s bR - bt (b1 tfb 1 )-lb1 tH
where b == [bo - bt (b,tfb4 )-1bl vfbo] 2.28
The displacement of the structure ca~ now be obtained from
r ==bot[fS + H]
~ [bot.fbo - botfb, (b 1 tfb~ )-lb 1 tfbo]R 2.29
[bot - botfb 1 (b 1 tfb 1 )-lb 1 t]n
+"
= FR + btH
where F = botfbo - botfb 1 (bf 'iifb 1 )-l,b1 tfbo 2.30
F is ·the flexibility matrix of the structure.
k1j • • • • kln
kzj • • • • k~
= 2.na
k .. k.
JJ Jn
knj • • • •
or more consicely as:
S = [k] r 2.37c
An observation of equation 2.37 shows that the geometrical configu-
ration, the stress condition and the elastic properties should all
be reflected in the stiffness matrix. Also if the material is
homogeneous, isotropic and linearly elastic, then by Maxwell reci-
procal theorem, the [k] matTix is symmetric about the leading
diagonal and positive definite. To conform with the finite element
definition, the numerical value of n must be kept finite. Figure
2.4 shows two types of triangular elements with their boundary
stress patterns, namely a triangular plate element subject to plane
stress, and a triangu.lar plate element subject to flexural stresses.
The system of joint forces in fig 2.4b will be assumed to represent
the resultant effect of distributed stress patterns shown in fig 2.4a.
I K, 0
•
2 .. 41
•
0
I!
I
'·
If the matrix multiplication of equation 2.40 is carried out we
45
obtain N equations of the form of 2.38 showing that the [K] matrix
is therunassembled' stiffness matrix for the structure. It is per-
tinent here to remark that the kg's need not be of the same size,
this is equivalent to formally expressing the fact that a complex
structure can be imagined to be composed of elements of different
geometry and elastic properties.
Matrix [2 6]. The boolean mat:rix operations are simple and will be
illustrated under the section on computation.
Let the matrix [a] be partitioned in such a way that the trans-
formation atka is possible, thus (a] takes the form
a= \a1 a~.,. ag···· aN} 2.49
where N represents the total number of element. The stiffness matrix
is thus seen to assume the form
K = 'J-agtkgag 2.50
g==l
the summation being taken over all the elements.
K = rKII
LKot
KK.,0¢]
~ ~
2~55
"'· .
50
But K == atka
=i~~t~J k[a1 a0 ]
ao':l
=t· a ~ka; a,!ka 0 'j 2063
a. 0 ka 1 a 0 ka. 0 ,
==1" Ku Kio l
._Ko, Koo .I
The procedure is thus seen to lead to a gradual build up of the various
components of the K.matrix in their required form to enable the ana-
lysis to proceed as indicated by equations "2.57 and 2.61.
J. 70 Calculation of Stresses
THEORY II
54
ks =
( kr
I" kct
kc
k
d
l 3.1
where kd is the membrane or extentional stiffness
kr is the flexural or bending stiffness
kc is a coupling submatrix.
kr;
I
The absence of the coupling submatrix is apparent. Although the
order of the matrix is not reduced, the separate submatrices have
been reduced from three components to two and what is more from the
computational point of view can be ,senerated and stored independently
as
k8 == f kd kr j 3• 3
with considerable saving in computer storage. Also a general
purpose stiffness matrix results. This type of element has been
adopted for analysing the hyperbolic paraboloid shell in this thesis.
in which
Pi 1:1-i Vi w·J. lI
= { ilr. t
, J.. J 3.8
(3 xl)
Next natural forces and displacements are defined. The natural
force and displacement vectors are arranged in the sequence
PN = {PNM P1iJ:B i
PN = tP}m PNBi 3.9
where the subscript Nha.s been used to indicate the natural system.
The order of the matrix vectors PN and PN is influenced by our
definition of what we called the natural system. If we merely
ref erred to the natural system as mere transformation to a conve-
nient local coordinate system the P, PN,P,PN are of order 18 x 1,
assuming we include all the degrees of freedom in PN and PN• If
on the other hand, we neglect the rigid body freedom as explained
later and identify our Pi-..i and pN with straining forces and straining
57
PNM ::: l
:& p ~ p y J i P1m = ~ TS TA ~ 3 • 10
in which Pp Py denote any three independent forces system which
P~
the relation
-1
C.'N "" EN O'N 3.14
where EN is the 3 x 3 natural moduJus of elasticity and may include
arbitrary anisotropyo EN is of the
r K11 K12 K131
Eii =E I K21 K22 K23
l_K31 K32 K33_i
where E is the cartesian modulus.
3.19
3.20
where
= f V fam;ft 0 -J EN f aml[ 0 1dV
l0 aNB L0 a:r._i:MI
= J Fm11t Ew amA: ; a:r:rn t EN aNBj dV 3.21
= rKM K13J
59
where
c cl
-
c
(ccx.x
=\cfjx
Cay
c {:ly
Ca·
c 13z
1
Cyy ~Cyx
Cyz
. /
I
and Cij is the direction cosine relating axis i to j.
w = [l x y x 2 y 2
2. Conforming DisElacement...J'unction
3.33
Thus ~k = ~oi,j
6ijk
therefore
Now 6. ijk
I 1
X·l Yi
I~
1/2 Xj Yj
Xk Yk 3~37
1 x y
6. ojk = 1/2 1 x·J Yj
11 Xk Yk
where the vertical bars have been used to indicate the determinantal
evaluation of the terms within the brackets. Expanding equation
3. 37 in terms of the first row and denoting D.ijk by the scalar 0
then Si
=
6. ojk
6. ijk = <lxXkj Yjl
Ykl -x 1
11 Yjl +y
Yk
where ai and '3i have been used to denote the coefficient of x and
y, and Yi is the constant in expression 3.38 above.
Similarly 3.39a
and
3.39b
kp/ jk = sl/sn
therefore sn = sl 1 isj
-2
vr(s} = a + CZ 3.44a
w(s) 3.50
w ::; bz + dz 3
- -e 2
= ~a(l - Si)
3.52
With one edge displaced and the other two supported, we can now
write
73
Thus 3.53
and
3.54
74
l~ 3.57
1 == o)
LI,
where L == f li
lkj 1.
J
3. 58
and ln has been defined as the side length of the triangle opposite
to node n of the triangle. If now we denote by X the row vector
of the polynomial expansion in Si~j up to the third power thus
x =mi "Sj ~i~j si2 $j 2 si2 °Sj~i~j 2 ~i 3 Sj 3 ]3°59
then we can write the set of displacements derived in the previous
section using the equation 3. 36 to eliminate-~ k as
Idx2
II d --~!.
2
3.61
=
. a.y2 I
\2d2w 1·
i-
l dxdy)
Similarly with respect to the natural triangi:.lar coordinate system
defined in the previous section, we also define the functional µ
given by
r
2
,
'
I
I
d
clf
w
..... i·2
!
I
.. I d2 I
µ··
5 = w
d't.. J.2
u.2 w
I 3.62
d$.d:$.
l.. J.. J
By partial differentiat.tm of the displacement function Wll with
respect to x and y making use of the relation (3.49) between x, y
and~ i ~·j $k we obtain
µx = [a] µ0s 3.63
where the matrix a is given by
,- a·2 a·2
I J.. J
a =
I
. 2(o:i/3i) 2(o:j ~j)
'-··
f3i2
µx = aXc 1 PN 3.65
where the explicit form ofµ is presented in table (3. 2). The stress
matrix rix is def·ined by
76
where
for a material with an isotropic elasticity, k has been evaluated
and presented in table (~3). In the integration involved we have
made use of the relation
r. m nd,.. 26m! n!
JS$i sj : ; == (;+n:2St
where 1::::,. denotes the area of the triangle and the natural stiffness
matrix is given oy
where w = ~w·
t l. Wj wk[
e == ~e·
\ l. ej Gk{
¢ }t <b.
' J. ..
¢; Cl>k}
and e = - dw/dy
cil = dw/dx I
=A p 3.. 8lb
The A matrix is presented ~n table (~4)
Now solve for PN we obtain
-1
PN = TNN A p = aNp 3.82
k matrix is now obtained by the congruent transformation
k = ~t 11i 8N 3.83
78
. -1 . . 1 .
-1 . . -1 • .
. 1 . . -3 .
l 1 -1 2 -2 .
[c] == l . . 3 . .
. . . -1 3 1
. . . -3 1 -1
. . . . 2 .
. . . -2 . .
. . 2 . . 252 . 6!1 .
[x] == . . . . 2 . 2~1 . 6"%2
. . . 1 . 2gl 2'§2 . .
D13 033/2
A 8 c E
F G H J K
A =
2011/3 + 2 013
B = (D 13 +
D 3 3 ) /3
C = 2(012+023)/3
E = ( O11 + 03 3 + D13) I 3
F = 2(012+013)/3
G = (023 + 033)/3
H "' 2(022 +Dz 3 )/ 3
J = (033 + 012 + 2013 + 2023 )/fJ
K = (022 + D33 -t- 023 )/3
L = (D11 + 2013 )/2
M= (2012+ D13)/2
0 = (0 12 + 0 23 )/2
p = (2022+ 023)/2
80
4 4 _2 2 2 2
D11 = al + f31 + 2val 131 + 4yo:l !31
]) 2 2 2 2 [ 2 2 2 2]
12 = a1"0:2 + !31 f32 + v al 132 + a2 131 + 4Yo:1a2f.31P2
+ 4ya2f32[i:x1f32 + i:x2f31J
y = (1 - v)/2
v = Poissons Ratio
+:i ~. '8. I
82
z
X FORCE (disp)
7--------~\:---,o..
~-\\
0 x 0 x 0 x
n·I
~·h·
I I
r
} t .. k
lx·y·) I ( xkYk)
Jl
)(
F~G 3.6
ELEMENi
-
GEOMEi~Y.
-
00
0\
i<t.I
_ CHAP.f'ER_L
GOHPU~R PI:O~"F®!
matrix: for a shell elemant, both the inplane stiffness matrix and
the bending stiffness matrix fo::;::o the element are required.. Quite
a number of acceptable fOfllls of the inplane stiffness matrix for
the plana:i.' trianglllar elel!l.ent are available (26,29,32) and the one
in refer~nce (29) has been chosen for inclusion. in the shell program.,
Of the bending stiffness matrices for the triangular element reveiwed
in chapter 31 two have been incorporated in-Co two separate shell
programs main:y for corr.pariso:.:l with t:r,e one derived in this thesis.
The details of these prograr.is are no-.v prese:..'lted.
r
I
Pm = 1
1 Pmi Pmj Prok
a:ad
Pr,;. \Fti pbk~
Pb Pbk~
where = !Umn Pmn { ~m vmn ~
and = nri
i. xn
VJ l p
n! bnl n :·n
== ) 0
n \
n w \.
Pm
where p
1 is of the same form as p in equation 3.30 but in the local
system.. The explicit form of r,\n and M:b are &s in table 4.1 and 4, 2
-&·
I
~-I
~·I
u.
I
1 __J__
v. , -- - U·
.J
r
I V·
·-- J
U· 1 w·
J
V·
J
- , I
!
I
! I
-&·
j
J
c:
I
I ¢·J.. co
1 I
"kj
vk
I
, ~·
_l_
Uk
M vk
wk
-&k
~k
~k
s
TABLE 4·1 COEF OF Mm IN EQ. f..f.
LI·
I
V·
I
W·
I
-e. -e-.
I I
¢·I
~-l
>-----+-------· - __________ , _ __, _______________ ~ -------- ---+---·+--- +----!----+
~i
w.
I
-~---,- -~------ - · - ----1 U·
J
_.
-fr. >------>---+------.--·---+----+---
r··.
VJ
w.
J I ¢·
c---"-- - .-- ---- - ~-- ----~---+----- _ _ _ , _ _............._ __,___ _+ . . . - - - - . . i . - - - ' - 1 - - - t--------+-___j_---+-----i J
~j
Uk
vk
wk
-e-k
~k
~k
TABLE 4·2 COEF. OF M.b IN EQ. 4 ·4 - s
so
11
c c c c c ·
c ~ ~ ... ' 18 x 18
l he global s-tiffness matrix is now obtained 11y the congruent trans-
formation
k 4.7
.... 4-13
PU
• IJ9o<•)
Pm
I 6 = 6x 6 unit rna.tri x ...
(. •)
\..0
EQUATION 4-13 I--'
FIG 5-1
92
N
The entire set of 3 x N numberc is read in one go. For each node,
starting with number l~ the set of 3N figures in the above table
is scaned three at a time. If within any scan, a number ooeurs
which is equal to the noC.al n·..:1mber uncler the control, the element
~umber in which it occu~s i8 noted ac well as whether its position
ic 1, 2 or 3 corresponding to ~ 1 j or k. Also of im~ortance is a
counter which registers the total nw:z:.ber of elements meeting at
the l'!.odal point in question. This is 11seful in compiling equation
4~14 ; as well as finding the average stresses at the ncdal point
as will be explai~ed later on.
[~:]
I
partitioned by rows thus:
k ..
kii
'V:· .
·-Jl
kkj
kki kkk
lJ
kjj
kik]
kjk
[~~]
l
J
.th
.th
row
kth row
row
4.15
. (n + 1) (n)
ri - ri 4.18
and ::; kii-l r;i - >kijrj(n+l) - ..:;>__ k l.J ~
· ·r J· ( nJ .
4 19
where n is the cycle of .aeration. In theory the algorithm ought
to work, but the result of the computation showed a divergence
after only a few iterations, and the method was therefore abandoned.
K 32 K 33 K 34
I
'
I
K 4.20
I
1.... Kqq.!
By symmetry considerations
t
kii = Kii
t
K. . l = K. l i·
J.' J.- J.- '
== R 1
== R3 4~22
+ KqqWq
Kq q-lwq ... 1 ==Rq
The first equation is solved for rlo This result is then used to
eliminate rJ. from the second equation which may then be used to
solve for :r.·2. This may then be used to eliminate .r2 from the third
and so on. The sequen(!e of ope:cation is represented thust
Eqn. 1. r 1 == Kl l -l [ 111 ·~ Kl 2 r2 ]
1 _, 1
R1 - X11 -K12r2 == R1 - Q12r2 4.23
:::ubstitution of l.'J. in equatirm 2 give..;
= R21 .. 'i23~
and generally
~
obtain 'the va.l-4_C':S of the r's .. It will be observed. that at any stage
only Qij submatrices need be stored, thus leading to a g:re.a.ter
saving in space compared to the ruethod described by Paullin~ 6
( l9 4).
Machine size limits only the size of the submatrices Qij but not
the total number of nodes.
97
the· eq_uilibriurn of a n0de under the act ions of the forces contrib~1 ted-
/tNO '1~'1'€~S
er·
__,o
1
The objective described under 1' was primary, while those under
'-2' and '3' were secondary as far as the investigation... cnr3:-iod out
in this thesis is concerned.
_2~actor Governing the Ch~ice ~j.__:~ca1e and Material_~f.. the
----
Mc,del
~aron
(1962) reportei an exploratory test on a timber c0noid
shell roof, carried out at Southampton University. The object was
to compare the behavior of the model with that of another model which
was a half scale of the prototype& The material used for the labora-
tory model was balsa wood.; '11he choice of balsa was influenced by
the fact that it grows uniformly than any other species, thus ensuring
some degree of control on the density and hence the strength proper-
ties. The apparent success of the model in predicting the behavior
of the prototype in a rather qualitative manner prompted the choice
of the material for the model of hypar shell planned in the present
investigation.
Two types of glue were tried for the membrane. The first was
the traditional balsa cement normally used for gh1--ing pieces of
balsa together. The adhension was found to be poor and large
J_Ol
The edge beams were made from the same veneers as the membrane
of the shell. Each beam was l/4 in. wide and 3/4 in~ deep. Eigbt
lengths of 24 in. each were prepared. After the shell dried, the
first set of four beams was placed arounC:. the top surface, and after
drying, the shell was trimmed and taken off the former. The bottom
set of edge beams was then fixed~
.2.~4--2.£rner Detail
An air bag was used for applying the load on to the shello
The choice of the air bag was influenced by the need to apply a uni-
formly distributed load and to obtain an arrangement which permited
a continuous incremental loading to failure, with minimum of inte:?:Ven-
tion with the test setup., The loading face of the bag, which was
supplied by P. B. Cow r.tdo, was shaped to the same surface as the shell
and it was enclosed within a box fitted to the frame within which the
shell was supported~ The enclosing box is of the same shape and size
as the former shown in fig~ 5ol except that it was completely hollow.
The air bag had the internal dimension of the enclosing box. The
al:i:.· supply was· obtained from an 80 psi supply using a reducing valve.
The p:r:·essure of the air in the bag was measured by means of a water
manometer connec·t;ed to the supply line going to the bag&
c~ The Young's moduli obtained from the tension test specimens are
p:r:esented in a graphical form in fig. 5.20. A typical stress-strain
relation is shown in fig. 5.18.
a. Bag Loading:
be Su~port Condition
The Youngs moduli curves presented :i.n fig8 5.20, show that
superimposed on the variabi1ity of material, there is a definite
evidence that the glue has a considerable effect on the tensile
specimen.
Fig. 5.21 shows the plot of the displacement along the edge
beam. The actual displacements ha-ve been -plotted to show in parti-
cular the deflection of the low cornerg If however the deflection
values as shown in figo 5.14 are compared for the low corners and
the high corners 9 there is definite lack cf symmetry and the reduct~~on
The material chosen was a type of plastic under the trade name
of VYBAK. It was chosen because it has a low softening temperature
of 90°c and could be easily formed to any shape. In contrast to
perspex, it dces not creep very much at room temperature. A~nroxi
mately 2 ft. square of the material, which was l/4 in. thick, was
placed in the steam bath and softened. The softened vybak was then
placed over the former fig. 5.1, weighted down and allowed to cool.
When cooled 1 the model took the shaue of the former. A certain
amount of local irregularity was obse:i.."'ved on one of the edges, but
this was soon cured by reheating and reforming as described above.
One quarter inch wide ~ieces of the same material for the edge beams
had meanwhile been cut and glued by means of Tensol cement which had
been recommended by the manufacturer~ The joint was found to be
structurally unsatisfactory, and other types of glue were tried
with little success. A bolted connection was therefore chosen both
for the ed~e beam and the corners. The edge beams consisted of
two 1/4 in. by 1/4 in. laminations symmetrically placed at the top
and bottom surface of the edges of the shell. The edge beams were
attached to the shell by 1/8 in. diameter studs at 2 in. interva~
along the edges. Fig. 5.6 shows the edge beams in position and
also the details of connection.
lower corner plates. These plates were provided withcl/2 in. wide
horizontal slots to allow for free diagonal horizontal movement of
the shell if this is desiredo A Y-p:...ece plate was provided at each •
corner and formed in such a way that the V--prongs o.f the piece follow
the profile of the edge beams and was attached to the shell edge
beams by means of eight 1/4 in. diamet~r bolts. The tnil pieeP v:::·."· pro-
vided with two vertical slots, to ensure that before bolting to the
piece on the frame, the shell was not constrained in the vertical
direction. All the corners of the shell were fixed4 A general
view of the shell mounted on the test frame is shown in fig. 5.7.
The model was loaded by means of lead shot which were measu:L'ed
into cans suspended from the model. 1/16 in diameter holes were
drilled through the membrane on grid points at 2 in. inte~·al as
shovm in fig. 5.13. There were a total of 121 such holes. The c
Strain gauges were fixed on the model both at the top and
bottom in accordance with the distribution shown in fig. 5.10.
Two types of strain gauges were used. They were the 5mm rosettes
and crosses. The crosses were used on the diagonals. The numbering
of the strain gauges for the purposes of interpreting readings
108
obtained from the Solartron are shovm in fig. 5~11 and fig. 5.12.
For the deflection measurement, dial gauges reading up to 1/10,000
in~ were used. The distribution of the dial gcuges are as shown in
fig. 5.13.
reduced strains fo:r. all the gauges at the top and bottom respectively,
Fig. 5.32, 5e34 inclusive show the plots of the strains at the
indfoated positions shown in fig, 5.16 anci fig. 5.17.
lowcornerr
I-'
0
Ftr.. 5.').
112
direction of load
applicnt ion
( b) corner detail
~·
11
3" 3" 3 3" 3" 3" 3" 3" 3" 3" 3" 3" 3" 3' 3"
I I I I I I I I I I I I 1 I I I I I _ . ./
Hz ._15_ _ _~1._6_ _ _ _ _ _14-.Lz_ t H) 15 L2
(-r)1
J_ 16
H -1-
• 17
12
•
:t
: I
1
11 II 12 131114
11
•
10
• .a
9 ("f')I
•
._______...____-----tr--~-.~-,.---H1 ~1 -
5
-
7
- -
1234 5 6 7 2 4 6 8 10
(a) posit ion of dial gauges ( b) posion of strain gauges
0
reaction
plates
d isplace<rent
trans due er
-I I I 1a''
·1" thickness I
variable
0
1
FIG 5.5 TENSION TEST SPECIMEN
Tl M BER AND VY BAK MODELS.
Fl G 5.5 CORNER AND EDGE BEAM DETAILS VYBAK MODEL
FIG 5.7 LOADING POINTS AND STRAIN GAUGE DISTRIBUTION ON VYBAK MODEL
llG
Hi h Low
,(.
>- ~
>- ~
I
Iz,~ ~ ~ '"""
2"
I I
~
'""
~
>- -'
,(,. "'"' --" -' ~
"""' ~
'( y """""
y
.
y
'<cross }- rosette y
y
y
Low High
?~,,., ~ 'r'b"y1f!
C
"'
-<
/()
t;.O
ff"'/)j
I I
7'~
1+
4-5.J..7,,,,_ I
I').
r,./
4
4/-"' 71.
~:!."""'
7'!
4-?> .....
'7+
C\ 1-b b 7 8 9
\( 37""";,e 3,5-'.,"' ~9"""'1c .fC..,.11 I
I '). ~ + 5
3'2.~
. li~ ~..l_.64 *'""-e; ~5'"" 61o 'Po.A..67
low high
high low
low high
high low
!<
2~
+
Ii. +
0
...
0
+ . ...
0
+ .... + +
0
+ . + + ... +
0
...
low high
VYBAK MODEL
122
high low
~ 16_ ~
0
·O-'
• t:>f I
·0.2& I
0
- - - - - - !'JI
l'"'q
•/ID
-
·.11. .g:z
'"""
·DIS
•g"" I
-11 -·""'
·DIS
·DS'- '"''
·DGI.
.0•7 •/l"f
1l 10
- . b'10
'""7
•OIZ.
. 0~1
~
·DOZ
.. ,,, -·4'01. -·o•I
-·015
•.OH •ODO
• llOft
.010
-·PIS
• 11•1
-.11115
-.0111
."""
.001 ·'"'
.ooJ
•(10~
.."""
•oJ.t • 010
."'' .OID
""
2 3 4 5 6
low high
readings are for loads corri:>sponcl!ing to ~·~"}
5·/S f''f
J/J..·SO
,,.0 ·.50
FIG 5.14 READINGS OF DISPLACEMENTS
TIMBER MODEL
123
high low
l(q
"-s-. 0;· /
/~
~ . o-6;·
o'
"•
•Oil o~
(,
~·
•
·Db8
•
•071 •
·DOa
A-
.4178
•
.(J~~
• ,
.,57 ·D'J
•
·PJf
•
·D7~
• -··.A
• •
·071
,cf'/
•
"""'
.mo .c-15 .~61.
·"~ ·"::§8 .()16
8- • • • • • • -B
•
•tJ3'5 •
·c~
•
,Cf() ·~If
/
C-
low
.Ot15
-
.qlf
- -
.tJlf ·~3 .~""
"""6 -c
high
A
6
A
5
A
4
A
3
A
2
A
1
,,
·""- A
2
/\.
3
/\.
4
/\.
5
/\.
I"
124
-<.
S>
" l
•
~~
~ 3e\ -
- ---- -
~ ~
'.;>
J ~
·'~ .$'
I
.99
~
I~
-;;..
a ~
%-
,.- I ' ~
t
.9.9
/~ I~
~ ~ \
t ~ ~ I~ ~I
~ ~o,C
~ (S'
~ ~ ~ ~
/
c,
-iJ. ~ ~
1t U"
~o."\
low t
D
strains >< lCf 6 high'.-<.
hig n low
~
,
'
I
~
/
~
.....07
I
,
....
~
I>
I
'
~ f
'"a
- - I
- - -
-1~ ~,\k -~
<.y ~>
'~ -~ I ~,I/'
"'\. ~
~1' -~ ~
IO
~ '>. ~
,
~ :k ~ ~ ~'
o..:o...t
l( ~ I
!I.
)'
G ~.... ~
~ 11'
-~
....,. ~,,. ~ 3-
~
'
,1
Ill
Cl-
800
Vl
Vl
/
-
.... /
C1I
0
Vl /
/
/
_q,.I) iJ.O
3
strain 10-
FIG 5·18 TYPICAL STRESS-STRAIN RElATfON 3 ply.BALSA SPE.
0
20
......
Vl
a...
Ol
c 16
·-
"'O
d
0
~ 0
-
C1I
u
~ /0
....
::J
Vl
/ff ,.
/
0 of O·Z
displacement at centre (ins)
FIG5·19 LOAD-DEFLECTION RELATION FOR CENTER OF TIMB::R SHELL.
tens.ion specimen No.
127
"°',_____1_ ___:_i_ __::.o_ __,_4-_ __:_s_ _6_ _;,.1_ _a_~~--10_-.
0·6
1·5
/·0
x
Vl
:J
:J
"'O
0
l·O
E
Vl
CJ')
c:
:J
0 0·5
>-
thre pli s.
mea : 1·4- 0 x10'-
o.____;,___.....___,__~----L-~---'----1----1----1----'
5·0
~
~
I
~
~
I(
II)
0
.50
~ c---- l
-
'J- .iru·
't· ~·~
VJ
-100
~~
'll
I
~
1·.20·5 ('$1
()
~ •JH ('I/·
)(
V)
~ -:z~
~
~
"'l -:o ---·
'I
w
~ p
~
"-
~
f e·J
~
~
129
I 5
-- ....
;
~
~~
------- ~
·02 1--------~--ji---------+
.QB'-------'---------'-------+--:-~~~----'
__)--
~
---
_/""
~-
·CZ.
130
~ ..... -
P/)J/1"1t>N er GALI(,£~ PAI THE "'iy6Mt:. "1"1>6/,. .
t I 2. 3 4- 5 6
a
~
~ ·t/Z
'-.:.
"-
~ ·O+
~·
Iii
~ .()6
~
....
V\
~
·Ob
/)
.......
~ ·C"2
~
"-
i-
Iii
\J
·04
"I: .p(,
~
Ill
{S
•08
-......
! ':I ~·~
~
"'~
t tJ• I
~
...
\I\
~
~
~
tJ d ·o? t:NtJ
t:, t1"V I LA-1.. /:J/ SPJ.,+cE"'f6 N I {.IA./ S)
,C:-1 ti s-: 31
131
I 6
I
/P8 ------ I
:t=-=-----11:-------t----......=.~-1-·- .----- ---
+- - --- j.
I
-------r - I tI
-1'4 J · · 1;-:::,::::L
I
II
--T----
1
--r-
i
I
-~1-------- -f
() _ _ j____ I
'
-JDC
rit:. ~33
Jco
'()
/40
•
~ 0
x
~ -100
~
~ .J()()
"I ---
I 111 /l,fVu f
-3Co
J .80/10117
--~----·
6 _1lus
..f/6 5·:?f
132
240
2.10
/80
/50
---....z
-......, /20
\')
Q
~
\I\ 90
60
r;.szx10' tb/111 2
2
0·80 X fO' lb/M
OL--~~~~~~!__L:__~~~~~~~~~~~~..__~~~~~~~~~~
.
AUTHOR MESH CENTRAL DEFL. o/o of Exact
ZIENE:IEWICZ 16 x 16 1.130 90
TIMOSHENKO EXACT' 1.27 100
an error of 1 fo .
Fig. 683 shows the result of the computation for the plate on
only four corner support~ This problem has also been discussed by
Timoshenko. The exact solution has been given as 25.7 and the
finite element solution for a 20 x 20 mesh is 26 again showing an
error of +l cro.
136
Fig. 6~4 shows the plot of the displacements for shell type
of problem 2. The data for this problem are as follows: length
100 cm, rise 10 c:rJ. and thickness 0.8 cm~ We note that the 8 x 8
mesh which was quite adequate for probleB 1 is no longer accurate
enough for problem 2. Thus we had to employ successively 12 x l~
and the experimental result, does not seem to be adequate. But more
disturbing i8 the fact that when the theoretical values are compared
with the experimental values, both results do not agree and are not
of the same or1er of magnitude. The experimental result is shown
on this g:raph but to a scale ten times that of the theoretical curves·
scale~ Table 6a2 below giYeB an idea 0£ the order of' I!lagnitudes
:Lnvol ved. The displacement has been compared at A - the centre of the
shell, B - a point midway between the centre of the shell and the
edge and C - the edge po:!.nt on the central lineo
0
M:E:Sll DIS PL. "!0 of EXl'T DISPL.
I
.§...d-.2 Comparison of' the curved pla;te solution and the f'lat plate
solution
In f'ig. 6.8 the computation using the shell data f'or a simply
supported plate has been plotted. Comparison between the simply
supported plate solution a.rid the curved plate analysis shows that
the shell stif'f'ness is roughly of' the same order as that of' a
f'lat platee This f'act has also been pointed out by Batchelor (l9G3)
The results of' his investiGation on a concrete model with boundary
conditions similar to the one conducted in this thesis also £'ailed
to be correlated by the shell theory and he therefore compared the
behaviour of' the shell with that of' flat plate under fixed and
simply supported conditions. When he used a flexible tie rod
his results compared with a clamped plate solution, while with
144
a more rigid tie rod, the results compared with that of a plate
which is sin;ply supported. His broad conclusion that the shell
behaviour could be approximated by the clamped plate behaviour must
be open to eq_uestions. Clearly the results must be governed by the
shell rise. If the warping constant is reduced to zero, the shell
tends to the plate on four corner support to which a solution has
been given in section 6 .. 1. Thus with very low shell rise one might
expect the result to tend towards the solution of the plate on four
corner support~ The little information available cannot make a
generalised conclusion to be drawn between the solutions., However
one can say that the shell does not behave as a plate under identical
boundary conditions.
6• 5 GENER.AI, RBMAR...TCS
behuviour and
.
out on ·the shell with the flexible edge
the
b,:r assuming the shell type
curved plate action seem to suggest that it is
associated with the membrane stresses. The portion can be taken
as the slowly converging 'Jortion of the analysis. As shown in the
comput~tions on problems 2 and 3 its effect is minimised by division
into finer meeh. I'he wor;:iening of this e!'f'ect in the computation
on the shell with the flexible edge support however suegests that
the boundar.1 conditions could also have some part to play. 'l'his
effect is totally absent on the plate problem~ and the curveO
plate analysis.
6.6 CONCLUSIONS
The shell theory has been shovm to fail to predict the beha-
vioUI· of the hyperbolic paraboloid shell on flexible edge supports.
This failure has been explained by the fact that the shell does not
develop membrane stresses as assumed in the shell theory for this
particular boundary condition, An experiment conducted on a Vybak
model established that the shell transmits its load by means of
bending stresses which are associated with plate type of actions.
A modified shell theory which neglects membrane actions has been
advanced to explain the behaviour of the hyperbolic paraboloid shell
on flexible edge support. This theory captioned the Curved plate
theory has been shown to predict the shell ·behaviour under the
flexible boundary conditions.
148
t TI
I '-----,/'---->'--~'----,l''-----,1'~----JL---IL--.I -
I
I X·
f-71'--7!'-----71"--,.J"--,.J"--,JL____-,l"-----J,L__J.L---.j'I
I l'--,!----7r----,,~'---JL___J"---JL----J."---~~ ~;
~I { f--7!'-----?f----7:"-?~"---j"--)L___J.'--+~ ~.
~
t---- - £ - - -~
~~ ~____,,____,, i
f· /../~ -+ i~ fl no. or .f'4'vit1- 'x 1
x:
150
·-=7~-----·---+--------+
g,5 - - - -
__.
--------+----
I 7t!"ftJSf{£1(K.O - -- -
1
._ a. aour.H - -o- _ 1
.
J 'r!#lf(.!EWIC;t - -
J.~,_____-_-e_-_-_--_~-------~---4~_$_E(~•U _______~_
F'/4 G·:2 .Stf/U!-1!..E Pt.A 7£ Nt7# ti."1f'£b E~&U. IJ~!!. (),/J. /.. (Q)
c~fvffiJ,£/S~IJ "F R.6Sfk.T.S IJ$ir.J~ Vl'cR.ID/IS F'VM:TIC#..16 r~ /Ct-..6u.:~
,cu•~N7 ~ th"'tb #/£..SH 1>1YrS11!1Aif 1>r 7~ PJ.ATF
--'
--·
-
-
flt. G·3 5~lJMU PlAT5 (:N Fcv~ e~EA!.. JuPP6R/S
.f1A11r£ £1...Eg.;"'1 S"1.v11"AJ 111.1w.f. 1't""'1~v~A1(' Iltt1'Jtl?(s { Je4t.1111)
151
l
Or---------,--------,--------....--~~~-::=---.-
()./ ~-------
--
(}/}. -
03'--------->-------~
·-- 5t/.1E$ 5oJ."'· ((!, f/,t.,r2J.ft>AJI(,>)
Ct.r;11~µ}
SE6UN
16 Kl6 1-fE.># .f'iv11f. !wrl/.Jr
+ £?,1.oge,/A : EJtp'f. (£.!Il.!Lr5
()
--.,
"'~
-.::,
......
~
~ O·OI
"~
"
~
c:;
f"f 6. 0·5
~
/, .
/
·001 ----------'+-------------+---------+------,~"-T-------+-
v .
,//~~
--
---------~·-·---·
----
\..!.J
/" .~ L
OOit===;;i-::e::-~-====::.;···t=-==-=-=-~--~...;.--:e=::-~~~=------,-?i"i..JEif
-rr- :t.
s
-r1A11rr
_ _ _...___ _ _ _ _ _ _ _
"
1&1i1..,.;.
61.6"'11NT
,, ~
~-· MESK(jf&tN
i•·r~ .. I
-------<
.r:i6 6· b
t
o <i ~
,__~~~~~~~~...-~~~-.-~~~~...-~~~-,--~~~~.--~~~_,.....~~~--to
1a
~ \\l
fl
~ ~
"'~ J.·5 ~
~iq
~·
~~ ~~~
~
'><
)\
~ "I
~ I-.
·~ 50
t-
!--'
~
~
~
~ ~ \J1
"'...
~
'U
\()
~
~
~
['\)
ft."""
'II'! :\
~
~
.Q
">
~
'l 1·5 ~ '
<l
'.I.I
\:. ~
•41 .._
ti.I. ~
~
. ______,____ L__1.~- - - - ____________ -~ /. S°¥/'I· fes11f1. Vy&A* ~r:>£L ~
'><
l- !2. .f. E. 'htrt 1'.1~i ( lb•l~rtt4 '4
g. " p (f-$ • )
N ---------- 10
FIG. o.]
--+-------~~--
I I
]: ·00!251--------+------l---+--- ----~-__,,,_
1 1
~ L I
~
~
~
"'>:.
~
~ ·P"5"1--- --+-------+-----~- __ _..,,o~
~ I
~ I I
l~
--.1
·POJ5"1-----+------+---_J____
I
-i---+-
I
~
V\
1
1 ~
I
SHELL l~kt-1.S~TfOA/
,
1
{i.2) CtAv•t:> ~r{ 1
1
_ _ _J_~_J______j _______
I I I
J_ __
, PIO'-----~--
I 3 +
I
II
• I
';'
~ P·I 1-----~-;----- _ _ _ J_
I
'-.J I
k
~
~
"
. 'I;:
CJ·~
~
~
.()
O·'/J
Au 6. /{)
155
If
I
I
/:J. I
I
I
/0
8
H6 '"' THE S/2.N!it:ET /NIM<.M'S
0 8 /6 .2+
f!G. . 6.J/
15'6
'4
3
~
~
~
.. 1
""-l
~
t--
"'-....;
~
0
"'::.
\II 0·(
~ -I
:::...
• ia,,.lll. i:'t~l'Tti E.L;~EAJr l'UIJN)
-2 - IC>wl~ ~rU•i& S.tFFirQ.6MC:'
-o
0 ~.J.lj ~.5
·/
~
....... ·J
..~le
,,-...,
""- ·i
~
~
·II-
, /~w../2 ~J..11TI kWflA/1
~ f~1'/0 .f/AJJT£ b1~nle#t'~
~ ·S
:l
1 -b
8EAJb/AI' ~#fEAIT
~16. 6·13 $1.AfPI..'(_ .(IJ~l'P• T#b -'1Ulf£fi l'l.,,r$ ()l./fJI~ P.Altf:"DllJ.f UM 'f_
(j) 11>1/1 "/Esll ~JI.JI TE ELE"f6AJT
(g} £;tpT. /!..£U/,.7!, (&Ea&IA)
~·---
I
I
® -----
;
BIBIJIOGRAF.dY
1 ..
2e Parme , .A. L •
Hyperbolic Paraboloids and other Shells of Double Currature
(Journal of the Sb.·uctural Division. P:cocs. of American
Society of Civil E:J.gineers. Sept. 1956. pp. 1057-1-32).
3. VLASOV, V. Z.
The Basic Differential Equations in the General Theory of
Elastic Shells
(National Advisory Committee for Aeronautics Tech. Memo.
No. 1241).
4. Munro, J.
The linear Analysis of Thin Shallow Shells
(Journal of the Institution of Civil Engineers, July, 1961
VoL l pp. 291··306).
5. Amba:r.tsumyan, S. A.
On the Calculation of Shallow Shell
(National Advisory Committee for Aeronautics Tech. Memo.
no. 1425).
7. Bouma~ A. z.
So:ne Applications of the Bend:i.ng Theoix regarding nouJ)ly
Cured Shells
(P:t:oCo on the Theory of Thin Elastic Shells. Delft Aug .
1959. VP• 202-239).
9o ApelMd K.
Stress Analysis of Translational Shells
(Procs. of the American Society of Civil En~~ineers? ·
Engineex·ing l11echanics Division, ]'sbo 1968. p:p. 111-139)
10. Reissner, E,
On some a.spe".}t of the Theory of Thin Elastic Shells
(Jou.r.!1al of the Boston Society of Civil Engineers, VoL
42 19?5).
11. Vreeden~uygh, G. J.
'I'he Hyperbolic Paraboloid Shell and its Mechanical Properties
(Phillips Teclm:lca~ Review No •. 1 9 V«Jil,., 20, 1958/59)
12. Loo:f,H.W.
Eenvoudige FoiY,'!Jules voo:r de Buigings-Storingen in H"yperschalen
die volgena beschrijvenden zing begrensd.
(Rapport 8-61-3-hr-l Stevin Laboratory, Delft, March 1961)
13~ :Proceedings of the Symposiu.111. on Shell Research
(:CeHt ~ Augast 1961).
(a 'i
• I
RJwe ., R. Eo
,
14~ Jones, Lo L.
'.fosts on a One-tenth Scale Model of a Hyperbo::.ic
Pa~aboloid Shell ~oof
15. Ahuja, E, M.
Stress Distribution :L:n Sh.ells cf Negatitre Garssian Curvature
(Ph .. D Thesis, London Ur.iversity Oct. 1961).
16. i'j.moshe:::.ko 7 S. P.
1
18. 1iargaere, K.
Zu~ Theo~ie der Gekrummten Platte Grosse Fornar.derung
Proceedings of the 1":!.fon Inte:r.l:..s.t:!.onal Congres8 i.'or applied
Mech (1938) pp. 93-101.
/bl
19. Purcher~ A.
13<?.itrag zur ·rheory '.I'ra.gender Jn.::,,0hen
(:Dissertation Tedmische Hoo1.8chule g:rzz, 193l).
20., Tester Kc
Bej_tral? zur Be:cec:hungden Hy}Jerbolishen Pa:caboloidshale
Ingeniew ·· Archid 16 1 (1947/48).
21. Candela, F~
22.. Tottonham, H.
T~1e Anal;;rsis of H;n:ierbolic :Paraboloid Shells
I MeI!lbrane (Hone~1·i;le.=~s) f.:'l'llysis
'i'irnber Dev-elopmer~t Association Lond.on 1957.
23, Flugge, W~
Stresses in Shells
Springer - Verlag 1960 pJi~ 181-184.
25,, Argyris ~ J. IL
EneI·gy Theo-r·ems and Structu!.·al A:::Je.lysis
Butterworths 1960~
~6. A?.:gr:~is J. H.
Hecent advances in natri:z: me,:;hod of Structural analysis
Perg8Jiloll Press 1964.
Elastfo ... plastic :Matrix Displa.c13;:r,2nt Analyc:5.0 of three
Dimensiona.1 Contirn1a Sept., 1965.
Tria:r . gular Elenents vdth linearly varying stra;i..n for the
Matrix Displacement Metl:.od Oct .. 1965.
R~info11 c~d Fields of Triangular Elements witt1 Linearly
Varying Strain; Effect of Initial Strains Nov. 1965
Matrix Displacement Analysis of Anisotropic Shells by
Triangular Elem.e11ts lfov~ 1965
Jourr.al of the huyal Aeronautical Society, 1965.
28. Argyris, J. JL
1foie:T1 E'usela(?,'e lmalysis and tbe EJ.astic Aircraft
:Butte;:worths, London 1963.
30. Turner M. J.
The direct Stiffness method of Structural analysis
l~ga.rd 72 Aachen Sept .. 1)590
31. Meloah R. J.
A stiffness IfiE..trix fo:c the .P..nal:n=ds of Thin Plates in :Rending
Journal AIAA 1960.
32~ Clough, R. W.
The finite element method in pla.vie stress analysis
Proc, Second Cor..f. on EJ.ectronic Computation, A.SaC.E. 1960.
33. Adini, .i.
Analysis of Shell Structures by the finite eleTient method
Ph.D. Dissertation., Univ. of Ua1ifo:;:nin., Berkeley 9 1961.
34. llelosh, R~ J.
Basis for deriYation of matrices fm: the <lirect stiffness
net hod
Journal .A.IA.'\ No .• 1. 1963.
36., Herenikcff 1 A*
$oJ.ution of problernc of e 1.astici ty by the framework method
J. Appl. Mech.~ J.941.
method
(Univeroity of Newcastle - Unun ···· Tyne Rep. No. C/R/42/66).
164. ,I
44. Wilson, L. B.
SoLrLio!"~ of Certain Lar,:-e sets of Eq 1r.+.ions on Pe{;asus
u3ing.~atrjx tlctho~s.
I
ANALYSIS O? A HYPERBOLIC
PARABOLOIDAL SHELL
by
MELVIN L, ~URDORF
A MASTER 1 S REPORT
MASTER OP SCIENCE
Approved by1
11
\:./15
D~o"J,:,_:1- TABLE OF CONTENTS
~ <.n-ts
General Definition
Surface Definition
Conditions of Equilibrium 5
APPLICATION OF STRESS ANALYSIS 16
Surface Definition • ••••••••••••••••••••. , • • • • • • • 16
St re a a Conditions ••••••••• , • • • • • • • • • • • • • • • • • • • • • 18
Secondary Stresses .•.••.••.• , • , •••••••••••••••• , 22
DESIGN EXAMPLE
Shell Design • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 27
Edge Beam Design •••••••••••• , • • • • • • • • • • • • • • • • • • • 29
Detailed Drawing • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 33
CONCLUSIONS • • ••••••••••••• • •••••• •. • • • ••• • ••••• • • •• ,, 35
AllALYSIS OF A HYPERBOLIC
PARABOLOIDAL SHELL
By Melvin L. Burdorr, 1 A. M. ASCI!
SYNOPSIS
INTRODUCTION
General Definition
Surface Definition
Conditions of Equilibrium
S Ibid . , pp . 3-9.
ffil. STRESS RESU...TANTS AND LOADS
z
t
I
~i
)
\FK3.2. STRESSES ACTN3 ON A DFFERENTIAL aEM:NT
8
Txy •
j_t/2 r 7-z
'I' {--)dz, Tyx
{t/2
•
rx-•
'I' {--)dz,
t/2 xy ry -t/2 Yx rx
Qx •
it/2 r 1-z
'I'xz{--)dz, Qy •
it/2 rx-•
'I' {--)dz,
t/2 ry t/2 yz rx
M
x
it/2< (-
• r 1-z
-)zdz, My
f_~/2~ <--lzdz,
•
rx-•
t/2 x r1 t/2 1 rx
Mxy
{t/2
•-
ry-z
'l'xyC--)zdz, Myx •i~/2'f (--)zdz,
rx-•
t/2 r1 t/2 yx rx
l
Tj;p
dp coif• dy (la)
dq co•~· dx (lb)
TxpdPTxcoa<t> dp,
Spdx•~ dx,
co~~
T dx•T COi ~ dx
J'P 7~ '
spdY·~m ~ 41.
12
Typ •T~
YCOS lj>
(2b)
SP • S. (2c)
wzp • wz 1 (4)
cos 'f cos 4>
With the forces acting on the element varying from the
one race to another, equilibrium of forces in the x direction
13
a_!u~ o (5)
ax+ay.
Likewise equilibrium in the 1 direction yields
a...:u + ~. o.
() y <IX
(6)
S dp sin 'f'
and by substitution from Equation• (2c) and (la) gives
{7a)
(7c)
15
The three Equations (5), (6), and (7c) are the basic
membrane theory equations for this case, Using the defined
shell's middle surface, a direct solution or these equations
may be tried. Since we have three "dependent• variables
Txp' Typ, Sp, each depending on two "independent" vari-
ables x, y, the solution becomes very complicated. By
assuming the stress resultants described by a single stress
function F of x, y, instead or Txp• Typ• Sp, the equations
are reduced to one second order equation. The stress
resultants are derived from the "Airy stress funct1on• 9 F by
differentiation, as follows:
Txp ·- a2p
ay2
(Ba)
• a2 '1t'
Typ (8b)
ax2
SP --~
axay
(8c)
(9)
Surface Def1n1t1on
E' c
Streaa Condition•
Likewise,
*. f or c • RZ • (lOa)
(lOb)
z • c~ • chf>i • xy C~al •
Letting k • ~b ,
z • kxy. (11)
)(
a2F
- 2 am k. - wzp
sp -~.
2k
(12)
~-
ax
and
3Sn
n· o.
Substituting these values into Equations (5) and (6) yields
(13)
z
0
(B)
Secondary Stresses
12 Ibid ., p . 16 .
25
~I)
fl
l300.-----.,_._,.-=r---,--,-~-;-~.---,
·~
f .50
~001---+--f-+=:--lr---+~+----t~-;--.
~
~50 1---l--li~::--+~..-t-~-+-~;-~r---i
~
<:;;, 100 1---~~~-f~orl-~;:-t--;--..:t::---i
~
1so~~~~~+:::::::'"i-=:::=-f"'-=:::t:=::::I
~ o0~~LJ:J~~~~~~~~iij.
10 20 30 40
~
(B)
ht • 1 p~~
iO x ). 0 • 010 •
From Pig. 8-a, for ~ • 0.010, and a distance of 5 feet or
more from the corner, it is observed that the secondary
bending moment ii becoming less important.
Pig. 8-b, representing the secondary bending moment
as secondary flexural stress, brings out the importance of
curvature on the magnitude or secondary streaaea, 1 3 For
example, a square shell whose thickness is 3 inches and
~ • 1/5 has a secondary stress in terms or w equal to -o,
at a point 5 feet from the corner. Another square shell
l3 Ibid., p. 16.
27
with the same thickness and at the same point from the
corner, except having ~ • 1/7 has secondary stress in
terms of w equal to 115. From comparing the two values
of stress, the importance of curvature is quite evident.
DESIGN EXAMPLE
Shell Design
sp • :t ~
2k
• :t !'.!.P. ~ • :!:
2 h
72 (15) (15)
2 (3)
+ 2700 l /f
• - b. t.
28
f • ll.22. • ± 75 psi.
c 3(12)
~ -~l . 0.0033 .
For the 3 inch thickness, this gives the L length strip
placed out an x distance 5' - 6" from the corner . For the
D. L. plus L. L. or 72 . 0 psr. the maximum secondary etress
equal a
f
c
• 175 x 72. 0 •
144
a1 • 5 pal
•
H • 2,700 x 15 • 40,500 lb .
30
.
A • .
~Mgg • 2. 025 sq. in .
from which
82 620
Ag • 540 + 20,000 x 0,01 • 111 ' 6 aq, in,
d • 18
111 . 6 • 6 , 2 inches,
~·3•7"aachwd!J
mw "in3"8/ab
c-*6•20~0" -..,.-;1>1
~*6•30~0"
f.._..._lHfi~====~====:E;=rtl..=:J
S!lction A-A
S12ction C·C
&ction 13-B
CONCLUSION
ACKNOllLEDGf'ENT
BIBLIOGRAPHY
T!moshenko, Stephen
Theory of Elasticity, New York, McGraw-Hill, 1934.
Timoahenko, s. and w. Woinows ky-Krieger
Theory of Platea and Shella, New York, McGraw-Hill, 1959.
ANALYSIS OF A HYPERBOLIC
PARABOLOIDAL SHELL
by
MELVIN L, BURDORF
B. s., Kansas State University, 1958
MASTER OF SCIENCE
1963
Approved byr
J.aJor Professor
Many article• have been published on hyperbolic
paraboloidal ahells, However to the knowledge or the
author, no complete a1mplif1ed analysis covering various
loading conditions haa been developed to date.
The shell analysis explained is a result of reviewing
several references using the membrane theory solution with
uniform vertical loading. By the membrane theory 1 no
bending moment is allowed within the shell, therefore the
shell acts mainly by axial forces. Applying this theory,
the basic equations are developed for one quadrant or the
hyperbolic paraboloid. A consideration of secondary
stresses, which cannot be included in the membrane theory
solution, la given. The basic equations and the considera-
tion of secondary stresses are then applied to a practical
design of an inverted, single support reinforced concrete
hyperbolic paraboloid .
A brief detailed drawing is presented showing the
dimensions and reinforcing resulting from the design of
the hyperbolic paraboloid.
'lliE STRESSES DI HYPEROOLIC PAP.AOOLOID
SHEUS USIJ."'(j 'IliE MEMBAANE 'IliEORY
by
MASTER OF SCIE!CE
1965
Approved bys
LO
?..&b<t:,
4 ~ .:
I q"
DJ 7 I TAIU OF CCllTEnS
c. 2.
• • • • • • • • • • ••• • • • • • • • • • • • • • • • • •• 1
•••••••••••• •• ••• • • •• 7
ES FOR A OOFOltlLY DISTRI11JtEO LOnL> • • • • • • • • •• • • 11
. ..
ltUS • • • • •• • • • ··- 19
DESillJ OF 'Di~ OF-OIJ:O VAULT • • • • • • • • • • • • • • • • • • • • • • • 20
lSCUSSlOI OF .. UL1S • • • • • • • • • • • • • • • • • • • • • • • • • • S.3
C<X.LUSIC. • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • Sb
•• • • • ••••• • •••• • •••• • •• • •• •• • SS
by
Syncp•lt
The r1 ten o! th• part 1•1 <11!f t.1•1 a u-,e otr 1 ln d0\bly-
c !4'V stl'U'Jturn 1a prn ed. Ula• ot a tre ~t!on in caleul:at.lng •
bi'tne str• lt given. Eq:atlo:w o! 1tt•u ot •~ ec tel cH• are
d r!..:tld. A atutr o! the e: •?¥1 Ull of th e typ• ot 1tr\l:tUffl 11 • •·
'!he ded;n o! • h)tpefl>l:>llo nbololdall grolntd vault la t.akc."l • en 1: 1•.
DTlQllCTIOI
Th• IC tane ~eery 1n m1ny cNat glveit fi cl »1' lnslght into the 1trc11
dittrlbutlcn ot oubly curved ~Ills. It 1s • good by-pGH to avoid the
•l•borat. c1lculat1on1 •J¥:0Uftttrcd ln Ulln th• lng theory.
1tres1
--
T1JQC»henko2 ct al. stated that 1n
can be neglected,
ru prabl- ot shells th•
fie• ot th• lh•ll• nHd to bl cmalder d. Th• doubly curved 1h1lll hW• been
rate to dcterain• th• 1treHa ln oubly cwrved 1urr1e... That 11 bee• • th•
d1lor•tlon1 In 1w:h 1tructura are very s:all. FltX\11'• 1tru111 are po11lbl•
only 1t 11r;1 dctct•tlon1 oscur. However, bencSlng atHHN •y bl expected
ln 10M ca111 ln th• vie lnlty ot th• eds;• beHI •nd at ••11 zon• nar th•
·~ort.a.
0
·. ~
c<m.lnu loid pdr ip lly by direct st.r HI. He also • ntloned tht pOI•
11bllity of th• ccci.:rre~• o! locellzed ~ing -'*lits n er th• ed~ts d..c to
th• ettect ot th dlspl c ot th• .. rs• but for th• est p•rt the
r.hdlt are tree o! £1~1.tHl tore . •
Uou~cr, the j.lSti!ic11tion o! th• ue;:bran th~ory ;;Di its 1u;~ • 111r~
cl tly C\.tlnCCtcd. with 1a e l tor.s \t\ic:h " se:v• c1:c!ul 1t.uGy. 1h•• tac.tori
the bo· ~conditions, t:;pc of loodlng, n1 rlsc-t.o-
pen ot lh•ll ratio.
n.1 cttcct of. the boundary c0!¥1!tlons on the • '"'In th ory 1JQ
bllam&d tore.a.
Ac:cordl to th• •bav• dl1cusslco, a ;en ral atatu nt can be •411 th•
ane theory l• ap;>liclbl• only 1t th• 1>0t.11n<1ary condltlcns are c 1tibl1
vlth the ccndltl o! quUl.btl • Hovw1:, r tult1 of testl
ln the dctlg ot this part ot th• 1111. Curves that show the He ry
- . nt n ar th• • are pt11 cc1 by th• Portland Cacn Assoc latlcn8.
.y tf!eet th• vsltdlty o! the ne theory. · .The xia~ •train 1s th• c1u11
ot th• ex!ct•nc• ot tht ·aec ry btndlna . ocicnti8. f'or a 11oduate rl•e-to-
\' ' .
•;> n cf ..ii.ii ntlo, th• •f'feet of axlal ·1ttliin1 11 unilport1nt •1¥1 can be
. .; '.
11tt1y ignored. Howw r,, llhcn th!t ratio dee~c111c,, · th• ef!cot c! axial •tnlnl
• 4 I t' • 7
beglnl to acrt e ' dord I .lnt1uenc• 'on tht .b9hav!Ol' ot th• lh•ll· 1h• d ar-
tute !n bthtvlor fr• that lmtuted tw. th• llCllbranc th•ory 1~ •nel~oua to
, • it I •
horlaoftttl COllpOftent dtUeuts II_ the ntlo ot· t!H to •n deer..1t1. With no
. .
rise th'e .horh: 11 : ~~on• d~rtHes to t:ero,, thw th• HCondtf1 bending due .
" .
to U1al :atfl Ina 1pproechtl tht l111plt• baft' bending aOMnt • 1hc Pert.land
+ . • •
c...nt
AHoclaUcn e wt• that "1.tn' th• rlsc-t<>-.pan ot. lhell rat.lo. it on••flfth or
f I .C,
.
aore, the effect ' of ~' extal. 1tralne
..
11 · unlaport&nt. and can ·M 1itdy ignored. In
other vord1, tor 1w:h tlee-t.o-ep1n o! ihdl r1Uo th• Mnding aoaitnt haa •
negligible ettect.
tut.. , that 11, lhdls vith n•gtttive curvature. 'lb•• tw•• ot lhella develop•
atate of 1tre11 i.vorahle to the
'
uH ,ot cC1¥!ret•. 'Ibey .anov re•rklblc res 11• '
• r
tance to cxploeton; ~r~, U~cluake. little HnUUvltY to !o tlon
1ctt1C1Mntl• . 1h~ ar1 •Clft•ic~l d.ua .to ·~dr ruled sur~a \blch •them
"' ~. "' • ·' t • - .,. , '.' .. • .
11ore ... uy. tor•d by straight plink 1'.or.t. Also. th•H lhcll1 give the archt•
~ ' ' .
tect.I I chance to .depart frca th•: conventional prectlee ot beea ind col
fl ~ • ' I
"' • t
building• to aore L._ghwUve •nd gr11<:etul m;>u of, ttru:turu. There!ore,
they ire vlddy !ltvored today.
1hc scope o£ thi1 peptr lnol\d th• ..crlptlcn· ot the surface ot th•
6
boundary conditions end • ••mrl• d•tlgn. A dalgn ot th• Groined Y1ult t11lng
th• ....wane theory •nd in aceordllrv:• vlth th• C.rdele6 Ket.hod vlll bl prff•ntld
r ruent•d in Ftg. 31 by th• grid Un • hn •rd ln' llhicb l"JH provision 1)¥
ne d the generators. tvery point on th• surt;ce y b• c(¥'lsidercd to be th•
intersection ot two • h 11 • coatlln•' ln the surtsce.
Th• equation ot the surface ln blra:ttnguJ.Ji'r coerdinates cen ba deriv
by conatdGrln the bflsic quadrent, HO , shown in f'ig. 1:. In tr!engle HA•A
5.. !.
h I
ot
c. r.
l
cz.Tb'.
z•-
lb XY
• KXY. • • • • • • • • • • •• • • • • • 1
nts th• 1 l ltsi pc:+!lt lhl• aqumt ton o1 th• 11c ~ d grH
,.
/
,. I
~
..,, Cl.. Cl
\ ,.--
,.,.-
Ul
(.., I (
I I \
,~· \
(/\
I
c
.,,
;llJ
)>
l)
m
\
\ .
\
0
m ,. \
..,, \
,. \
z
....
\
0
a m--- --1m,.
I
',, z I
(/l
I
N ,. I
I
~ I
I
I
I
I
I
I
, /
<-\
I
\
\
I
\
I
I /
~~-
\
- --
\
\
9
Fi g . 4 -GEOMETRY
I vy-+~·dy
?Jy
s~Tic
!
.I
I
z I
Fi g.5 - MEMBRANE F 0 RC ES
il
ds2 c ~ • dy~ • • • • • • • • • • 3
The horizontal cc.ponant of the nor 1 1tresse1, Tx, 1s
In order for the projected de ent to havs th• same total strc11 11 th•
actual en•
Cl'
12
Vx • T c •
X COll(i • • • • • • • • • • • • • • b
• •laUat pr • ut• one ctn snov that
Vv•T ~.
" y cos p •••••••••••••• s
Fr Eq. l
~•lY
1> x
Ind
h.xx •
1jY
1htflfcre,
V • T
x x
-f 1+K2XJ-
l+K.2.'f.. • • • • • • • • • • • 6
In. •1•1111'. t
~-~1~~
" . . . .. . . . . . . 7
Since,
thtrc.tou,
S • T. • • • • • • • • • • • • • • • • • • 8
111bti of atrcas per unit 1 ngth • flcUU •1•-
t 11 th• Y-exl1, con1ld1tl th• varl•Uon of th• forcu troa th• n•r
CM to the f'lf cm, I only vertical le11c1 thl lhtllt I ;tltl~ I
'0() Jx +']..I_
;;;Y • o• • • • • • • • • • • • • • •9
Sl•Uarly, qulllbl'lua alo the X-t!xls glva
~v ]1
~. °1)-X. o. • ••••••• • • • • • .10
Therefore, the vertical c onent of T.x per unit lc~h ale the Y-exll
~
Yx •· ~·
S'l•Uarly, th• vertical c~o.nint ct Ty prr nlt ct length along th• X-exls la
()z
v~ ~ 1)V.
1-~z '02z 2
Vx~ + Vy ~ + 2T LL • ·z • ln ..., • • , , • c • • 11
{IX ?J Y'- '15~-Y
Tn• u•• ot • 1tr· s tunctlon3 de!lnlng ~11 three strc11 c nt s vlll f'ac:ll•
lt•t• the solutlcn of the dlf!erenttal au on. This 1tr 1 !'unction, F(x,y),
reducll th• th.tee 1lll\lltln'l Eqs. 9,, 10 end 11 to one equation, the coa-
q. 11, therefore,
lS
of• dc:iubly Cutv 0111 llh•r• th• algsbrale 1olutton of th• ditter t11l
· ~·ri;al pro::c.dut •Y be used. However, In
th• caH ot th• hyperbollc paraboloid shelll, subjected to 1 ltora aurface
1 d ltor.ly diatribe: load • horizontal projection, dlrect lnte-
ration i1 relatively 1~10.
---oz -o . ~ o2
z • KXY'OX • KY,o • KX,o-XW. K'o"X7. o, ...... '-'t""'[;" 7
7},
-2K~ • 'z aln w,
r:~
0 XQY.
Thtrafot•,
T• ~ sin w. • • • • • • • • • • • • • • lJ
For hyperbolic paraboloid ell• havln~ a gr.. t rlao, the dead loe c•n•
not contidet unl!oraty distributed on th• h bontal projected ara.
'nlete!ora, •n re111Q\ tot th• dtl loa t any point ln ter ot Wz should
be duiv •
~ ,ff)~0 ~
c2x • see 'f' •0 x - 1 ittt r. • • • • • • • • 11
av 2_
A
• 1ec-i.r
~ -~y •ji+x.2?.
C/ e e e e e • e • 18
1h t!o:e,
T • ~((i • • • •• • •• • ••• • • • • 22
• •••••• •
Then by 1 tltutl £q. 2h ln Eq. 9 th• tollovltJQ can be •in da
..wj2
• _.s. 2K Y sin2\l+2K2Y cos2v-2K2x c01J t1 d.X
2K 2 {ff
2 £. 2 2
+We l,...,JcOI w(21\2rc v-2K Yl dX:fK ~ln v • ~
·a- f11
1~1'1.iln zeto.
1'11ualng fer convenlenc• that
'lh rcforc,
logcusly,
r w • 1~
1
.Ji w y: c
Vv • ~! C:OI v{cf • ~ X s1Jw log -/
2 2 + 1'2{X). • 26
., " it; 11nv l+KX
Yx • ~ ucs{~:};} t 1(T) • • • • • • • 28
ti
We
Vv • -:;- X log
..
~KY+
_J
.Jl+K2~
fi] + t2(X). • •• • •• 29
We 1ln v
T• 2K • const11nt. • • • • • • • • • )>
1h• econ le aspectl at tbia tw• cf 1tt\lltut• erivcs tr th• f1Ct th•t
it dcvdcp1 • 1t1t• ot 1trc11 favc •bl• to oncrtte, 10 that thin saetl art
anticipated. In 11ao1t all th• 1hdl1 c~tructcd, th• • Uons ted vary tr
tvo la;, ct to f l~t1. Their ru.lcd s~1'aa11 • th• aorc tily to cd
~ atrlight pl• tcr • In 1ddlt1on, If\! hyperbolic p raboloid yp )root
c~otcd ot llU!ti• it1 ot 1laller hyper llow1 the rt•uslng of the 1 •tor•
J• 26 d g. J4 ainutc1.
D lg Dlt11
t•c • ))00 pll.
ts • 20000 p1l., d1!or blrs.
v• 60.oo pat.
Codtt
1h• ICI code (.ACI 316-6.l) •rd th• racaa.tr¥1SUCC'll of th• .ACI c lttH
)J41l IH Ulld.
Prcacdure ot Dc1lgn1
~
._t -_ _7o_~oo
1
\ ...,
PLAN
fv (a)
I
_ £LE V~TION
t
(bJ
Sttp la
y .. -
.. • • • • • • • • • • • • • • • • 33
•ln f
a• v-n •V-L tin f• ........... .34
t 2•t+ii •••••••••••••• • 3$
Frca Eqs. 33 1 34 and JS, wa obtain
)!
v+dtan 2
y. - -• • • • • • • • • • • • • • • 36
2 1in f
A110,
v •n •L
1ln 2"• X• COi ¥• • • • • •• • • • • ••~
Ft • JJ, .34, 3S, J6 and 37 lt can be lhown that
utan r .,,,
x. 2 l JI. • • • • • • • • • • • • • • • .)8
•n2
Col 1 •nd 2 or Tsblt 1 lhow th• valu ct X •Di Y of th• ?>ounda•ty
P(><, Y)
y
v ··:
ltlere1
v ·-fll
• 11n v ll+K2Y2
c• iwc x 11n2w
2)
!..!-- • !f.
•1n • n tt•til) . .JL..
iln-A<
Fr \blch Is 11
1ln (8~)
"2 • •ln D<. •
1tn
11> • 1x sin rx.
2
e
+ 2T s!nA!ln
sin 0c.
v.aj + Ty tln2otn"'
CB-~)
• • • • • • • 39
26
K2y2
t-olnt
l
2
r
0.0000
2.8000
x y
2
0.0000
2.8o0o
l
0.0000
E
0.0000
7.8400
y2
r:
0.0000
7.8400
K2x2
6
0.0000
0.0024
1
0.0000
o.o02li
- 0.1363
) s.6000 s.6ooo - 0.5451 )1 • .)600 31.3600 0 .009$ 0.0095
4 8.LOOo 8.1&000 - 1.2266 70.$600 70.$600 0.0213 0.0213
s 11.2000 u.2000 - 2.1806 125.4400 12s.hhoo 0.0319 0.0379
6 14.0000 14.0000 - 3.4071 196.0000 196.0000 0.0592 0 .0)92
7 16.6000 16.8000 - 4.9063 282.2LOO 282.2400 o.OSSJ 0.08.SJ
a 19.6o00 19.6000 - 6.6700 .334.1600 )8h.1600 0.1161 0.1161
9 16.8000 22.4000 - 6.S417 282 .21.00 $01.7600 o.08Sl 0.1516
10 14.oooo 2s.2000 - 6.1329 196.00CO 6)S.0400 0.0592 0.1919
11 11.2000 28.0000 ... s.hSJ.4 12$.4400 784.0000 0.0379 0.2,369
12 8.4000 )J.8000 - 4.4974 70.;600 948.61!00 0.0213 0.2867
1.3 S.60oo 33.6000 - J.2709 31.;soo 1128.9600 o.OQ9S 0.3411
lh 2.8000 J6.4ooo - 117717 7.4800 1324.9600 0.0024 o.4004
lS 0.0000 39.2000 0.0000 0.0000 1$36.6400 0.0000 o.h~.3
16 - 2.aooo 42.0000 2.0!Ji3 7.4aoo 1764.0000 0.0024 o.S.330
17 - s.60oo 44.8000 4•.3611 31 •.3600 2007.0400 0.009$ o.606S
18 - a.4000 47.600\) 6.9So6 10.S6oo 226.$.7600 0.0213 o.6847
19 -11.2000 so.Loco 9.6126 12.$.4400 2S40.16oo 0.0379 0.7676
20 -14.oooo 53.2000 12.9472 196.0000 28)0.2400 0.0.)92 o.esso
21 -16.Booo S6.oooo 16.3$43 282.2400 31.36.0000 0.0653 0.9476
22 -19.6ooo SB.Booo 20.0000 .384.1600 JJ67.1:400 0.1161 1.1608
23 -18.1000 56.oooo 16.20h4 349.6900 31)6.0000 0.1057 0.9476
24 -11.1800 S3.2000 16.lt4.30 316.8400 2830.0000 0.09$7 o.8550
2s ·16.Booo so.0400 1417184 282.2hOO 2540.1600 o.08S3 0.7676
26 -lS.6800 47.6000 13.1381 252.8000 2265.7600 0.0764 o.6847
27 .14.esoo h4.8ooo 11.S6S1 220.$200 2007.0400 o.o666 o.6o6S
28 •14.0200 42.0000 10.2300 196.$800 176h.oooo 0.0594 o.S.330
29 .13.0800 39.2000 a.9127 171.0900 15)6.6400 0.0517 o.h64.3
X> -12.1800 36.bOOO 1.1069 lh8.,3SOO 132h.'600 o.0448 o.h004
31 -u.2100 33.6000 6.S828 127.0700 1128.9600 0.0.384 0.3411
32 -10 •.J)OO )).8000 S.Slh6 106.0900 948.6400 0.0321 0.286?
.33 - 9.3600 2a.oooo 4.5sss 87.6100 78h.OOOO 0.026; 0.2369
34 - 6.J600 2s.2000 3.7019 11.4000 63$.0400 0.0216 0.1919
3S - 1.s100 22.4000 2.9244 S6.booo so1.160o 0.0170 o.1S16
J6 - 6.saoo 19.6000 2.2419 43.3000 384.1600 0.0131 0.1161
37
38
- s.6lm
- 4.7200
16.8000
14.0000
1.6SS9
1.1488
31.8100
22.2800
282.2400
196.0000
0.0096
0.0067
o.OSS3
o.os92
39 - 3.7700 11.2000 0.7362 14.2100 12.).hhOO 0.0043 0.0379
40 - 2.8400 e.uooo o.4116 8.0700 10.S6oo 0.0024 0.0213
41 - 1.9100 s.6ooo o.16S8 3.6500 31.3600 0.0011 o.oo9S
h2 - 0·2;!22 2.eooo O.O!i2~ o.8600 z.el.ioo O.OOOJ o.002h
27
Table 1. (Cont.)
Table 1. (cont.)
2 I 2
l+K2 ~ l+K2y'l •in v l+K2.x2 sin v l+K Y2 ccs w 2' sin w
16 18 21
Point
1
2
o.&s
o.&>20
1.0000
1.0020
i.Moo
l.0020
o.tiooo
o.eo16
12
0.6000
0.8016
20
c.4boo
o.4812
0.0000
0.8960
3 o.eoso 1.0040 J.• 0040 o.6o.32 o.eo32 o.4829 1. 7920
4 0.6110 1.0100 1.0100 o.eoeo o.eoeo o.4866 2.6880
s6 o.8180
o.8,J>O
1.01so 1.0150
1.0290
0.8120
o.e232
0.6120
o.82.32
0.4908
o.49eo
.3.SeliO
4.4500
1.0290
7 o.e420 1.0410 1.0410 o.8.328 o.8326 0.$158 s.J16o
8 o.ssoo 1.0550 1.0$50 o.e.s24 o.6.$2h o.Sl.36 t.2720
9 o,8600 1.0410 1.0110 0.8238 o.8408 o.Sl60 s.316o
10 0.6140 1.029'J 1.09$0 0.8232 o.s44o o.s244 4.4800
11 0.9020 1.01.so 1.1100 0.8120 o.aaso o.s411 3.,S8Lo
12 0.9290 1.0100 1.1320 o.soao 0.9356 o.SS73 2.6880
1.3 0.9600 1.001J0 1.1sao o.6032 o.926h o.s1s9 i.7920
14 1.0000 1.0020 1.1610 0.6016 0.9Lh8 0.6000 0.6960
1S 1.0500 1.0000 1.2090 o.eooo 0.9672 0.6299 0.0000
16 1.1000 1.0020 1.2350 0.0016 0.9680 0.6599 -o.8960
17 1.1610 1.0040 1.26.30 0.8032 1.0104 o.696S -1.7920
18 1.2210 1.0100 1.2990 0.8080 1.0392 o.132s -2.6880
19 1.2850 1.01.so 1.3300 0.8120 1.0640 0.1109 -J.S840
20 1.3.)00 1.0290 1.3600 o. 232 1.0680 0.8099 -4.4800
21 l.4l$O 1.0410 1.3950 o.8328 1.1160 o.6489 -s.3760
22 1.446() 1.0780 1.4720 o.BS2h 1.19.36 o.867S -6.2720
2.3 1.43ao 1.049() i.3950 0.8392 l.116o 0.8627 -!).93l:C
24
25
l.JSfO
1.3)6o
l.OhSO
1.0410
1•.)600
1.J))O
0.8.360
o.8.328
1.0880
0.0640
o.a13s -s .6896
0.801.$ .5.3760
26 i.2940 1.03)0 1.2990 0.8280 1.0392 o.7763 -5 . 0016
21 1.2440 1.0:330 1.26.x> o.8264 1.0104 0.71£6.3 -4.7520
28 1.2010 1.0280 1.2.JSO 0.6224 0.9880 0.1205 -4.1'864
29 1.1.$90 l.:'\2)) 1.2090 0.8184 0.9672 o.69)3 -4.18$6
)J 1.1160 1.0200 1.1810 o.816o Q.9448 o.669) -J.8976
31 l.076o 1.0180 1.isao o.8144 0.9264 o.6h55 •.3.606li
32 1.0370 1.oiso 1.1320 0.8120 0.9356 0.6221 -J.29£0
33 0.9760 1.0120 1.1100 0.6096 o.eeeo o.S85S -2.9952
34 0.9660 1.0100 1.09So o.808o o.6Wio o.S79S -2.7040
J.5 0.9340 1.<X:>70 1.0110 o.BoS6 o.8408 o.S603 -2.1.io32
J6 0.9040 1.0040 1.osso 0.8032 o.e524 o.S423 -2.1056
31 0.8770 1.00~ 1.ouo 0.8024 0.8320 o.s262 -1.8048
.38 o.6S60 1.0020 1.0290 0.8016 0.6232 o.Sl.36 -1.5104
39 o.8.)So 1.0010 1.0150 0.8008 o.a120 o.so16 -1.2064
40 0.0200 1.0000 1.0100 0.8000 o.80&> 0.4920 -0.9088
41 0.6090 1.0000 1.0040 o.eooo o.eo32 o.46S4 -0.6112
42 0.80~ 1.0000 1.0020 0.8000 0.8016 0.4818 -o.22z6
29
Table 1. (Cont.)
!2 sln2w c
A IL D ~ .12:iW2.
? o!nt
l
2;:
0.0000
2j
0.0000
"!1
1.0000
B:
0.0000
26
l.0000
21
0.0000
2(j
0.0000
2 o.09(;o o.oa96 0.9761 o.. os90 0.9i61 -0.0243 ...0.0243
.3 l. 7920 0.1792 0.9538 0.1792 0.9533 -0.0471 .0.0471
4 2.6680 o.~688 0 .. 9314 0.2688 0.9314 -0.0710 ..0.0710
s6 3.$840
4.4eoo
o.J.$84
o.4480
o.911Ji
o.e911
o.35a4
o.4480
o.. 91lh
0.8911
-0.0932
-0.115h
-0.0932
.0.1154
1 5.3100 0.5376 o.8708 0.5376 o.a1oa -0.1.383 -0.1383
a 6.2720 0.6212 o.84hJ 0.6272 o.8443 -0.1707 -0.1707
9 1.1680 0.1168 o.9s34 0.5376 0.7839 -o.0481 -0.2h33
10 8.0640 0.6064 1.oses o.4h8o 0.7069 o.osa2 -0.3467
11 6.96oO o.8960 i.12SJ 0.3554 o.6S.S3 0.1088 -o.4231
12 9.86)0 o.9865 l.1802 0.2688 o.59SS 0.16;$ -0.5184
13 10.7520 1.0752 1.3095 0.1792 0.5h07 0.2700 -o.6149
14 11.6!:80 l.16h8 1.4086 0.0896 0.4946 o.34.36 -0.7032
15 12.sW.Lo 1.2544 1.5083 0.0000 0.3999 0.4121 -0.9163
16 u.W+oo 1•.3440 1.606o -0.01.!96 0.4250 0.4731 -o.8551
17 11.i.3.360 2.4336 2.828S -0.1792 o.4031 o.6o03 -0.9088
18 25.2320 1 •.52.32 1.7065 -0.2688 0.3787 o.s344 -0.9702
19 16.1280 1.6128 1.6847 -o.35a4 0.3597 0.6337 -1.0217
20 11.02ko 1.7024 2.9744 -o.ueo 0.))92 o.6eos -1.0817
21 17.9WO 1.7920 2.0.s.30 -0.5376 0.3198 0.7193 -1.1394
22 18.6160 1.8516 2.0106 .0.6272 o.2s1s 0.6936 ..J..3567
23 11.9200 1.7920 2.1032 ..o.$984 0.3211 o.6981 .1.13JS
24 17.0240 1.7024 2.2256 41().$689 0.3309 0.7990 -1.2347
2s 16.1280 1.6128 2 .031.aJ ..o.SJ76 0.3493 0.1102 .1.0S26
26 1s.2320 1.s232 1.9931 -o.$082 0..5961 0.6881 -1.0J.k7
27 l!i.))50 l.h3J6 1.9488 -0.4752 0.$714 o.6678 -0.9795
28 l~-4400 1.,3440 2.9009 -o.h486 0.6418 o.6416 ..0.9288
29 12.s440 1.2544 1.8538 -o.4186 o.61S3 o.6153 -o.8734
30 11.6480 1.1648 1.6070 -o •.3898 0..$91) o.s913 -o.828.S
.31 10.7.520 1.01s2 1.16.54 -0.3606 0.6231 0.$678 ..o.1as.;
32 9.86SO 0.986.S 1.6958 .0.3296 o.S,306 0.$.306 ..0.7318
.33 e.9600 o.8960 1.6104 -0.2995 o.h762 o.4162 -0.7339
34 ~.8064 o.8806 l.S798 -0.2704 o.b574 o.1674 -o.61S2
3S 7.1680 0.7168 1.)173 -o.2403 o.4187 o.4187 -o.SS34
)6 6.2720 0.6272 1.h1S7 -o.21o6 0.3471 0.3471 -o.4894
37 s.3160 0.5376 1.3830 -o.1eos 0 •.3241 0.3241 -0.4232
)a 4.4aoo o.Wiso 2.3178 -0.1510 0.27.)2 o.21s2 -o.3S81
39 J.Sh8o O.JS48 2.2s20 -o.12o6 0.22!6 0.2245 -o.2890
40 2.6880 0.2688 1.1861 ..0.0909 0.1697 0.1697 -0.2200
41 2.7920 0.1792 l.12h2 -0.0611 0.1129 0.1129 -o •.lh91
h2 o.6960 0.0896 1.04$9 -0.0298 0.0438 o.0438 -0.0693
.30
T""..l:>le l . {Coiv.:l.)
-
~.].cs C lgs D
.)2 ..
T .~ft. T COlf 1'
-~
vx ';</ft.:,.
lit.~~~·
?oi t :i2 ___}} ,?:!: ...... ~
l,.. o.oJOO 0.0000 • 2.29tu l •.;l93 i.3793 1.379.3
-0.0022 -0.0022 2.)016 1•.3011 i.3;69 1.37e9
'3 -o.ooas -o.ooes 2.3100 1.,3561 i.:,776 1.3776
u -0.0191 .JJ.0191 2.3268 le.35'73 l.J78~ 1.3782
s ..v.o.358 -0.03$0 2.3449
2 • .)8:;5
l.4o69 1.3711 1.3111
6 ..o.0;16 -o.OS16 1.4.;ol 1.;7s5 i.,;78S
7 -0.0744 ..,.01U 2.4166 l.b512 1.;;768 l • .)768
e -0.1011 -0.1011 2.4600 1.4160 1.)689 1.)689
~ -o.o.344 -0.1~2 2.4658 1.479S 1.Lw.s1 1.3493
10 0.0388 -o.1SS4 2 •.$160 1.5100 1.$~96 1.3$54
11 o.om -o.1s1~ 2.$6~8 1.55;;9 1.6514 1.40~7
12 0.163.) -0.1393 2.6610 1.))66 1.7601 1.4573
l3 o.:scs -0.1102 2.1ss1 1.6552 1.9457 1.5450
14 0.4000 .o628 2.b1$ 1.72.41 2.1241 1.6613
1s 0.5160 0.0000 3.0189 1.81lJ 2.3273 1.8113
16 o.6JS1 0.0767 3.1600 1.8$60 2.5311 1.9727
17 0.6611 0.1629 3•.)4£;1 1.%41 2.76S2 2.0680
18 o.e100 0.2616 J.$011 2.1007 2.9107 Z.3625
19 1.0196 0.3660 3.68)9 2.2139 3.2335 2.5799
20 1.1.$8.S o.4850 3.6788 2.3273 3.4ess 2.6123
~1 1.2810 0•6221 4.06.)h 2.4392 3.7202 J.0613
22 l.))4S o.as1e 4.1soo 2.4700 J.7945 3.3hl8
~) 1.2.Sll o.6686 4.1.354 2.4812 3.7323 J.1498
.:4 i.;oos o.701e 3.68)7 2.J3Ja .).694.3 3.0356
2S l.l.484 o.S6SS 3.S.)C9 2.~21 J•4.SCS 2.8676
26 1.0499 0•.5160 3.;210 2.2326 J.2825 2.?486
27 0.9)99 o.4659 3.s1a9 2.lh?.3 J.1072 2.6132
28 o.8643 o.4166 J.16co 2.0700 2.9~3 2.4866
29 0.7718 o.36.58 3•.)2.).} 1.99,52 2.7670 2.3610
.:0 o.6~ 0.3229 J.j)JO 1.99ao 2.6800 2.3209
.31 0.6182 o.2a2e 3.oass 1.CSl3 2.h69S 2.1.341
32 o.s2.;s o.241e 2.s191 1.7578 2.3113 2.0.:96
33 o.4267 0.219$ 2.eooo 1.6800 2.1067 1.8995
34 o.403S 0.1661 2.7689 1.661.3 2.o648 1.8374
JS o•.J>OO 0.1338 2.679'.1 1.6o77 1.9077 1.741)
J6 0.2810 0.1029 2.5942 1.)$65 1.837) 1.6594
37 0.1731 0.0754 2.s1a1 1.s109 1.6840 1 •.$863
38 0.1234 o.osJ.i2 2·!:579 1.4747 1.!)981 l.$289
39 0.0796 0.0349 2.4000 1.4400 1.;196 1.4749
4C> o.o4S6 0.0200 2.1)00 1.hloo l.hS.56 1.4300
41 0.0203 0.0091 2.3199 1.3919 1.1'122 l.1&010
g2 O.OOJ2 0.002J 2.3)00 2.~00 l·J§J2 1.~2J
31
• ere
\fb 11 t? • ptoJ Uon of. ~11)?, m th• XY49hnfl.
~b 1• the Ulgon tr!cal tang t or i.hc ,..mle !orMd "1 th• •trdG}lt lin•
repr s nt th• s tlo vlt.h the XY~lAne, ~. •·•
A 1lgn c 9'\U tor the prc4uct ltY <::o should bt cptc her. In rdtr to
avoid alsta ••• The velue KY · 11 podtiv. en both sldca ot •~1• v are
""-
"X • .. ,
,.. 1 !p (,!'.-es,l. I
sin J1 t • • • • • • • • • • • 42
't): • 'I"
•T ' n fl
(§::otj • • • • • • • • • • •• h.l
U, wlth llb 1 to 116 ~ · iU .34 !nut. , tt
valu~ •r~ shown in ol 12 of Ttible 2.
~in::~~~
KY.g'b+
Lb. ~ KY•li-
6
cos "aa.
Point 2
0 • lol 0.0000 O.OOYJ .oooo i.o.s 0 0 .. 0000 -o •. 7
9 1 .. 11.i; o.osos 0.0026 1 .. 0011 1.0722 -0 .. 0197
...().0716
..t>.4669
-o.s1es
10 i.12s1 0.16.)S 0 ..0261 1.0125 l•lOSJ
11 i.1266 0.2726 0.0743 i.oJ6S l.l.SOS ..0.1327 ..0.5799
12 l•lbJl 0 • .)816 o.. US6 i.0703 1 .. 2116 -o.20hl -o.6$1)
l.) i.1626 o.h906 0.2407 l .. llhl l.29'Jl ...o.2a6$ -0.7338
14 1.16.34 o.sm o.»96 1.16S6 1.3766 -0.3794
-o.h629
-o.8266
lS 1•2090 0.7087 o.so22 1.22s1 1.4819 .0.9)Jl
16 1.237s o.8177 o.6656 1112916 l.S9Sl -o.sno -1.0442
17 1.2681 0.9267 o.esss 1.)634 _i.7220 ·<>.7217 -1.1669
18 1.3120 1.01)8 1.0729 1.4397 1.8702 -o.es10 -1.))42
19 1113499 hllc48 1.3106 1.s201 2.0217 -1.00~ -1.4s02
20 1.mh 1.2SJ8 1.s120 1.6037 2.1810 -1.1595 -1.6061
21 1•4S22 1.)628 1.es12 1.690) 21113$82 -1.3267 -1.7739
22 1.$868 1.12 , 2.1266 h768l 21o6029 -1.L90? -1.9317
'hble 2. (Cent•)
Tabl& ~. (C .}
!"' ~-~
-!.~ ~i~
0.9074 1.005 0.997 • f)S -2.. ·Sb 1.00JO
9 0.9128 1.0222 0.9780 -2.s~22 -2.4158 0.9754
10 0.920$ i.os2s 0.9501 -2.6152 -2.35$5 0.9321
11 0.9.)J 1.0784 0.9273 -2.7$'28 -2.4015 D.9lh4
12 0.9331 1.1067 0.903S •2.~9 -2.4042 0.8922
13 0.9339 1.1.3S4 o.6807 -3.1~2 -2.la96 o.8670
14 0.9329 1.16,?> o.859S -3.3433 -2.h69 o.e4e4
15 0.9312 1.1960 o.a,;61 -).6!06 -2.S2hl 0.6271
16 0.929~ 1.2294 o.Bl.34 -J.68li9 -2.5703 o.s113
17 0.9210 1.2S!il 0.7973 ..Q.169S -2.6631 0.7949
18 0.9097 1.2693 0.7878 ~.Lh39 -2.7582 o.111S
19 o.C989 1.2902 0.77)1 ..Q.7607 -2.8600 0.16.32
20 o.e909 1.31-,s o.1S90 -.s.1103 -2.91&40 0.7S66
21 o.a774 l.J.316 0.7$10 -$.hl.3$ •J.OS31 0.7462
22 o.84 2 1.222~ o.zz~ :2·~26 ·J· 2 o.z~2J
Table 2. (Coi¥:1.)
l+l
~· Av)/tt. llYyd{ft.
,. +k
2I
-J.~77
n I
22 't
a
9
i.oooo
i.0299
-~.w;~
-2. 661
-2.~M
-2.L811
-J.ih
.3.9112 .3.e.301.i
10 i.062S ..2.~7S -2.s;ea .4.0111 -).8935
11 1.09)5 ~.ssn -2.626) .1~.2os1 -4.0290
12 1.120 -2.6274 -2.69~6 -1..i.)87S -L.1$19
-2.f.C~J .-!i.6613 ~ •.;uu
~
1.1)34 -2.71$6
1.17 -2.6J6S -2.9109 -li.9606 -la.s122
l5 1.2090 -2.9863 •J.OS16 ,.3136 -h.8629
16 1.232s •3.1Sl8 •J.1679 -S.6f-29 ,.lho6
17 1.2seo -J.3.JX> •).,1,;02 -6.09S2 ,.h182
18 1.2861 •J.hSS1 •J.S47l -6.36$8 ..;.909
19 2.3103 •.3.633b •J.747S -6.6669 -6.3274
20 1.3217 -J.866$ -3.e911 -7.3$23 -6.70~
21 1.3400 -4.0J9S -4.0911 -7.7$97 -1.1s2
22 ·~2 -J·SZQ - ·.ltl22 -z.z212 ·Z·Z.lQ
Step hs
Ue l E.qs. 6 and 11 th• valu.1 of Vx- • Vy are t 22.
1heH ar• own ln colmns 22 1 23 cf Table 2.
lh• lues of Yx v1 shown ln col 33 • 34, Ta lt 1 tor point•
-22, ire th• UIUllpU that f1(Y) •ir.1 f2(X) are l to ctto.
Table 2.
Sttp 61
cncs, 11 lt ICh generator were • tie or 1trut. Such tren1tora1t1on v111 pro-
d c, ot c:ours1, •ltcratl ln th• •t•t• ot 1tres11s at the lntcdOl' points ot
the 1utfk•• Hovevtr, thtH •ltcratl are lnalgnlticant 1in:1 tb• l>Oul¥lary
point• i;overn tht dttl;n.
Accord! 1y, t.'°\1 atr•ssca at th• tr•
.3S
tra formed to the groin, ;>olnta 22•42•1. Th• values obtained frc this step
ere shoa-n in colwans 1 and 2 of T;ble 3.
Step 7a
Th• aodl!lad values of V~ and Vy at the groin, shown ln colmns 3 and
Sttp 91
As in step J, V1lu11 of 1tresH1 along th• other boundary line, 22-42•1,
need to be lnvcstigattd. This can be ecc Ulhed by using Eq. 41, with Wb
equal to 71 deg. and 34 minutes in ordH to !11'¥1 th• angle fi tor th• point•
22-42•1. then, by ualr.g Eqs. 39 and hO, V1lu11 of TiJ and Sb can b• obtained.
The necusar,y ca.putations tor this step are shown in colu:uns 9 to 41 of Table
I • interval
C • th horizontal projection of th• interval, t.e., 2.357.
~Vyk/tt.
Po 2
2 .7.72 -1. JJ ·J· i 0 • o. J2J • •
23 ·1·1S91 -1.sJ&> -b.0274 ~.)862 o.7S20 1.329 -S.35S6
24 .7.JS23 ..7.Jh$2 -J.6SBO -Li. )'.)96 0~7684 1.))14 -lJ.760S
2s "'6.8669 -1.1S2h -3.3164 ..u.2~e tJ.7880 1.2690 -1J.2oes
26 -6.365 -7.0028 •J.08)) -ls.2Sla 0.7968 1.2ss1 -J.869
27 -6.09S2 -6.8S31 -2.96eo -1i.2399 o.8166 1.221£6 .3.6$91
28 ,.6829 -6.70.Jh -2.7466 -1.i.2168 o.8))4 1.2QL3 -J.)101
29 -s.3136 -6.5781 -2.Sh66 -h.2171 o.Bh41 1.1838 .3.0147
l> -b.9606 -6.1621 -2.2726 -h.1)16 o.66.37 1.1578 -2.6312
31 ~.661) -6.327b .1918 -u.1933 o.a s2 1.1297 -2.4761
.32 -L.:;a1s -6.1882 .0762 •h.1Se6 o.e91a3 1.11~ -2.3216
33 -h.20s1 -6.0090 -2.0964 -h.1L9$ 0.9036 1.0988 -2.30S7
1l -4.0171 -s.9098 -1.9S2l -ll.0724 0.9266 1.0792 -2.1069
» .3.9112 -s.1L6o -4.0035 -4.()()4S 0.91:02 1.0636 -2.1.J>9
J6 -J.8341.& -s.sa21 •1.9969 •J.9227 0.9524 1.(1.)02 ~.0971
J7 .3.9112 -?.b182 -2.2272 •J.832.S o.964h 1.0369 -2.))94
38 -h.0171 .5.3257 -2~190 -:h1968 0.9725 1.0279 -2.486)
y; .2os1 -s.2331 .68.SS .3.15e2 0.9833 1.0170 -2.7312
40 -4.JS?S -s.14o6 -2.9319 .3.7106 o.m1 1.0100 -2.9612
41 .6613 -S.01460 •J.2h91 .3.6470 0.9970 1.00.l)..3.2se
IJ2 -h.9606 -b.sisss •J.S767 -J.h806 0.9980 1.0020 •J.S8~
1 -~·Jl~ •.8622 -~·~J ·J·k ~ !·0000 .oooo -J·2lil
J7
Table 3. (Cont.)
l.~868
~
o.,gu -!1
KYffb
-0.1156
-
12
o.5115 l.~~12
U:• l 4
1~
l.8457
2.) -3.2999 1.4634 0.7499 -o.7289 o.S602 1.2499 1.7411
24 -3.311.; l.4212 o.734S -o.677S 0.)399 1.2400 1.6892
2s -3.3764 l •.37$2 0.612.s -o.5690 o.4S23 l.2021 l.604.S
26 -3.3897 l.34L5 0.6700 -O.S.$49 o.Uh69 1.2008 1.$588
27 -J.462) l.)J67 o.S6So -o.4398 0.3202 l.1496 l.4526
28 -J.S016 i.2727 o.5632 -o.hll7 o.316o l.1424 1.4148
29 -3.$622 l.2.)89 o•.;124 -o.,3465 0.2628 1.1269 l.,3647
:30 -J.5686 l.2046 o.1.i1as -0 •.3024 0.2297 1.1066 1..3099
.31 .3.7119 l.1707 o.45h6 -0.261:) o.2o63 1.0979 1.2626
32 -3.7190 l.1s20 o.Lo12 -0.2207 0.1678 1.0807 1.2266
33 -3.7495 1.12.;3 0.,3621 -0.1761 0.1316 1.0641 1.1833
34 -J.773$ 1.1001 0.3311 -0.lhSl 0.1096 1.os34 l.1462
35 -J.7650 l.01BS 0.2900 -0.1135 0.0841 l .OlJ.38 2.1179
J6 -J. 1359 1.0612 0.2497 -0.0848 0.0623 l.0327 1.0916
31 -J.6961 1.0431 0.2160 -0.0646 o.4660 1.0212 1.0620
38 -J.693' l.0,320 o.11s1 -o.0426 o.o.306 1.01)8 1.0463
.39 -J.4000 1.0190 0.1362 -0.0267 0.018,$ 1.0092 1.0274
~ -J.6739 1.0100 o.09ao -o.ol.45 0.0096 l.0048 1.0148
41 -J.6361 1.00~ 0.0591 -o.oose 0.0034 1.0019 1.0049
L2 .3.4736 1.0020 0.0192 -0.0009 0.0037 1.0018 l.0038
1 -J.~806 1.0000 0.0000 -0.0000 0.0000 1.0000 1.0000
Tabb .3• (C .)
...
zfxr-tcos v K~q b 'fCOI Vb ~ • !!.!l• D .• t.ln. £!.a•l'!!n.
~ ;r- 4!-~
ol
22 o.~17 -o.ei84 o.HS6
COi
:o.!fss d {>(
2
ro
10414L 1
2J•J2
23 o.28JS -o.h127 0.1937 -0.2.)68 78-50 103-42 2h-$2
24 0.3133 -o.3613 0.2204 -0.2139 77-16 ~02-21 2S-OS
2S 0.3459 -0.2726 o.2s1s -0.1700 7S-26 99-48 24-22
26 0.371.) -0.2367 0.2762 -o.1SJ2 13-S1 90-li? 2h-S2
27 0.3989 -0.12.36 o.JOSJ -o.oeso 72•13 94-5.3 22-40
28 0.4220 -o.o9SS 0.3316 -o.067$ 10-38 93-Sl 2.3•13
29 o.44li1 -0.0323 o.35e9 -0.0236 68-58 91-21 22-2J
):> o.!:6S9 0.01.38 O.)e68 o.01os 67..J;.h 8'~ :2-10
31 o.h849 o.os11 0.4lh2 0.0409 6S·32 81•39 22-07
32 o.so.34 o.o9SS O.h370 0.0100 64-0S S•.32 21-27
JJ o.s202 0.11+01 o.4623 0.1181 62-.26 eJ-n 2o-45
l4 o.SJSl 0.1711 0.4660 0.1492 6o-SS 81-2S 20-)J
JS o.5487 0.2027 0.5088 0.1820 S9•2S ?9•30 20-0S
)S o.S6o6 0.2314 o.s2a3 o.2us ;s-01 77-l.i7 19~0
37 o.s112 o.2S16 0.5476 0.2,368 56-48 76-18 19-.30
JS o.S797 o.21lS o.5617 0.2617 SS-So 7h-SO 19-00
3~ o.sa10 o.269S 0.5761 0.2820 si.-u9 73•37 18-48
kO o.S926 0.))17 o.S867 0.296S Sk-ok 72-47 !8-lil
hl o.S966 0.3104 o.s9ue 0.3091 53·.JO 72-00 l8•JO
h2 o.S992 0.3153 o.s9 0.31)8 Sl-16 71-Ul 18-27
1 o~~ O.Jl62 0 ·222.2 o.~162 2~ z1-~ lS-26
T&bl l• (C ->
l l /){_
Point 1
22 20 ~ o. J o. o., o. 0.1 2
23 207-2.b 128-34 0.9511 0.9716 0.9442 o.h20S 0~1768
2li 204~ 127-26 o.97S4 0.9769 o.9S4S o.~239 o•.1796
2s 199-36 l2S•J9 0.9679 o•.96S4 0.9712 o•.4126 0.1701
26 197• 12h-h1 0.9610 0.9882 0.9762 o•. 20$ 0.176
27 169-46 117•33 o.ss22 0.9964 0.9910 o.38S4 o•. t46S
23 187-.42 116-0Z. 0.9434 o.wn o.99so 0 •.3942 o.1SS3
29 162-l,2 113-44 0.9334 0.9997 0.9990 o•.3808 o.1hS1
.» 178 111·.34 0.9221 0.9999 0.9999 0.3773 o.1426
.31 17S•18 109-L6 0.9102 0.9992 0.99 0.376.S 0.1416
32 171-oh 106-59 o.en 0.9910 0.9940 o.36S7 0.1337
33 166-26 103-SS 0.8861 0.9930 o.9e6o o.JS4J 0.12.se
3b 162-?0 101-SS o.e139 0.9888 0.9781 0.)$02 0 •.1226
.3S 159-00 99•.lS o.&)09 0.9832 0.9668 o.3434 0.1179
)5 lSS· 96-27 0.8491 0.9773 o.9S41 0.3)66 0.1132
37 152·36 SS-h o.8366 o.971s 0.941 O.JJ.)8 0.1111
38 149-40 9.3-50 0.8274 o.96S2 0.9316 0.3256 o.1os9
3? 147-14 92-2S 0.8173 o.9$94 0.9200 0.3223 0.1037
ha 116-.34 91•)) 0.6077 o.9ss2 0.9124 0.3209 0.10.30
bl "44-00 90-ll o.so39 o.9!)11 0.9016 0.3173 0.1008
li2 143-26 90•10 o.eo14 0.9h9S 0.9015 0•.316$ 0.0999
1 II ~ 9Q=OQ 2· 0.2!i§l o.am O.J162 0.0228
Tlhl• 3• (Cont.)
Tabl1 l• (C 1.)
Tb
iol
22 -o. 19 .2 • -1.
-el IX
LO
-2.311 - • 7
2
-2.)400 ..3 • .3
23 -o.s1e9 1.249S ).2.)65 -1.221 -2.1752 -J.26U -2.2131 .3.3)6.)
24 .o.S91'8 0.9947 3.08?.3 -1.2699 -1.99b9 -2.8121 -2.0453 -2.8800
2s -o.S7kS o.?OSO 3.1166 .1.26S6 -1.$950 -2.sseo -t.6 99 .2.6W)2
26 -0.6000 o.S862 3.os99 -1.2915 -1.2GS7 -2.,3Sb6 -1.).38$ -2.4h?S
27 -o.Sll&S o.Jl!i 3.1700 -1.2387 -1.40.$1 -2.21'61 -1.47SO -2.3(,04
2 -o.Shli6 0.2226 J.10.i; -1.2.389 -1.1347 -2.0872 -1.4,X>l -2.2186
29 -o.siss 0.0766 3.ol.sOO -1.2s11 -1.0046 -1.8649 -1.0712 -2.oooo
.YJ -o.So.$6 -0.0263 3.099b -1.21'6.) -o.6222 -1.5268 -o.61SS -1.9 00
.31 -o.S2SJ -0.1011 2.9032 -1.2937 -o.68o) .J..So64 ..0.7476 -1.6;2
32 -o.L9$6 -0.1602 2.646J -1.2613 ..0.1001 .J..1;048 .o.1eoo -1.;618
)J -0.4717 -o.27o6 2.12s1 -1.2)16 -o.5728 -1.2229 .0.9849 -1.3791
34 -o.4600 -0.):)99 2.7096 -1.2320 -o.62$S -1.1677 -o.111s -1.3J8S
35 -o.4h2S -0.3809 2.6400 -1.2111 -0.6211 -1.0450 -0.7219 -1.3.300
.36 -0.4232 -0.4322 2.s1so -1.1e1s -0.7163 -o.961) -o.eL;o -1.1322
-o.t,106
~
-o.SJ18 2.~000 -1.1600 -1.0209 -o.B052 -1.2199 -0.9665
.0.3910 .o.62)1 2. 00 -1.1312 .1.1696 -0.6937 .J..~us .o.8J7S
39 -o.3528 -o.7J8S 2.4000 .1.0335 -1.3912 -o.6280 -1.1023 -0.7700
LO -0.)780 -o.8)7S 2.3500 -1.1188 -1.6373 -0.3937 -2.0280 -o.456$
h1 -0.)5;9 -0.92L9 2.3199 -1.0916 -1.90S7 -0.3034 -2.)800 -0.3790
b2 -o.)468 -1.06S4 2.JCX)O -1.ob20 -2.1864 -0.1926 .7396 -0.240
1 -o.J2z2 2.22 -2. 062 -o.oz6a -2.2600 -o.o~
2
~ coa'f • ffiSb
ere
Ch point.
the UH of CllCul lt Cl ba sh that th dlr Uon coalne of th•
nor 1 to• pl to 1 cutain 1urt1 • u iv n b,y
43
Te 1• ii. r~c •
1+122
FSi, k.
-.i
•
Point ' 1 ~ 6
i. 9;, J.2 2. • ·.3· -5· 1895 o•
2) 1.7614 3.102s 2.9b11 2.9529 -6.SJSO .9.e;23 o.S999
24 1.7246 2.9742 2.92o6 2.9))6 -?.994h -6.4407 0.$885
2S 1.seoJ 2.4973 2.0373 2.s1~9 -'..i· 749S -7.6009 o.ssse
6 1.s1YJ 2.4643 2.8320 2.el!£6 •J.7941 <,.9377 o.$SS1
2.10ss o.L90S
27
28
2g
1.3291
1.3233
1.2ose
1.16SS
1.7600
1.h$86
2.7037
2.6h76
2.768
2.70h6
2.67S6
-~.0840
.3.8678
-2.8661
,.mb
-6.S.35S
-s.3)12
0.4900
o.hSb
)) 1.1241 1.26SS 2.611S 2.629) -1.'/S62 -!).206U 0.431
31 1.06a2 1.1400 2.sea!i 2.sm -1.9h,;8 -4.2962 o.us2
.32 o.9S91 0.9187 2.$415 2.$660 -2.oois .0072 osns
33 o.6SlS 0.7286 2.$600 2.$)18 -2.s133 .3.;192 o.3333
JU 0.777) o.ShSl 2.b69 2.Slli9 -1.6044 -J.)662 o.314e
3S 0.6825 o.i.t.) 2.hS.36 2.!i617 -1.7771 -3.2741 o.2a90
~ o.ssco o.Jt..34 2.L2B9 .~1.&12 .0600 -2.S6S6 0.2417
37 o.scn1 o.2s11 2.~lOl.J 2.4196 ~.8134 ~.242S 0.21~
.)6 0.4126 0.1704 2.3917 2.!tOlO ..3.3900 -2.0213 0.1727
39 O.)H6 0.10.34 2.3790 2.3853 ....-l.0496 -1.sm O.lJSO
ha 0.2,aa o.os12 2.~s 2.3737 ....!,.5007 -1.1521 0.0967
h1 o.11.os 0.0196 2 •.)601 2.3643 -?.622S -o.89$S 0.0831
L2 o.Ol6J 0.0021 2.3579 2.3S90 -6.1616 -o.s68o 0.0892
1 0.0000 0.0000 2-~1.2 1.178.$ -J.1879 -0.1130 0.0000
Table L. (c~.)
?olnt
ill·&!·
1.0 --
llEc. 11ln.
G-
22 o. • ) •J -1. • 0.7300
23 o.eooo -0.6986 134•19 -1.0241 0.619 Sl-42 0.764
24 o.aoss -o.71S8 13$-4.3 -o.91S3 0.5747 S4-SS 0.8183
2S o.8))9 -0.7.lSO 137·16 -0.9228 o.s1oa S9·17 o.8S91
26 0.8320 -0.75~7 138-00 -o.6693 o.4822 61-10 o.8760
27 o.s1os -o.11So lh0-48 -o.6156 0.4000 ~s o.916S
28 o.a?OS -o.194s 11.2-37 -o.76U1 0.3740 68-02 0.9274
29 o.689S -o.8lli 144·.34 -0.711) 0.3226 71·12 0.91'66
)) 0.9006 .o.8.)21 146-19 -o.666S o.2s79 73-16 0.9577
.31 0.9091 -o.6Sho 148-39 -o.6092 0.2534 7.)-19 0.9673
.32 0.92.so -0.812> lS0-16 -o.S600 0.2119 77-U6 0.9773
33 0.9200 -o.e 152-59 -0.5099 0.1700 80-13 o.96SS
J4 o.923s -o.90he J.Sh-1.&6 -o.1;706 0.1401 81·29 0.9890
JS 0.9600 -0.9250 157-40 -0.1'106 0.1166 83-11 0.9929
;;6 0.9696 -o.9t-OS 16o-OS -o.;6ll 0.0872 64-00 o.99hS
31 0·977S -o.9SS6 162-52 -0.308; o.0646 66·16 0.9979
~ o.9aso .o.968S 165-35 -0.2571 O.C444 87-27 0.9990
Y) 0.9 90 ·<>-91~ 16S-2S -o.,oso 0.0261 c-28 0.9996
hO o.9a94 -0.9899 171-?1 -o.1432 0.0133 89·12 0.9999
41 o.99as -o.99h.) 173~ -o.1o;o 0.0001 B~.30 0.9999
h2 1.00)0 -o.99n 179•11 -o.olhJ 0.0001 90-00 1.0000
l l.OCOO -1.0000 180-00 -0.0000 0.0000 ~-\>O 1.0000
Table h• (Cent.)
-.
~olnt
22 o. • -2. 0 -i. 71 • l • 0
2.3 o.11ss 1&.68$0 .Ji.SS69 .J.S7SO -2.620) 7.89SO s.9000
24 0.6982 h.1692 -4.J:)l.&8 .3.S208 -2.4779 6.1ess h·9S13
2s 0.6782 3.22o6 -3.4965 •J.006h -1.7821 6.))56 1i.22so
26 0.6$61 2.$043 -2.6.)99 -2.$087 -1.3723 S.7SS2 3.8471&
27 0.6320 2.ses2 •).1647 -2.9000 -1.26$9 S.69$1 3.2011
28 o.6o71 2 • .3478 •J.068S •2.&M -1.11s9e s.2019 2.e.;-,s
29 o.S798 1.6.$88 -2.3542 -2.2274 -0.7)67 4.7SOS 2.4296
)) o.SSh6 0.9746 -1.~611 -1.boS6 -o.ti2ss h.6SS6 2.2401
.31 0~203 1.01os -1.6~ -1.6022 -o.4150 3.9000 1.781.S
32 o BM 0.9772 -1.7 -1.70S3 -o.)662 J.9000 1.3100
33 0.1543 1.1140 -2.2400 -2.22s2 -o.)810 .).2.;86 1.1710
.34 o.!i2S8 o.16e1 •1.629.3 -1.6091 -0.2416 J.1008 1.060)
35 0.3800 o.67SS -1.6422 -1.6324 -0.19:$0 J.138.3 0.945)
J6 0.3398 0.7000 -1.9m -1.9262 -0.1690 2.1aoa 0.6899
31 0.2946 o.a300 -2.66o6 -2.S100 -0.1662 2.1974 o.472S
J8 0.2490 o.w.o .3.2800 •J.27SO -O.J.168 1.9892 0.348J
39 o.2008 0.8099 -J.96S3 .3.96JS -o.1osa 1.811.3 o.2478
40 0.141 o.6816 -h.7498 -b.7498 -o.06SS 1.lli17 0.1113
41 o.10liS o.s66S .t;.60S7 -S.60)7 -O.OS60 o.ee97 0.07~
b2 O.Olh.3 0.092$ -6.4516 -6.l.S16 -o.006h o.S6eo 0.0496
1 0.0000 0.0000 -2.1ez2 -~·1812 -o.OCXX> 0.11~ 0.0000
6
Tabl• • (Ceft:l.)
PH FV
• •
6~;
Point 72
22 5.1"1
2) s.0141 lo.saso
24 b•.X>76 9.1405
25 J,.52l'J 1.W66
26 h.3529 6.:!)17
27 4.4292 s.1e..s3
28 h.0$21 ~.18Sl
29 3.99.)6 4. 84
.)J h.26ol 3.2147
Jl J.h5SO 2.7920
32 J.SJ18 2.2872
33 2.6576 2.2aso
34 2.8590 1.7290
l) 2.9433 l.6210
j6 2.6118 1.3699
37 2.0312 i.~2s
JS 1.5434 1.1923
39 1.10.ss l.OS77
~ l.0762 0.7929
41 o.a337 o.66o9
b2 0.5616 0.11&21
l o.11p 0.0000
h?
ere
f 11 th• • 1• we the llOl"•l • th• oaltlv• dlrcctl t th• z-
axle.
1hl• r•l to th• t• lftt 1• 1110 •1 to th• dlrteU of th• tore•
f'Tb "1ic 11 contal ed In th• ta tnt lane. Tlltretore, the torcu Frb •k•
•n angle 180• ( vlth th• XY-pl•n••
Tb• vutlc11 component• ot Ffb can bl obtained, thcrcfot• 1 !ff
Mb• PTb •ln (180-/ ) • FTb atnf
th• projected COlllpon•nt• of rrh on th• XY•;:>lan• are equal to
•
•110,
~erctore,
t
48
·.
flg.9 - ANGLE €. ..
-2 • c2 + b'J. •2 Cb COi 6-e
Th•t•t re,
a2 +~-Cc t•nt- t {)2 • c2 + t:J. -2 c COi 6·
•ca obt•ln
cos~• -tan 1tt f
ire
1J lrdf Ir• the • le• •fine ln 1tep1 11 Ind 12.
en th• 1 111 ~ are cbt11ned, th• fore 1 f'RTb can ruolved. lnto tvo
ca11p nt1. 1h•H arc noraal to the proj«:Uan ct th• groin, t.e., M'b •
th• c •nt• t• ntlal to th• project! Cf th• roln,
t.1., Tb • f'Rfb x c e: • Tbtse nor•l c:e111:>cr21nt1 are lance .by th• or•l
c ~cnt1 t th• Jlcent ~tr.
z tal proJ t I • 1beretort I 1n Cll ul•t lng th• ltfelHI du• to thl• type or
. 1 din I • JJ, Jl .32 •1 be ed. 1hb hpl111 thlt 1 thod 1lallar to
that uaed ln st • 1 to 11 l c:alc l•t the 1tre1111 d to ad lced ould b •
tolloved. Hevner, th• 1p 11'1 Uve loed t roor1 a hotlz tal proJtctlOl'l
c1 be r ..ce to • aurtllce le1d !nee an curate c1lculmtlon ot th• curved
1urt1e1 b p dbl•. lhl1 surflc• 1re1 can obt•l~ dividing th• total
vuttcal forcu c trlbute<t to th• groin, ln col 2) 1 by th• d•d
lced ptr cnlt tr a, t.c.,
zs.112
1urt1e1 are • o,os • 15lla.25 1q. ft.
50
the L. L. to be equal to .)'.) lb. per aq. ft., th• ratio ot tct 1
1urt1C• to th• bork t•l projtctlon 11
~ • 2.472.
1ht L. L. pa t 1urt111 ere• 1• then
~-~·02U3
rr;i:; so.oo • •
To lrolud• • e.ttect L. L., th• w1 • t 1tHtH1 ind tote•• 1lntd
st.-,, isr
OlrtcUQ Of prL-¥:1;>•1 ltHIHS can be detua!ned by l•tU s ln Eq. 40
be 1 to ztro. 'there.tore,
Sl
2
T2 •••x ci'2p . ~ C$n! JJ C!! (£..P,) T era CB~l.
• n °' + iea aln P<.. + :I •ln O<..
F point 22 it Is t that
•
jJ. 66 ~J·.
• 0.36 )( 1.243
ft' • J x 12 x 1000 • 12. 4.l psi.
S2
TbeH valuu ara vtll below th• allowable liJllt1 spsclflld by th•
Jlalldlng Code CACI .318-6J).
By lmpectlon or otb.•r v~lu•s e! T, Tx and r1 , thown In oolu.n 3l ot Table
Discussion ot Results
by the cover onr the reln!orcemcnt, better ln1ul1Uon, •nd fire proofing.
Th•nfou, lt 1c.• that th.r• ls no netd. to reduce the th1cknu1 of th• shell.
An approxl•t• check ot th• accuHcy of the result• can be obh!ned by
C:Olllpltlng the ttal surface ar• of the tr11ngl9 1-B-42• 1 to Its projection on
tht XY-plant 1 as 1ugge1ted b;y C1nd1116 1 l .e.,
1
•
*~:~s 2.472.
Thls ratio Ht• to be r•sonable, e.peclall;y 1t lt ls COllpH'ed to the
ratio 2.8 obtained by Candd• ln his exanpl• of • groined v1ult 20 •· sq. ln
plan and with 1 rls• ot 10 ••
Sh
Coaclualon
t1vortd bec1use they ara vcll ..dlpt•d to th• use o! ceff:rcte. They show a
l'eurkablc rcd1ta11u to any fora ot vibtatiOflll lhcck. They ire 11eon011lc1l
•nd. 1110 give the erc:hitcct an q>portunity to dtvdcp aore l•gln11Uvc 1nd
gl'lc•tul snap•• Of •tructures than th• c:c..ventlonal l!taa and colu. bsildlngs.
SS
Acknowltdgacnt
by
AN AffiTRACT OF
P.. MASTER•S REPORT
M!>STER OF SCIENCE
1965
AI~ .AB3TRACT
The .me:'!lbrane theory in :nany c;:;ses £;ives ci reasonable estimate of the stress
distribution of dou!Jly curved shells. Hi th t!"le prov is ion that they are sub-
conditions of equilibri~~ , and the rise-to-span of shell ratio is not less than
lelogra;il formed by r.: cving a straisht lir.e along two nonparallel, nonir.tersecting
lines in sp:::ce.
T'ne equation Z•1XY represents the sinplest possible equation of the second
T:te seneral equations for membrane stresses for doubly curved shells c2n
be obt2 ined :by writing the eJ..?ress ions for the eqt;.ilibr iu;11 of a:i ele;ner.t for med
ele;Je:it into fict it i0us stresses zcting on the ;;>reject ion of the ele:n.ent on the
XY-plcne. Tne resulting expressions 8re in the form of three partial differ -
ential equations of the second order. 1be use of the stress fu.4ction will
reduce these three equztions to one. This will facilitate the solution of the
f::::vored because they are well-adapted to the use of concrete. They show a
a:1d also sive the architect an opportunity to develop ~r.ore imaginative and grace-
ful shapes of structures than the c onve;it ional beam and c olur.m buildings.
Tne sroined vault is taken as a design example. Tne basic stresses in the
However, readjustment of stresses is required i!'l this case due to the existence
o! a free edge ·which cannot withsta!'ld any types of stresses, since there is no
cancelling out the stresses calculated at the free edge by introducinG a set of
stresses equal and C?posite to those obtained at this edge. TI1.is involves the
shell are calculated, 2nd a suggested procedure for desi£n of the groin is pre-
sented. TI1.e principal stresses are also obtained which are used as the b2sis
Shap e Characteristics
Each method of generating shell
hape imparn. specific and character-
Fig.1: Rectangular ~hells 20X 14 m. shape Fig. .J: P11eummic shape derfred from an ir- istic qualitie · to the shape it create:..
found by pne1111wtic method. 1955 regular frame For instance. the pneumatic method
Structural En ~in cc ring lntem alional 3/94 Shell and Spatial Structures l.J:>
helpful in the creative process of shell
design. but only when the structure
the) propose have been built and mon·
itored over a period of time.
An accurate physical model on the
other hand i the Lil..eness of the real
building. and contain all of the struc-
ture ·s problems and questions. also the
one not yet known. Unre tricted ex-
ploration of the unknown i. a most im-
portant tep in innovation. '111erefore.
phy ical models are likely to remain a
necessary design tool for new "ihdl
form in the future.
Fig. 8: /11tloor swimming pool. Brugg. Swir:.erla11d. Four-pnim-!tupporrecl sht!ll. 35 x 35 111. Shell Construction
1981
Form work
A crucial step in the con truction of a
shell i the de ign of the formwork.
Curved gl ue-laminated timber beam
or flexible metal trus e upportcd by
light caffolding are favoured methods
(Fig. 17). The curved beam cll.!fine Lhc
shape and upport the light timber
strips over which wood fibre . lab
[iJled with rockfibre insulation arc
placed (Fig. 18).
The curved beams can be reused when
Fif;I. 9: fl1ree-poi11Huppom•d sliells 11'irlt free edge.~. ltigltway serl'lce swrion. Sll'ir:.erland.
the same shell form is re peated. Dif-
1968 ferent projects can share fonmvork el-
emenL in order to keep co ts IO\\. 111c
resulting designs need not look al all
the same. nor even very imilar (Figs.
12. 13).
Cladding
Cladding i an even more important
cost factor in hell constructio n. Nor-
mally. double-curved and especially
spherical surface cannot be clad with
quare or rectangular panels. as can
single-cun·ed cylindrical surface-.. To
cover a sphere. every panel has to be
cut to differing angles. making the
work very expensive.
Pig. JO: Site/I:. for sports cewn te1111i~ lwll w1d indoor \11·i111ming pool. lleimberR. S11'ir:.er· Pneumatic. hanging and llo\\ -generat-
land. !97R ed hell . however. although doublc-
cun·ed. can largd) be clad by square
or rectangular panels. without cu tting.
For statically accurate physical model For shape-finding. accurate phy ical There i no direct explanation for this
test . a great deal of knowledge and modelling remain invaluable. The au- surpri ing fact. yet it ha functioned in
experience, a well a patience. i nec- thor ha done all shape-finding for new practice for decade .
e sary. Such a time-consuming process shells merely by physical experiment
can only be undertaken in exceptional This contribute con iderahly to the
or by design. This is possible when
ca cs due to its inherent expense. For economy of shells. Indeed. without thi
careful and very accurate techniques
ta tic or stability a sessmcnt. computer advamage. shell construction would be
are applied. lbis natural way of find-
programs arc undeniably quicker and substantially more expensive.
ing hapes leads to the answer directly.
cheaper. 'oneth~lcl>s. a healthy critical without computers or other aids. To- There 1s still a son of secret to the cffi.
ense hould be employed when inter- day. some computer programs aspire cient cladding of hell : the operation
preting the result derived from com- tu simulate the c direct experimental must tart a t particular lines on the
puter. processes. Somcda) computers may be shetrs surface. otherwise all the panels
l nsula1io11 Panels
The cladding panels arc at the same
time insulation slabs which remain a
part of the building. Again. cost is re-
duced by u ing one element for two
purpo es. The inf>ulation labi. arc ex-
Fig. 11: Remmg11lt1r shell~ for ll!lllli' 11,11/. \farin. S11'i1::t·rltmd. 1983
po cd to the inside of the shell. On the
upper side. their rough surface adheres
them 10 the concrete. which penetrates
it when poured.
About one panel out of 1en thousand
tends to fall after the formwork i re-
moved. Therefore. additional hooks
are placed on the panels to guarantee
. ccure attachment. The hooks must be
non-corroding and must not allow heat
loss through the insulation panels.
In ulation panel with noi e-absorbing
Fig. 12: Fo11r-poi111-mppor1ed slrell 11•itli [rel! edges for a garden center. Ca111ori110. Swa:;er-
surface arc recommended. This leads
land. 1971
to excellent acoustic properties in the
building. eliminating echoes. Perforat-
ed metallic heets are rather ineffec-
tive for this purpo c.
R ei11forcemen1
Experience has shown that shell thick-
ness should not be le s than 8-9 cm.
Two layers of reinforcement arc rec-
ommended (Fig. 19). Jn this way. defor-
mation and cracking under local
bending are kept to a minimum.
StTUctural Engineering lnlcm ational 3/9-' Shdl and Spatial Structu r~:.- 1-15
Therefore. the workers have to be fit
and \\ell 1raincd.
Monitoring
After a hell ha been poured. the con-
crete hardl'ned and ripened. the pre·
stres ing set and the formwork re-
moved. the real lifetime of the building
begins. The hell has tow ith tand grav-
itational loads. temperature changes
and all the innuence of ''ind and
weather and atmo pht!ric aggrc,...ion.
Shell h<l\C <,tatic propenie... quit~ dif-
ferent from ordinal) building Their
behaviour can lead to a long sen ice
life. but al o to premature collapse.
llle pecific peculiarity of '>hell., i'
their lo'" change of shape. From the
Fn:.. I .J. Pril'dtt rt»Htlence. 511 i1:.erlu11d. 1986 moment a hell stand on its O\\ n.
light hape denection · commence.
Denections C\Cn in Lhc ft~t fc,, day'>
after completion can be quite con 1d-
erable. and they continue to mcrca..,c.
Such dencctions are harmlcs-; for some
hape : but for other.; they can be latal.
rr. for instance. a deflection leads to a
flattening or a curved part. the rndius
of cun a tun.! around it increa. cs. \\ hich
results m mcreased tre es and de-
creascd -.1ability.
Inadequate or weak shell shape'> arc
characteri<,etl b~ their inability to limit
deflection before they reach a critical
tagc. For ..uch hells. collap c b onl) a
mailer of time.
u ~ g
fig.15. 1\1r Forn• \Jmewn. Diibemlor.f. 511'11:.erlwul. 1987
146 Shdl Jnd patial tructure-. tructur::il E ng ineering Jnlc rn::ilio nul 3/9.t
Shell ... tructure.., should be monitored
pl!riodicall) after completion to gain
information about the character and
degree of <li.,placements. 'ormal hell
bcha\ iour indicate that di placement
arri' cat an end 'aluc after some years.
thus auaining a degree of st!curity. But
monitoring should continue.
There an! definite tabilit) limits for
-hell de igns. EH!f) hm compres. ion Fi~. 17: Formwork. for .\lu•ll /8X 49 m · g/11e-lm11i11att•ti be"ms un "light _\ct1ffoldim:. 19"'8
member ha., a critical load which may
not be exceeded. Theof). tests on-site
and ml!asuremcnt ha,·e hcd ome
light on the stability limits for hells.
hut the subject is not yet full) under-
stood. E pccially the combination of
instabilit) modes and long-term be-
ha' iour still need., a grca1 deal of sci-
entific investigation.
Future Development
The shell as a -.tructural form i o effi-
cient that it i" as..urcd of hright future.
Sooner or later. ecological concerns Fig. 18: Tu11her stripJ ""d im11/u1io11 1/ah.1 011 /1t•\ible mewl trllS!>e.i;.. Sic/1 Bldi:. Gem•rn.
"ill compel their \\ idcr U'ic. Dc\'elop- S11 it:erland. 1969
ing countrie., arc rich in manpower.
but poor in re ource Pro\·ided the
craft men and engineer.; in these coun-
tric., have the neccs af) skills. shells
should be \\Clcomcd and \\ idely ap-
There is a general trend toda) in the 121 ISLER. H. \ e11 Shapes for Slrcll1·
de ig.n of building. to brt!ak out of the Tin!lll\ Yean Aftt'r. JASS Bulletin 'r.
plied. 71 Tl. \ladrid. J9"'9.
boring uniformit\ of cubic architec-
Even in the richer countries. hell ture. Many new building are decorat- 13) ISLER. H. Typologie 1111d Ted111if.. dt•r
ha\'e their place. not only becau e of ed "ith ome form of curn!d elements. 111otleme11 Sclwlen. \\'erk. Bauen und \\oh-
their long-term economy. hut bccau...e impo ed for aesthetic reasons to inter- nen. Di:c. 1983.
of their pleasing 'ihapes a well as their rupt the monoton) of the hox-hke
[4] BILLI GTO~. D. P. The Tower 1111d
great varict). structural forms. Shell rorm · can pro-
tire Bmlgt•. Princeton Uni\'. Press. 1985. pp.
vide a remedy for design monoton). 210-232.
Grid . frame-. and other geometric spa-
offering a rt!al altemati\e.
tial 'itructure arc \'cry frcque111ly u'ied (51 RAMt--1. E.: CHL'1 CK. E. //<'i11:
toe.lay to co,er la1 ge areas. Indecd. for /\ll!r. Sdwle11. Karl Kramer Verlag. Stull-
large spans in excess of. '>a). 90 m. Education of the tudent "ho "ill be
the architect and engineer-; of the fu - gan. 1986.
shelb arc not appropriah.:. Ho"c'er. ture ha to find ne\\ (or rather. old) (6] BAL/.. ~I .Arc/1irecmral .A~pect~ oj Or-
for medium '>pan.... shells present con-
way . Student must be mo,ed a\\3}. at ~1111ic Forms 111ul Desil{11 of Concrete Sl1t•ll.1.
siderabk adn1ntages. IA Bulletin ;-..o.10632.19<'9.
least occasionall}. from the computer
monitor. The hands-on adventure of (7] ISLER. H Ele~mue .\.lc>delle. Deut'Ch\.
hapt!-finding is part of the proccs Bau1eitung. 7 1990.
Creation demand discipline. i...no" l-
edge. and opennes · Without a\\ 111 for (8) ISLER. H. Ge111•ra1i11e Shell Sh11pe.{ b'
l'hnirnl F•.xperm1e111.1. IASS Bulletin '\r
implicit~. shell hapt!s in ma<,scs
111 J.t Copenhagen. 1991.
would become tedious. \\Or c than
\\hat we complain about toda). (9] ISLER. 11. .\.fotlemer Sclwle11bt111. Ar-
~- cus 18. 1992.
Perhap the architecture of tomorro" I IOJ
COPCLAXD. J. A. Str11c111ral Art.
\\ill disco\'er interesting nc" territOr) ! Concrete Quarterly. Winter 1992.
[Ill ILER. H. The \\ay 10 Slr"pe. Int.
d flu? lt'.\I. i/ltotrtl//llllJ and :.lid/ :.IW/11'\ 111 tlit:. llrtl
ymp. on Innovative World of Concrete.
de 11rt' the illft•llec11111/ pmpt•rt1 of II /1/t•r
Bangalore. India. 1993.
[ 12] ISLFR. H. Longterm 8t•ltm·io11r of
C( • ~ Related Literature Slre/lr,, Seiken IASS-S)mp.. Tokyo. Oct.
c..
1993.
Fig. 19: Dt'for111<11io11 t111J.:le with one layt'r [I] ISLER. H. .\'e11 Sltape1 f<n '>lie/1.1. (13) ISLER. H. Concretl' Sire/ls Toti".' ·
of rt'111fonc111c111 and ll'itlr 111·0 (helc>11 > !ASS Bulktin :\r. ~ C-3. ~ladrill. 19:W. IASS-A CE ymp.. Atlanta. L"SA. 1994.
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
a r t i c l e i n f o a b s t r a c t
Article history: This paper sets out to explore the potential of combining a cable net with a fabric, in particular to scale
Received 3 November 2013 the concept of flexible formworks to the size of large-span roofs and bridges, especially when applying a
Revised 20 May 2014 thin coat of concrete or mortar to form a shell structure. By carefully designing the cable net and its
Accepted 21 May 2014
topology, and calculating and controlling the prestressing forces, it is possible to form a wide range of
anticlastic shapes, beyond those of the hyperbolic paraboloid.
A complete workflow for the computational design of a shell shape and its corresponding flexible
Keywords:
formwork are presented as a proof-of-concept for future work. Two prototype shell structures were built
Shell structure
Cable net
based on this workflow to validate the overall approach, to compare the built geometry with that of the
Fabric formwork design model and to identify further challenges when developing and scaling up the concept. In addition,
Flexible formwork a comprehensive overview of flexible formworks for anticlastic shells is presented to frame the present
Form finding research.
Shape optimization Ó 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2014.05.036
0141-0296/Ó 2014 Elsevier Ltd. All rights reserved.
40 D. Veenendaal, P. Block / Engineering Structures 75 (2014) 39–50
1.2. Objectives
The innovations in this and future work are rooted in the longer
history of using (prestressed) fabrics as a flexible formwork to
construct thin-shell concrete structures. This section serves as a
first comprehensive overview of flexible formworks for anticlastic
shells. The overview is meant to frame the present research and
highlight some of the similarities and differences with our partic-
ular approach.
vault with a 6m span in San Bartolo, Mexico City and again for a although it is doubly curved, it can be described by straight
rural school near Victoria, Tamaulipas in 1951 [7]. elements. This presented a uniquely efficient way of constructing
shuttering made from straight, reusable timber planks (Fig. 3).
Although both this and the Ctesiphon system were ingenious
2.2. Hyperbolic paraboloid shells construction systems at the time, their disappearance from prac-
tice is most likely related to the general decline of shell building.
After that, Candela moved on to geometries based on the hyper- The rising cost of labor and declining price of steel, made shell
bolic paraboloid (hypar). This is a ruled shape, meaning that building less attractive.
Alternative structural systems to cover large spaces became
available through the development of steel and timber gridshells,
and more importantly of tensioned membrane structures. In the
latter case, the membrane is the structure itself, requiring no
formwork and falsework. These systems became possible through
the increasing quality, strength and affordability of coated polymer
fabrics. At the same time, these fabrics gave rise to new opportuni-
ties for fabric-formed concrete [8] and the present method.
Fig. 3. Falsework (top) and current state in 2008 (bottom) of the shell of the Chapel
2.4. Recent fabric-formed anticlastic shells
Lomas de Cuernavaca, Mexico, 1958, by Félix Candela et al. It has a minimum 18 m
span and thickness of 4 cm. Photos (top) ÓPrinceton University Library and In 2006, several experiments at Eindhoven University of
(bottom) ÓEduardo Alarcón. Technology were intended to demonstrate a proposed construction
Fig. 4. Latex-modified concrete hypar roofs by TSC Global, Port-au-Prince, Haiti, 2011. Photos (b and c) by Stephen Riley, TSC Global (Creative Commons License CC-BY-NC-SA
2.0).
42 D. Veenendaal, P. Block / Engineering Structures 75 (2014) 39–50
method using fabric and shotcrete. A 7 m high, 2.5 m wide proto- above 80% of the applied load. The wires, spaced 12 in. (30.5 cm),
type with a thickness of 7 cm was constructed (Fig. 5a). The project and prestressed up to 600 lb (2.7 kN) using ‘‘standard prestressing
concluded that the construction method would be feasible in a equipment’’ [23].
developed country, but that the surface could have deviations as As a result of these experiments, two 64 ft (19.5 m) square
a result of the application of shotcrete, up to several centimeters buildings were soon successfully completed; the Bay Refining
[14]. A follow-up feasibility report showed an additional experi- Gas Station and Carwash in Midland, Michigan [24], and the club-
ment replacing the fabric by a stiffer metal mesh to resolve the house at the Purdue Golf Course in West Lafayette, Indiana [25]
issue of form control, and mentioned the use of a cable net with (demolished in the mid-1990s), both an assembly of four hypars
rebar for the actual proposed construction method [15]. (Fig. 6). The Purdue shell had two layers of 0.135 in. (3.4 mm)
By 2009, the Centre for Architectural Structures and Technol- wires, 6 ft (1.8 m) offset from the straight line generators. The
ogy, or CAST, at the University of Manitoba, Canada, had under- bottom wires were spaced 12 in. (30 cm) apart, the top layer
taken several experiments in small- to fullscale fabric formed 24 in. (60 cm) apart. The shell used 3 in. thick EPS foam, a 0.5 in.
shell elements; single curved, synclastic and anticlastic shapes, stiffening mortar, a 3 in. concrete cover and traditional rebar,
many with local wrinkles corrugating the surface [16]. 16.5 cm in total.
At the Vrije Universiteit Brussel, a series of ten prototypes were
constructed from a prestressed fabric using shotcrete [17,18]. The 3. Methodology
final shell had a thickness of 5 cm and a 2 m span (Fig. 5b). After
applying the shotcrete, the deformed shape was calculated and To fully realize the structural efficiency of a flexibly formed
compared to a numerical model, but for the ten shells, deviations shell, it is crucial to both design an optimal shell within the pro-
from the expected deflections were measured ranging between ject’s constraints and to control the cable forces such that its form,
5% and 58%, to more than 100% for non-coated fabrics. These errors despite the formwork’s flexibility and the weight of the wet con-
were attributed to ‘‘slip at the fixation points, and the dynamic crete, is in the end exactly as required. A computational approach
effects of the shotcreting’’. to realize this goal was developed and is explained in this section
Yet another independent physical experiment was carried out in more detail. The procedure is shown in Fig. 7, and consists of
at the University of Edinburgh, as part of a student project, to eight steps:
investigate constructional aspects of the fabric forming of a hypar
shell [19]. The work is unpublished at present, but briefly shown by 1. establishing the boundary conditions (Section 3.1) and tar-
Brennan et al. [20] (Fig. 5c). get shape of the shell (Section 3.2);
2. generating a cable-net topology and mapping it onto the
2.5. Discrete flexible formworks surface (Section 3.3);
3. patterning of the surface for the fabric (Section 3.4);
An alternative tensile formwork system is to replace the fabric 4. computing the nodal target loads based on the target shape
by discrete elements, such as the system briefly shown by De and thickness (Section 3.5);
Bolster et al. [21] in the context of hypar shells. This reusable 5. best-fit optimization of the cable forces under load in the
and reconfigurable ‘cable net’ was proposed by Mollaert and final state (Section 3.6);
Hebbelinck [22] and was in fact constructed as a network of edge 6. materialization of the cables (Section 3.7);
chains and adjustable belts. The network was covered with EPS 7. determination of the cable forces prior to casting through
tiles before applying a fiber-reinforced mineral polymer on either static analysis of the unloaded cable net (Section 3.8); and
side (unpublished work by Sven Hebbelinck). 8. design of the external frame that ensures the chosen
A similar, older system is the ‘‘offset wire method’’ (OWM), boundary conditions (Section 4.1).
developed by Waling and Greszczuk [23] at Purdue University,
Indiana, USA. The method used two layers of wires to sandwich a Calculations for loading cases other than the design loading
layer of EPS tiles on top of which concrete could be cast. This gen- (self-weight of the formwork and concrete) need to be carried
erated a hypar shell, whilst avoiding the need for ‘‘a forest of out afterwards to further check the design, although they could
falsework’’. potentially be integrated in step 1 for the shell and step 7 for the
After building a 37 7/8 in. (1 m) square small-scale model, they cable net.
proceeded with a large-scale 20 ft (6.1 m) square laboratory model
with a rise of 7 ft (2.1 m). The shell had 3 in. (7.6 cm) thick EPS 3.1. Boundary conditions and initial shape
designed for a 2 in. (5.1 cm) thick concrete cover, deviating by
2.7 in. (69 mm) from a true hypar at its center. A coating of mortar As a starting point, the boundary conditions and initial shape of
reduced deflections to 0.43 in. (11 mm), but showed cracking our shell are based on a ‘simple’ hyperbolic paraboloid (hypar)
y2 x2
z¼c 2
b a2
1.2m
1.8m
1.8m
Fig. 8. Boundary conditions and NURBS control points, allowed to move vertically
within the bounding box while maintaining anticlasticity.
modelling them as discrete networks under a given selfweight of load-bearing and shape-defining part of our flexible formwork sys-
the concrete. A set of two force densities (corresponding to the tem. Several criteria govern the design of the cable-net topology:
two orthogonal directions) or a (fictitious) elastic stiffness (identi-
cal for all links) were varied to obtain a shape that fitted some spa- The valency of the cable net should generally be even to allow,
tial requirements of the designer. Guldentops et al. [29] also for practical reasons, continuous cables up to the boundaries.
compared such a form to that of a hypar of similar (but not identi- Continuous cables should terminate at the boundaries to allow
cal) size and boundary conditions, showing deflections to be lower. for more convenient prestressing (and thus control) throughout,
However, to reduce deflections even more, we apply shape optimi- rather than loop within the mesh.
zation instead of form finding. Continuous cables should follow principal curvatures of the
Bletzinger et al. [31] suggested to use a form-finding method as surface to reduce the amount of prestressing required, as the
a shape generator for structural optimization by varying the loads cables’ load capacity is proportional to prestress and curvature.
(distribution of selfweight). However, compared to varying Continuous cables should have as low geodesic curvature as
geometric parameters (e.g. control points of the surface), they possible, if they are to serve as guides or seam lines for second-
demonstrate that results are less optimal (for uniform thickness), ary fabric strips or cutting patterns (thus reducing material
since the degrees of freedom are reduced. This leads us to choose waste).
a geometric parameterization, using a NURBS surface, for the time The density of the mesh should be fine enough such that the
being (Fig. 8). The control points of the NURBS surface are allowed demands on the secondary fabric are low (in terms of strength,
to move vertically within the bounding box of our boundary condi- prestressing and patterning requirements).
tions, such that the resulting geometry never becomes synclastic. The density of the mesh should be coarse enough, such that the
The shape is optimized using a genetic algorithm by varying the total length of cable, the number of intersections and the
control points [32]. amount of prestressing work (thus material and labor cost)
When optimizing a shape to achieve the lowest maximum are reduced.
deflection (serviceability limit state), with self-weight as design
loading, we expect the material to stay within its elastic range. Given these considerations, we start by roughly orienting a
Therefore, as long as no stress limits are exceeded, we can apply a quad mesh along the lines of principal curvature (Fig. 10a). The
linear elastic model. Since the present project uses a custom end points are used to plot geodesics along the surface of the shell,
concrete mix design, and does not include material testing, optimi- corresponding well with principal curvature (Fig. 10b). Then, the
zation is carried out for the probable limits of linear elastic stiffness. end points are moved to optimize the mesh width (Fig. 10c). The
Based on Eurocode strength classes, from regular C25/30 to high- objective is to reduce the standard deviation of the mesh width.
strength C90/105, the linear elastic stiffness E of the concrete might By moving the end points, the total number of nodes along the
vary between 31 GPa and 44 GPa. The results show that the stiff- boundary increases, but in terms of construction, this simplifies
ness, within this range, does not significantly influence the shape. matters, since at each point, only one cable now terminates. We
The model’s thickness is designed to be in the range of 1.5– set our mesh width at approximately 20 cm with a 30 cm upper
2.5 cm, so we optimize for both of these limits as well (Table 1). limit. Table 2 shows the result of the manual optimization.
In summary, we minimize the maximum deflection by varying
the vertical position of the four control points within the bounding
box of the boundary conditions. The problem is not subject to any 3.4. Cutting patterns
other constraints, since displacements are very low and additional
load cases are not considered at this stage. As discussed, aligning the seam lines of cutting patterns along
The FE software used (see Section 3.9) provides one type of shell geodesics minimizes waste. The cables, already geodesic, are cho-
element, that is described as being similar to the TRIC-element sen as seam lines, and the surface in between is unrolled using
[33], which already produces accurate results for doubly curved an area-preserving approach. The surface is subdivided, triangu-
geometries at coarse meshes, avoids any shear locking [34] and lated and flattened. The degree of subdivision is increased until
neglects transverse shear deformation [35] meaning that it applies the difference between surface areas of the flattened pattern and
to thin shells. the original doubly curved surface are within a certain tolerance,
Table 1 shows the results from optimization compared to the in our case 0.01 m2 (Fig. 11).
initial hypar. We observe that the relative reduction in deflections
increases with shell thickness, and thus the influence of flexural
behavior. The shape for the lower bound uniform thickness of
1.5cm is chosen for further development (Fig. 9).
The next step, having settled on the required shape of the shell,
is to map a cable net onto the surface that serves as the main
Table 1
Results (height at midspan) and fitness (max. deflection at the tips) of shape Fig. 10. Optimization of the cable-net geometry.
optimization, identical for C25/30 and C90/105.
CTN Uq ¼ DN xN þ DF xF ¼ 0 ð1Þ
CTN Vq ¼ DN yN þ DF yF ¼ 0 ð2Þ
CTN Wq p ¼ DN zN þ DF zF p ¼ 0 ð3Þ
¼ DN x
Aq p N þ DF x ¼0
F p ð4Þ
Fig. 11. Cutting patterns, cut along cable lines, used for the fabric shuttering and
the textile reinforcement of the second prototype.
where DN and DF are block diagonal matrices, and remaining vec-
tors and matrices are vertically stacked.
Of the two directions in the net, the standing cables are chosen N and
We linearize Eq. (4) with respect to the unknowns q and x
for the seams, such that the fabric strips can simply be placed over obtain [39]:
the cable net without the possibility of sliding down. The patterns
DxN
are designed to skip every other cable, to compare the effect on the DN kA A0 q0
¼p ð5Þ
concrete imprint. For engineered membrane structures, the cutting Dq
patterns are shrunk to compensate for the prestress, such that the
where weighting factor k was set to 10 in our case. This is an under-
resulting surface is properly tensioned and smooth. In our case, for
determined system of equations, since the left-hand side matrix is
simplicity of construction and aesthetic reasons, the patterns are
3nN ð3nN þ mÞ. Such a system can be solved with a least-squares
not compensated, such that the unstrained fabric sags between
solver such as GNA. In each iteration i, one sets qiþ1 ¼ qi þ Dq, then
the cables once the concrete is applied.
updates DN as well as the residuals, and repeats the iteration until
convergence is reached, in our case if kDx N k < ¼ 0:001, occurring
3.5. Target loads after 707 iterations (22 s). Upon convergence, the required pre-
stresses under the load of the fresh concrete can be calculated by
To determine the forces in the cable net, we first calculate the simply multiplying the force densities with the lengths of the cable
target loads on the nodes of the net as an approximation of our links.
target shape. Each load is assumed to be the tributary area of a cor- Fig. 12 shows the results for an orthogonal net, one based on
responding node in the net, extruded by the average thickness of OWM (see Section 2.5) and one aligning with principal curvatures,
the shell. Although the shape is optimized for 1.5 cm, the thickness each mapped onto our target surface. Note that the first cable net
for the load calculation is changed to 2.5 cm to accommodate the would have infinite forces on a hypar, but for the optimized shape
expected sag of the fabric. This means the additional thickness is has a slight curvature, allowing us to find a solution.
assumed to act as dead load, but not contribute structurally to
the shell. First, the Voronoi diagram of the projected cable-net
3.7. Materialization
nodes is determined. The cells are projected onto the target sur-
face, planarized and then extruded normal to the surface by the
Having determined the forces after casting in Section 3.6, we
assumed thickness. The density is set to 24 kN/m2.
wish to know the forces prior to casting in order to construct and
prestress our formwork. To do so, we materialize and dimension
3.6. Best-fit optimization the cable net. The smallest steel cable available to us is 2 mm INOX
with a tensile strength of 3.7 kN. This is well above the maximum
We wish to find the required cable-net forces such that, under tensile stress after casting (factor of 13.5). The stiffness of our cable
given loads of the wet concrete, the resulting concrete shell takes
the form of the target shape. As mentioned, this problem has
already been addressed and solved by the second author through
a non-linear extension of thrust network analysis (TNA) [2]. This
approach used the Levenberg–Marquardt algorithm (LMA) as a
least-squares solver, which interpolates between the Gauss–
Newton algorithm (GNA) and the method of gradient descent. For
the sake of comparison (see Section 5), non-linear extension of the
force density method (FDM) [36,37] is implemented and adapted
to our problem. The approach uses GNA as a least-squares solver.
302 806 1214 [N] 117 216 364 [N] 95 193 274 [N]
The goal is to minimize changes in coordinates DxN ; DyN ; DzN
and force densities Dq, such that the coordinates x; y; z approach Fig. 12. Cable forces after casting from best-fit optimization, where aligning with
the target shape, while satisfying static equilibrium. The results are principal curvatures reduces the required prestress.
46 D. Veenendaal, P. Block / Engineering Structures 75 (2014) 39–50
is not specified by the manufacturer, but presumably it is matrices are not implemented in IronPython, the ALGLIB library
E ¼ 195 N=mm2 based on the relevant German DIN-norm. for IronPython was later used to solve sparse least squares prob-
So, having chosen a material and cross-section, with an axial lems with xalglib.linlsqrsolvesparse, based on the LSQR
stiffness EA ¼ 195p, the initial lengths l0 can directly be computed algorithm. Unrolling of the surface to create cutting patterns was
from the forces f and lengths l in the final state [40]: achieved by adapting a Grasshopper definition by Andrew
1 Heumann.
l0 ¼ I þ ðEAÞ1 F l ð6Þ
4. Results
where I is an identity matrix of size m, and EA and F are diagonal
matrices of all stiffnesses EA and the forces f respectively. Based on the computational approach presented in the previous
section, two prototype shells were built as a proof-of-concept.
3.8. Static analysis
4.1. External frame
From the final state geometry, and the newly given stiffnesses
and resulting initial lengths, it is now possible to compute the A timber frame was built to follow the boundaries of the shell
intermediate geometry and forces prior to casting. In that situation, and resist the applied prestresses, using 9 cm square fir elements
the fresh concrete has not been applied yet, so assuming the (Fig. 14). The frame was designed such that the upper part would
selfweight of the formwork to be negligible compared to the pre- be removable for demolding, whilst the lower part would support
stresses, we simply remove the loads when computing the residual the two bottom corners of the hypar. A tension tie connected the
forces and attempt to find a new equilibrium shape. This FE two bottom corners to resist the horizontal thrust from the shell.
analysis is performed using dynamic relaxation [41] (this method Because the frame edges might deflect due to bending in the order
is well-known as a form-finding method, but here we used real of 1 mm (based on calculations), it was decided during construc-
material values, making it equivalent to FE analysis with a tion to add an additional timber cross, connecting the midpoints
particular type of element and solver). of the frame edges (see Fig. 14, right) and a top member for the sec-
After dynamic relaxation, we obtain new lengths and recom- ond prototype (to allow access for measurements).
pute the forces. The cables were dimensioned for the forces after
casting and Fig. 13 shows that, as expected, the range of forces 4.2. Cable net and prestress
has decreased for the unloaded state.
The cable net was made from 2 mm stainless steel cable. Cables
3.9. Implementation were guided through the timber frame along cringles, terminating
at eyebolts using crimp sleeves. For the first prototype, nodes were
The procedure, outlined in Fig. 7, was implemented in fixed by winding a piece of wire around the intersecting cables. For
Grasshopper for Rhinoceros, using its built-in Galapagos the second, cross clamps were used instead (Fig. 15), which also
components for the genetic algorithm, plug-ins Karamba to served as measuring points for photogrammetry.
evaluate deflections, and Kangaroo for the dynamic relaxation. In At one end of each of the twenty cables, a turnbuckle was used
addition, custom Python components were written in IronPython, to introduce prestress. The prestressing was controlled by measur-
using the NumPy library. NumPy functions used were ing the lengths of the cable segments during prestressing, attempt-
numpy.linalg.inv (which uses Lapack’s subroutine DGESV, based ing to approach the digital model as closely as possible.
on LU factorization), to solve linear systems such as in Eq. (6), and Prestressing was then further refined by measuring and checking
numpy.linalg.lstsq (which uses Lapack’s subroutine DGELSD) the forces in several cable segments. In the first prototype, this
to solve the least squares problem in Eq. (5) using singular value was done by measuring the elongation of springs at twenty loca-
decomposition (SVD). Because NumPy algorithms for sparse tions inside the net. For the second prototype, a portable cable ten-
sion meter was used instead, allowing more measurements of
higher accuracy.
4.3. Fabric
Fig. 13. Cable forces after and before casting. The range of prestress is smaller in the Fig. 14. Design and photo of the first prototype, with a cross added in the as-built
latter state. frame.
D. Veenendaal, P. Block / Engineering Structures 75 (2014) 39–50 47
at our concrete lab. The mix design was not the focus of our present
research, so the only criteria were to obtain a mixture with high
slump, resistance to shrinkage cracking and tensile capacity. The
proportions by weight were:
4.5. Construction
Table 4 shows the materials and their costs for the first proto-
type formwork. The cost of the formwork is CHF 527.79, or CHF
162.90 per plan square meter with nearly 60% due to the timber
frame. We expect that the relative cost will go down at larger
Fig. 15. Construction of the second prototype (clockwise from top left): turnbuck-
les, stitched seam, surface after casting, final result, corner support detail, and node
with cross clamp.
4.4. Concrete
After prestressing the net, the fabric was applied and a 0.9–
2.9 cm (average ca. 2.4 cm, see Fig. 16) thick layer PVA fiber rein-
forced cement mortar was hand-rendered on top. The second shell
was cast while continuously measuring the thickness for better Fig. 17. Construction of the first formwork and shell.
control, by distributing the concrete accordingly. It was also fitted
with an additional layer of AR-glass textile reinforcement. The mix
design and choice of fiber reinforcement was adapted from Máca
et al. [42] based on further discussion and availability of materials Table 3
Labor involved in constructing , and adapting and reconstructing the formwork.
Table 4
List of materials, quantities and cost in CHF for the first formwork.
Component Type Qty. Cost per unit Total excl. Total cost
Timber 100 100 fir 25 m 11.50 287.50 310.50
Fabric Propex 60-7041 9 m2 0.71 6.39 7.60
Tacks 1.8 20 mm 0.5 kg 26.10 13.05
Cable INOX V4a 2.0 mm 100 m 0.52 52.29 58.10
Wire Steel wire 1 mm 12.00 12.96
Cringle 6 mm 60 0.62 37.26 40.24
Turnbuckle M5 20 2.00 40.00 43.20
Crimp sleeve 2 mm 60 0.09 5.49 5.93
Quick link 4 mm 40 0.56 22.32 24.11
Eye screw M6 40 20 0.56 11.20 12.10
Total 527.79
spans, where the ratio of surface to edge is higher, and for multiple average deviation from the target surface l ¼ 2:0 mm with a
use, since much of the first prototype was reused for the second. standard deviation of r ¼ 1:5 mm.
Table 5 and Fig. 18 show an overview of our two as well as
4.7. Measurements and accuracy precedent prototypes for which data was published. Here the total
difference between design model and final loaded state are
After the first shell had hardened, thirteen measurements were compared, where the difference is the sum of deformations due
taken at nodes of the cable net to see how much the actual proto- to loading and deviations between theoretical and physical model.
type deviated from the computer model. Measurements using The proposed design procedure accounts for the final loaded state
ruler and laser level revealed an average deviation l ¼ 22 mm of the formwork, so the second prototype is particularly accurate
with a standard deviation of r ¼ 1 mm. The second shell was made by comparison. In general, good results were obtained in each
to improve the method of measuring forces and geometry as the study, considering the inherent flexibility of these formworks.
main source of error. Forces were measured using a Tensitron
ACX-250-M portable cable tension meter, while photogrammetry 5. Discussion
was used to register nodal positions. The cable net was loaded
using weights of sand equivalent to the nodal loads in the com- A historical investigation yielded a large number of precedents
puter model. This also meant that the cables remained accessible of flexibly formed, anticlastic shell structures. A recurring topic
in the loaded state, allowing measurements of forces. The resulting among these works is the lack of control of deviations. Based on
Table 5
Comparison of deformations and deviations, relative to the span, in flexible formworks.
Ref. Surface Eq. thickness in Span s (m) Plan (m) Deformation d due Deviation D from Rel. difference
concrete (cm) to loading (mm) theoretical (mm) dþD
s (‰)
Fig. 18. Sequence of differences from design model to final built state for Table 5.
D. Veenendaal, P. Block / Engineering Structures 75 (2014) 39–50 49
this historical overview and our own study, possible strategies to the concept of calculating prestresses in the final state, and
improve accuracy are: materializing before finally calculating prestresses prior to
casting;
to reduce the applied loading by layering and curing the shell in the design and construction of two prototypes with a discussion
stages, thereby creating a partially or entirely self-supporting on their accuracy.
structure as soon as possible;
to increase the stiffness of the material (e.g. coated instead of It is clear that much work needs to be done for the further
uncoated fabrics, cable nets) and/or the prestress such that development of this construction method. However, the present
the formwork is less sensitive to inaccuracies of the applied research has enabled us to identify which particular aspects war-
load; rant our immediate attention. Future work will focus on improving
to design for the loaded state to exclude deformations from the the optimization procedure and tackling the challenges associated
final comparison; and with scaling up the construction method.
to accurately measure and correct the required prestresses.
Acknowledgements
Our approach is based on the latter three: a cable net for which
we calculate the prestresses required in the final state, then mea- The authors would like to thank Regine van Limmeren, Masoud
sure and control the prestresses required in the initial state. In this Akbarzadeh and Ramon Weber as well as Mile Bezbradica, Jonas
paper, we adapted a non-linear FDM for the calculation of the final Sundberg and David López López for the construction of the two
state. Its convergence was slow, but satisfactory for the present prototypes respectively. David Novák and Prof. Konrad Schindler
model. However, some preliminary numerical experiments with assisted with the photogrammetry. Further support at ETH Zurich
more complicated shapes have shown that non-linear FDM tends was provided by Oliver Zgraggen, Paul Fischlin, Heinz Richner and
to diverge, making it unsuitable for future work in its current form. Thomas Jaggi. We acknowledge our former students Paul Mayen-
In the method by Van Mele and Block [2] convergence is vastly court and Matthias Amstad, who worked on a previous formwork.
improved by reducing and controlling the amount of unknowns: We especially thank Manuela Tan, and Noah Nichols from Pro-
first by solving vertical equilibrium separately, and second by iden- pex as well as Kaloyana Kostova and John Orr from the University
tifying mi independent sets of the m unknown force densities. In of Bath for sponsoring the two fabrics; Jennifer Stevenson, Chivas
other words, the amount of unknowns decreases from 3n þ m to Brothers Ltd., for providing current photos of the Paisley ware-
n þ mi , where mi 6 m, and n are the number of free nodes. houses; Charlotte Erdmann, Engineering Library of Purdue Univer-
In further work, the objective will be to scale up the construc- sity, for identifying the Bay Service Station and securing additional
tion method. copyright permissions; and George Nez and Brad Wells, TSC Global,
In terms of computation, thickness in addition to shape optimi- for information regarding their hypar shell roofs.
zation will be included as well as stress constraints when checking
for multiple load cases based on Swiss building code and wind References
modelling. The material model will be updated based on results
from material testing of the textile and fiber reinforced concrete. [1] Van Mele T, Block P. A novel form finding method for fabric formwork for
concrete shells. In: Proceedings of the international association for shell and
In terms of construction, the timber frame and turnbuckles will spatial structures symposium 2010. Shanghai, China; 2010.
be replaced by steel and jacks. Development of the steel details and [2] Van Mele T, Block P. Novel form finding method for fabric formwork for
prestressing strategy will be an important focus. Partially leaving concrete shells. J Int Assoc Shell Spatial Struct 2011;52(4):217–24.
[3] Torsing R, Bakker J, Jansma R, Veenendaal D. Large-scale designs for mixed
the frame as a permanent edge beam will be evaluated, depending fabric and cable-net formed structures. In: Orr J, editor. Proceedings of the 2nd
on results from shape and thickness optimization, as well as based international conference on flexible formwork. Bath, UK; 2012. p. 346–55.
on non-structural considerations. [4] Billig K. Concrete shell roofs with flexible moulds. J ICE 1946;25(3):228–31.
[5] Anon. Curved roofs of large span. Concr Constr Eng 1959;54(5):171–4.
Finally, the full-scale project will serve as an enclosure, and [6] Waller JdW, Aston A. Corrugated shell roofs. ICE Proc – Eng Div
therefore integrating the structural shell with building systems 1953;2(4):153–82.
and insulation while retaining its thinness is another important [7] Faber C. Candela: the shell builder. Reinhold Publishing Corporation; 1963.
[8] Veenendaal D, West M, Block P. History and overview of fabric formwork:
aspect of the overall design.
using fabrics for concrete casting. Struct Concrete – J Fib 2011;12(3):164–77.
[9] Kersavage JA. Method for constructing a tensile-stress structure and resultant
6. Conclusions structures, patent us 3,927,496; 1975.
[10] Stepler R. Hypar structures, light, easy to build, cheap – and permanent. Pop
Sci 1980;216(2):73,74–77,168.
The field of fabric formwork represents over a century of [11] Knott A, Nez G. Latex concrete habitat. Canada: Trafford Publishing Company;
exciting inventions and discoveries, yet provides many opportuni- 2005.
[12] Balding D. The experimental seismic testing of hypar shells. Fourth-year
ties still. A historical overview of flexibly formed shell structures undergraduate project. Cambridge University; 2013.
and prototypes, often hypars, has been presented, many of which [13] Carlton WS. Material behavior of latex-modified concrete in thin hyperbolic
were constructed with comparable goals to the present work. Our paraboloid shells. Master thesis. University of Oklahoma; 2013.
[14] Pronk A, Houtman R, Afink M. The reconstruction of the philips pavilion
proposed solution has been developed to allow large-span form-
volume 1. In: Sources of architectural form, theory and practice. Kuwait City,
works with little or no falsework and enable a wider range of anti- Kuwait; 2007.
clastic shapes than those possible through traditional means. [15] Pronk A, Dominicus M. Philips-paviljoen; 2012.
[16] West M, Araya R. Fabric formwork for concrete structures and architecture. In:
The first prototype presented here is the first cable-net and
Krõplin B, Oñate E, editors. Proceedings of the international conference on
fabric formwork, and an important step towards a new cable-net textile composites and inflatable structures, structural membranes 2009.
and fabric formed roof to be built in Switzerland. Our main Barcelona; 2009.
contribution is a complete computational design workflow for [17] Cauberg N, Parmentier B, Vanneste M, Mollaert M. Fabric formwork for
flexible, architectural concrete. In: Oehlers D, Griffith M, Seracino R, editors.
the prototype shell and its required formwork. Proceedings of the international symposium on fibre-reinforced polymer
This includes reinforcement for concrete structures, FRPRCS-9. Sydney, Australia; 2009.
[18] Cauberg N, Tysmans T, Adriaenssens S, Wastiels J, Mollaert M, Belkassem B.
Shell elements of textile reinforced concrete using fabric formwork: a case
the formulation of design criteria for the cable net and its study. Adv Struct Eng 2012;15(4):677–89.
topology; [19] Ghaui B. My interests & thoughts as a part ii student; 2012. <http://
ghaui.wordpress.com/tag/hyperbolic-paraboloid/>.
50 D. Veenendaal, P. Block / Engineering Structures 75 (2014) 39–50
[20] Brennan J, Walker P, Pedreschi R, Ansell M. The potential of advanced textiles [32] Pugnale A, Sassone M. Morphogenesis and structural optimization of shell
for fabric formwork. Construct Mater 2013;166(4):229–37. structures with the aid of a genetic algorithm. J Int Assoc Shell Spatial Struct
[21] De Bolster E, Cuypers H, Van Itterbeek P, Wastiels J, De Wilde W. Use of hypar- 2007;48(3):161–6.
shell structures with textile reinforced cement matrix composites in [33] Argyris J, Tenek L, Olofsson L. Tric: a simple but sophisticated 3-node
lightweight constructions. Compos Sci Technol 2009;69:1341–7. triangular element based on 6 rigid body and 12 straining modes for fast
[22] Mollaert M, Hebbelinck S. Adaptent: a generating system for cable net computational simulations of arbitrary isotropic and laminated composite
structures. In: Fifth international conference on space structures. Guildford, shells. Comput Methods Appl Mech Eng 1997;145:11–85.
Surrey, UK; 2002. [34] Argyris J, Papadrakakis M, Apostolopoulou C, Koutsourelakis S. The tric shell
[23] Waling J, Greszczuk L. Experiments with thin-shell structural models. In: element: theoretical and numerical investigation. Comput Methods Appl Mech
Proceedings, annual meeting, American Concrete Institute, vol. 57; 1960. p. Eng 2000;182:217–45.
413–31. [35] Preisinger C. Karamba User Manual for Version 1.0.3; 2013.
[24] Maddex D, Dow Alden B. Midwestern modern. Archetype Press; 2007. [36] Linkwitz K, Schek H-J. Einige Bemerkungen zur Berechnung von
[25] Waling J, Ziegler E, Kemmer H. Hy-par shell construction by offset wire vorgespannten Seilnetzkonstruktionen. Ing Arch 1971;40:145–58.
method. In: Proceedings, world conference on shell structures. San Francisco, [37] Gründig L, Schek HJ. Analytical form finding and analysis of prestressed cable
CA, US; 1962. p. 199–200. Washington, D.C., National Academy of Sciences- networks. In: International conference on tension roof structures. London, UK;
National Research Council Publication. p. 1187–453. 1974.
[26] Ortega NF, Robles SI. The design of hyperbolic paraboloids on the basis of their [38] Schek H-J. The force density method for form finding and computation of
mechanical behaviour. Thin-Wall Struct 2003;41:769–84. general networks. Comput Methods Appl Mech Eng 1974;3:115–34.
[27] Jadik T, Billington DP. Gabled hyperbolic paraboloid roofs without edge beams. [39] Linkwitz K, Veenendaal D. Chapter 12: nonlinear force density method:
J Struct Eng 1995;121:328–35. constraints on force and geometry. In: Adriaenssens S, Block P, Veenendaal D,
[28] Tomás A, Martí P. Optimality of candela’s concrete shells, a study of his Williams C, editors. Shell Structures for Architecture. Routledge; 2014. p. 143–
posthumous design. J Int Assoc Shell Spatial Struct 2010;51(1):67–77. 55.
[29] Guldentops L, Mollaert M, Adriaenssens, De Laet L, De Temmerman N. Textile [40] Linkwitz K. Formfinding by the ‘‘direct approach’’ and pertinent strategies for
formwork for concrete shells. In: Domingo A, Lazaro C, editors. Proceedings of the conceptual design of prestressed and hanging structures. Int J Space Struct
the international association for shell and spatial structures (IASS) symposium 1999;14(2):73–87.
2009. Valencia; 2009. p. 1743–54. [41] Barnes MR. Form finding and analysis of tension structures by dynamic
[30] Tysmans T, Adriaenssens S, Wastiels J. Form finding methodology for force- relaxation. Int J Space Struct 1999;14(2):89–104.
modelled anticlastic shells in glass fibre textile reinforced cement composites. [42] Máca P, Zatloukal J, Konvalinka P. Development of ultra high performance fiber
Eng Struct 2011;33:2603–11. reinforced concrete mixture. In: IEEE symposium on business, engineering and
[31] Bletzinger KU, Wüchner R, Daoud F, Camprubí N. Computational methods for industrial applications; 2012.
form finding and optimization of shells and membranes. Comput Methods
Appl Mech Eng 2005;194:3438–52.
1
I
UNUt:Hi->1'ANlllNC TllE HYPERBOLIC P1\HAHOLOI D
P I
HE ..\Tl:'\G A:'\[) COOLL'\G
A SHOPl'f'\r, n::'\TER
._.,....,_
-
~
-
__ __
.........,____
....... ._ _
,........,. ... ,..,.......,._..,,,,, ,.....
H'Tltt:-..' .\S!\l.hl"' h>H "'V Hnunor)f- t'A.RAROf.(111)
.,
. .,.
,