Professional Documents
Culture Documents
Advanced Mechanical Sample Dissertation
Advanced Mechanical Sample Dissertation
Aberdeen
August, 2018
Lateral Stiffness of Piperack Support Structures
Abstract
Recent seismic events have resulted in excessive economic loss and has
been concluded that the level of structural damage, in piperack type struc-
tures, is directly related to excessive deformations [1]. This study attempts
to develop a preliminary design algorithm for approximating lateral stiff-
ness in piperack structures, that can be used as a tool to optimise them for
earthquake performance. Previous modelling has been done on this sub-
ject by Irvine [2]. The study presented here attempted to take the initial
preliminary work of Irvine [2] further and develop a more generally applica-
ble design algorithm for approximating lateral stiffness in piperack support
systems. This work had two main deliverables which were the ability to in-
clude, non-uniform and irregular frames. A further deliverable of this study
was to include bracing in the approximate algorithm.
The method presented here is an original approach based upon calculat-
ing the storey drift of each storey. Initial work on the Tanh approximation
developed a considerably more complex set of manipulation equations than
Irvine. Conducting further investigations into this method revealed the need
for a corrected first storey stiffness approximation. Once the method was
adapted to the first storey correction a number of irregular, non-uniform,
and braced frames were investigated.
The results of this study saw the first two deliverables met with great
success, while only partial success was obtained in including braced frames.
For non-uniform and irregular frames the vast majority of approximations
were found to be within 5% of the actual stiffness, with only a few results
over 10% and the worse at c.14%. The level of accuracy obtained from
this algorithm provides justification for the more complex equations used.
The investigation of bracing, while not a complete success, provided an
interesting avenue for further research.
ii
Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii
Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . xi
1 Introduction 1
1.1 Project Overview . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Project Aim and Objectives . . . . . . . . . . . . . . . . . . 2
iii
3.4 Hyperbolic Tangent Curve . . . . . . . . . . . . . . . . . . . 13
3.5 Frame Analysis Approach . . . . . . . . . . . . . . . . . . . 14
4 Method 15
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.2 Generalising Infinitely Stiff Model . . . . . . . . . . . . . . . 15
4.2.1 Bracing . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.3 Hyperbolic Tangent . . . . . . . . . . . . . . . . . . . . . . . 17
4.3.1 Coefficient Curve Fitting . . . . . . . . . . . . . . . . 20
4.3.2 Multidimensional Curve Fitting . . . . . . . . . . . . 22
4.4 Individual Storey Approach . . . . . . . . . . . . . . . . . . 25
4.4.1 Illustrative Example . . . . . . . . . . . . . . . . . . 25
References 49
Appendices 49
B Multi 54
iv
List of Figures
v
5.3 Showing Ke /Ka ratio for a number of five storey, multi-bay
frames at Lb /Lc ratio’s of 0.5, 1, and 2. Along with compar-
ison Irvine’s method, Data points in red are calculated using
Irvine’s Tanh coefficients. . . . . . . . . . . . . . . . . . . . 31
5.4 a) Three storey, three bay frame modelled using original ap-
proach. b) Three storey, three bay frame modelled using Ns
equal to one for first floor. . . . . . . . . . . . . . . . . . . . 34
5.5 Showing three storey, three bay frame modelled using Ns =
1 × ζf s for the first storey. . . . . . . . . . . . . . . . . . . . 35
5.6 Showing normalised stiffness against Ic ratio for nine-bay,
five-storey frame. Where, a) shows Lb /Lc = 0.5, b) shows
Lb /Lc = 1 , c) shows Lb /Lc = 1.5, d) shows Lb /Lc = 2 . . . . 37
5.7 Showing normalised stiffness against Ic ratio for nine-bay,
two-storey frame. Where, a) shows Lb /Lc = 0.5, b) shows
Lb /Lc = 1 , c) shows Lb /Lc = 1.5, d) shows Lb /Lc = 2 . . . . 37
5.8 Showing normalised stiffness against Ic ratio for two-bay,
two-storey frame. Where, a) shows Lb /Lc = 0.5, b) shows
Lb /Lc = 1 , c) shows Lb /Lc = 1.5, d) shows Lb /Lc = 2 . . . . 38
5.9 Showing normalised stiffness against Ic ratio for five-bay,
four-storey irregular frame. Where, a) shows Lb /Lc = 0.5, b)
shows Lb /Lc = 1 , c) shows Lb /Lc = 1.5, d) shows Lb /Lc = 2 39
5.10 Showing normalised stiffness against Ic ratio for nine-bay,
three-storey irregular frame. Where, a) shows Lb /Lc = 0.5,
b) shows Lb /Lc = 1 , c) shows Lb /Lc = 1.5, d) shows Lb /Lc = 2 40
5.11 Ke /Ka ratio for various Infinitely stiff regular uniform braced
frames, Lb /Lc equal to one. . . . . . . . . . . . . . . . . . . 42
5.12 Ke /Ka ratio for various nine-bay Infinitely stiff regular uni-
form braced frames, at different Lb /Lc ratios . . . . . . . . . 42
5.13 Ke /Ka ratio for various Infinitely stiff regular uniform braced
frames, at Lb /Lc ratio equal to one, modelled using corrected
ISM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.14 Ke /Ka ratio for various nine-bay Infinitely stiff regular uni-
form braced frames, at different Lb /Lc ratios, modelled using
corrected ISM. . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.15 Showing ratio Ke /Ka for braced nine-bay, two-storey frame,
modelled using Tanh curve developed in Section 4.3. . . . . 45
5.16 Showing ratio Ke /Ka for braced nine-bay, two-storey frame,
modelled using experimentally altered Tanh curve. . . . . . . 45
vi
B.1 Three dimensional surface fit, relating P1c and P2c to the
number of bays and storeys in a frame. . . . . . . . . . . . . 54
B.2 Three dimensional surface fit, relating P1d and P2d to the
number of bays and storeys in a frame. . . . . . . . . . . . . 55
vii
List of Tables
4.1 Showing the coefficients for various Lc/Lb ratio for 5 storey,
1 bay structure. . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.2 Parameters for Frame shown in Fig. 4.7. . . . . . . . . . . . 25
4.3 Comparison between frame analysis package “MASTAN2 ”
and Individual Storey Approach. . . . . . . . . . . . . . . . 28
viii
Nomenclature
List of Symbols
FL Lateral Force
KL Lateral Stiffness
ix
KS Lateral Stiffness of Elastic Frame Storey
Lb Length of Beam
Lc Length of Column
Nb Number of Bay’s
Ns Number of Storeys
Abbreviations
% Percent
c. Circa (Approximately)
m Metres
mm Millimetres
N Newton’s
x
Acknowledgments
Secondly, I would like to acknowledge fellow student Gavin Irvine for the
previous research he conducted on this subject, which provided the basis
for the body of work presented here.
Finally, I would like to thank my sister, Claire, for proof reading my work.
xi
xii
Chapter 1
Introduction
1
tool to optimise the frame for earthquake performance requirements, before
embarking on more rigorous computer analysis. Previous modelling has
been carried out providing the foundations for this project.
• Analyse, using finite element methods and the elastic frame analysis
package “MASTAN2”, a wide range of framework geometries.
2
Chapter 2
2.1 Introduction
3
2.2.1 Regular and Irregular Frames
Regular frame structures are defined here as frames in which the basic struc-
ture does not change between floors (the number of bays and storeys remain
constant through out the piperack). Irregular frames on the other hand
refers to frames which include setbacks and irregularity in terms of the
number of bays and storeys, such as the piperack seen in Fig. 2.1.
Frames can also be defined as uniform or non-uniform to describe how
the storey parameters, such as the area, length, moment of inertia, and
Young’s modulus, change for the beams and columns within the frame.
In a uniform frame all storey parameters are identical for every storey.
In contrast, the parameters vary between storeys in non-uniform frames.
Therefore, with in this study regular and irregular frames can be either
uniform or non-uniform.
2.2.2 Bracing
An effective and economical way to improve the lateral force resistance of the
piperack is to include bracing members in the frame [4, 1]. Fundamentally
there are two types of bracing, concentric and eccentric.
Concentric Bracing
Eccentric Bracing
4
this bracing it is not as suitable for an approximate algorithm, such as the
one in this study.
shows how the lateral stiffness (KL ) can be thought of as the force (FL )
required to achieve displacement (δL ) in the lateral direction, or the force
per unit length. The lateral stiffness term itself is defined by Manohar and
5
Madhekar [8] for an elastic body as the “reciprocal of flexibility”, which again
relates the stiffness to the extent an elastic member will deform under some
form of loading.
6
Figure 2.2: Showing stiffness ratios (Ke /K∞ ) for a nine storey elastic frame: a) 9th
storey; b) 5th storey; c) 1st storey, adapted from Schultz [7].
In Fig. 2.2 the x-axis variables α and β are parameters describing the
relative stiffness of the storey beams and columns. Assuming the beams
and columns are made from the same material and have the same Young’s
modulus,
Ib /Lb
α= , (2.2)
Ic /Lc
and,
Ic /Lc 1
β= = . (2.3)
Ib /Lb α
Where; Lb , Ib and Lc , Ic are the lengths and moments of inertia for the
beams and columns respectively. It can be seen from Fig. 2.2 that the fifth
storey is the least stiff followed closely by the ninth storey. However, in
the first storey there is a noticeable increase in the stiffness, this is most
probably due to the fixed base. The first and ninth storeys are examples of
the boundary storeys described by Schultz as having higher stiffness than
the intermediate fifth storey.
This study forms the basic for subsequent work on approximating lateral
stiffness in elastic frames. The similaritys of Fig. 2.2 to a hyperbolic tangent
7
curve provided the methodological approach for Irvine [2] and this research.
Figure 2.3: Simplified elastic frame, combination of flexural and shear deformations [14].
8
The construction of a cantilever analogy allowed for the development of
the simplified approach to the limits of stiffness developed by Irvine and
described in the next chapter.
9
Chapter 3
3.1 Introduction
This chapter presents a review and critique of the previous approach of
Irvine [2] to help contextualise the method and approach taken in this dis-
sertation. Irvine’s methodology was inspired by Schultz’s work, in particular
the graphs shown in Fig. 2.2. Irvine recognised that Fig. 2.2 resembles a
hyperbolic tangent curve and could be used to approximate the response of
an elastic frame. The initial work determined the upper and lower limits for
stiffness. This lead to the development of an expression for displacement at
maximum stiffness. Following this, a hyperbolic tangent representation of
the ratio of frame displacements at maximum stiffness to actual stiffness,
similar to graphs presented by Schultz [7], was investigated. Irvine success-
fully a means to approximate the lateral displacement for regular elastic
frames via the hyperbolic tangent curve.
10
and that the columns are identical, the lateral load being spread evenly be-
tween the columns [2, 15]. Furthermore, this assumption also implies that
the columns do not transfer any moment into the beam [15].
Irvine summarised from this simple model that the storey lateral stiff-
ness,
12ENc Ic
KS = , (3.1)
L3c
can be approximated in the same way as the stiffness of a fixed-guided
cantilever beam with te load applied at the guided end. In Equation (3.1),
E is Young’s modulus of elasticity, Nc is the number of columns, Ic is the
moment of inertia and Lc is the length of the column (or the height of the
storey).
3.2.3 µ Effect
The µ effect, as named by Irvine [2], is the ratio between column moment
of inertia to the cross sectional area that causes significant errors in the
simplified model presented in the section 3.2.1. The µ effect is comparable
to the stress distribution in a fixed-free cantilever beam with a point load
at the free end, with one side being in tension while the other will be in
11
compression.
In the case of the model shown in section 3.2.1 the columns will act in the
same way as described by the stress distribution analogy made above. This
effect causes axial deformation in the columns that increases the maximum
lateral displacement of the frame. To compensate for this Irvine presented
the following term to be added to displacement calculations;
Pi L3i Pi L2i
∆µ = + (H − Li ), (3.3)
3EIA 2EIA
where,
Ac D 2
IA = , (3.4)
2
is the structures moment of inertia. In equations (3.3) and (3.4) above, Pi
is the lateral load acting to the storey, H is the total height of the frame ,
D is the distance between outer columns and Ac is the area of the columns
in the storey. Equation (3.4) also assumes there are only two columns
in a storey, this was not mentioned in the original text but was implied.
This term is effectively used to incorporate flexural displacement into the
maximum stiffness model. However, there may be a limitations to assigning
a moment of inertia term for the whole structure as it may not account for
any variations in the frame.
Ns
X
∆M ax = ∆storey + ∆µ , (3.5)
Ns =1
where ∆storey is the displacement of the storey calculated from (3.1). The
fully expanded ISM expression,
Ns
L3c L3i L2i
X
∆M ax = Ps,i + + (H − Li ) Pi , (3.6)
i=1
12ENc Ic 3EIA 2EIA
where Ps,i is the summation of the lateral loads acting on the specific storey.
Equation (3.6) presented in Irvine’s work assumes the storeys are uniform.
12
Furthermore, it is also assumed that all the storeys have the same number
of bay’s and hence columns. This is a limitation of this approach as it limits
the user to regular structures with identical storeys. Irvine recognised this
limitation and offers a worked example to explore how the ISM could be
corrected. However, no specific equation was presented.
y = a · tanh(b · (x − 1) + c) + d, (3.7)
e0.5Ns × e−0.3Nb Lc
a = 0.02Ns + 0.33 − , (3.8)
200 Lb
b = 0.30, (3.9)
× e−0.16Nb
0.06Ns
e Lc
c = ln + [0.14 ln(Nb ) − 0.1], (3.10)
2.4 Lb
Lc
d = 0.1 ln + 0.55e−0.03Ns . (3.11)
Lb
For the above equations Ns and Nb refers to the number of storeys and bays
respectively. These manipulation parameters were developed using manual
rudimentary curve fitting. However, this time consuming approach, while
involving considerable effort and precision, limits the number of examples
which can be analysed. Within the final equations it is difficult to determine
how the Ns and Nb where related to the fitted curve.
These parameter performed reasonably well when tested against three
different regular uniform frame geometries with all but one of the results
being c.10% of the correct value. The outlier of these results was for a two
storey, three bay frame which produced an error of 26% for a β value of 6.
13
3.5 Frame Analysis Approach
Irvine’s approach for analysing frame structures was to calculate the re-
sponse with Infinitely stiff beams for the whole structure. Following this,
a hyperbolic tangent curve was approximated using equations (3.7 - 3.11)
to represent the entire frame structure. The response of the frame could
then be acquired. This approach allows for fast calculations which appear
to be reasonable accurate and ideal for preliminary analysis. However, this
methodology requires the hyperbolic tangent curve to be calculable for the
whole structure in one iteration. Consequently, this approach effectively
constricts the use of (3.7) to purely regular structures with no variation
between storeys.
14
Chapter 4
Method
4.1 Introduction
This chapter presents the approach used in this study for the approxima-
tion of lateral stiffness. The approach presented here aims to further the
work done by Irvine [2] and derive a more general method for estimating
the response of elastic frames under uniform lateral loading. Firstly, the
infinitely stiff model (ISM) will be generalised to accommodate individual
storey parameters, irregular frames, and bracing. Following this, equations
will be derived for new manipulation parameters of the hyperbolic tangent
function. Finally, the methodological approach will be described in full.
Ns
X
∆Kmax = ∆storey + ∆µ ,
i=1
for elastic frames with infinitely stiff beams can be estimated by summing
the response of each storey. The equation presented in the previous study
(3.6) can easily be adapted to accommodate changes in section and geome-
ters between storeys. Accordingly, each storey is considered individually,
Ns
L3c,i L3i L2i
X
∆Kmax = Ps,i + + (H − Li ) Pi , (4.1)
i=1
12Ei Nc,i Ic,i 3Ei IA,i 2Ei IA,i
15
and
Ac,i D2
IA,i = , (4.2)
Nc,i
where the subscript i represents the current storey within the frame and IA,i
now represents the storey moment of inertia. From this set of equations,
when compared to (3.4) and (3.6), it can be seen how the variables now
allow for changes in column parameters for individual storeys.
4.2.1 Bracing
As briefly described in Chapter 2, bracing is used in elastic frame structures
to improve resistance to lateral forces. To include bracing it is suggested that
the force taken by the braces in each storey be calculated and subtracted
from the force applied to the storey. This method would allow for the use
of equation (4.1) in its current form. However, a more elegant way achieve
this is to include the bracing stiffness in the ISM. Here, the stiffness terms
for a number of braces can be calculated using the Unit Load Method [9].
This method leads to the following set of equations for the bracing shown
in Fig. 4.1, assuming the braces themselves are considered plane trusses;
EAbr L2b
Kbr1 = , (4.3)
(L2c + L2b )1.5
2EAbr L2b
Kbr2 = , (4.4)
(L2c + L2b )1.5
0.5EAbr L2b
Kbr3 = , (4.5)
(L2c + 0.25L2b )1.5
0.5EAbr L2b
Kbr4 = , (4.6)
(0.25L2c + L2b )1.5
where Abr is the cross sectional area of the bracing member. Equations
(4.3-4.6) represent bracing types a) through d) respectively in Fig. 4.1.
This approach has been adopted here as it leads to a more complete
form of the generalised ISM,
Ns Nbr
L3c,i
X X 1
∆Kmax = + Ps,i (4.7)
i=1
12Ei Nc,i Ic,i i=0
Kbr
L3i L2i
+ + (H − Li ) Pi ,
3Ei IA,i 2Ei IA,i
where Nbr is the number of bay’s in a storey with bracing, and Kbr repre-
sents the bracing terms presented in equations (4.3 - 4.6). Although, only
16
concentric bracing has been considered here, this approach to bracing could
be extended to any type of brace.
Figure 4.1: Bracing types; a) Diagonal brace, b) X-brace, c) Chevron brace and d)
K-brace
Initial Procedure
The computer program “MATLAB ” was used to create a finite element code
in order to analyse frame structures and produce the type of Tanh curves
shown in Fig. 2.2.
The first step was to use the code to calculate the displacement at the
maximum stiffness (∆Kmax ) for a ratio of column length to beam length.
Once this displacement had been calculated, the maximum frame displace-
ment (∆) at different α and β values in the range of 1 - 10 were determined.
As with Schultz and Irvine, this ratio of ∆Kmax /∆ provided the y-axis for
the hyperbolic tangent graphs.
The next step was to combine the α and β values together into the x-axis
of one graph. This was achieved by using the fact that α = β when α or
β equals one. Therefore, to produce the graphs x-axis, one was subtracted
from α and β, before making β negative. This overlapped the graphs to
17
create an x-axis equal to zero, when α and β are equal one. Hence, in
relation to α and β the x-axis in the Tanh graph is defined as,
α − 1, if α ≥ 1
x= (4.8)
−(β − 1), otherwise.
During the curve fitting procedure, b was kept at 0.3 while a, c and d
were subject to change for each Lc /Lb ratio. To fit the data points the
curve fitting suit in MATLAB was implemented. This involved the use
of non-linear regression to model the data with (4.9) using a method of
successive approximations to determine a, c and d for each Lc /Lb ratio.
Table 4.1 shows the manipulation parameters determined for the one bay,
five storey frame structure shown in Fig. 4.2. This process was repeated
for 30 different regular uniform frame geometries, with an upper limit of 9
bays and 5 storeys, to build up a comprehensive set of data.
Table 4.1: Showing the coefficients for various Lc/Lb ratio for 5 storey, 1 bay structure.
Lc/Lb a c d
10.000 0.353549728 0.249672494 0.607722684
5.0000 0.390580224 -0.0001626 0.518484787
3.3333 0.391487715 -0.079564833 0.492009181
2.5000 0.390324276 -0.112753663 0.480996729
2.0000 0.389407394 -0.129357078 0.475482589
1.6667 0.388791303 -0.138748533 0.472360276
1.4286 0.388376281 -0.144546698 0.470431302
1.2500 0.388088678 -0.148366792 0.469160034
1.1111 0.3878831 -0.151012509 0.468279601
1.0000 0.387731913 -0.152918644 0.467645472
0.9091 0.387617916 -0.154336505 0.467174037
0.8333 0.387530089 -0.155419328 0.466814284
0.7692 0.387461152 -0.156264762 0.466533685
0.7143 0.387406163 -0.156937375 0.466310726
1
The MATLAB Script used for this process can be found in appendix (ref).
18
Table 4.1: continued from previous page
0.6667 0.387361684 -0.157481224 0.466130722
0.6250 0.387325265 -0.157927192 0.465983375
0.5556 0.387269948 -0.158608206 0.465759066
0.5263 0.387248743 -0.158871598 0.465672648
0.5000 0.387230753 -0.159096798 0.465598978
0.4762 0.387215393 -0.159290867 0.465535699
0.4545 0.387202205 -0.15945931 0.465480975
0.4000 0.387172384 -0.159850079 0.465355048
0.3846 0.387164898 -0.159951584 0.465322672
0.3571 0.387152603 -0.160123333 0.465268357
0.3226 0.387139208 -0.160322163 0.465206516
0.3030 0.387132732 -0.160426178 0.465174818
0.2500 0.387119753 -0.160677995 0.465101356
0.2000 0.387115194 -0.160876094 0.465050387
0.9
0.8
0.7
0.6
/
Kmax
0.5
0.4
0.3
0.2
0.1
Data Points
Fitted Curves
0
-10 -8 -6 -4 -2 0 2 4 6 8 10
- ( - 1) -1
19
4.3.1 Coefficient Curve Fitting
y = P1 · xP2 + P3 , (4.10)
where x is the Lc /Lb ratio, y is the coefficient, and P1 , P2 ,and P2 are the line
fitting variables. However, the variables in (4.10) were found to be too sensitive
to errors, when relating them to Ns and Nb , and a more robust solution was
sought. It was found that all the coefficients could be approximated by a linear
polynomial,
y = P1 · x + P2 , (4.11)
where all variables have the same meaning as in (4.10). The figures below show
a representative sample of the line fit for coefficient (a) (Fig. 4.3), coefficient (c)
(Fig. 4.4), and coefficient (d) (Fig. 4.5). Although, this line did not fit the exact
values for the coefficients, as can be seen the figures, it was less sensitive to error.
Coefficient d for the three bay, three storey frame in Fig. 4.5 shows the greatest
divergence form (4.11). However, the variation in the coefficient is so small it
is almost negligible. The calculation of P1 and P2 for each of the coefficients
can then be related to the number of bays (Nb ) and storeys (Ns ) with the use
multidimensional curve fitting, as explained in Section 4.3.2.
Nb =3 , Ns = 3 Nb =4 , Ns = 3
0.376 0.38
Co-effecient -a
Co-effecient -a
0.374
0.375
0.372
0.37
0.37 data data
fitted curve fitted curve
0.368 0.365
0 1 2 3 4 5 0 1 2 3 4 5
Lc / L b Lc / L b
Nb =5 , Ns = 3 Nb =9 , Ns = 3
0.38 0.38
Co-effecient -a
Co-effecient -a
0.375
0.375
0.37
data data
fitted curve fitted curve
0.365 0.37
0 1 2 3 4 5 0 1 2 3 4 5
Lc / L b Lc / L b
Figure 4.3: Sample line fitting for coefficient (a), at 3 storeys high and various number
of bays.
20
Nb =3 , Ns = 3 Nb =4 , Ns = 3
0.02 0.035
data data
Co-effecient -c
Co-effecient -c
fitted curve 0.03 fitted curve
0.01
0.025
0
0.02
-0.01 0.015
0 1 2 3 4 5 0 1 2 3 4 5
L /L L /L
c b c b
Nb =5 , Ns = 3 Nb =9 , Ns = 3
0.05 0.07
data data
0.045
Co-effecient -c
Co-effecient -c
fitted curve fitted curve
0.065
0.04
0.035
0.06
0.03
0.025 0.055
0 1 2 3 4 5 0 1 2 3 4 5
Lc / L b Lc / L b
Figure 4.4: Sample line fitting for coefficient (c), at 3 storeys high and various number
of bays.
Nb =3 , Ns = 3 Nb =4 , Ns = 3
0.519 0.526
data data
fitted curve fitted curve
Co-effecient -d
Co-effecient -d
0.5185 0.525
0.518 0.524
0.5175 0.523
0.517 0.522
0 1 2 3 4 5 0 1 2 3 4 5
Lc / L b Lc / L b
Nb =5 , Ns = 3 Nb =9 , Ns = 3
0.532 0.542
data data
fitted curve fitted curve
Co-effecient -d
Co-effecient -d
0.53 0.54
0.528 0.538
0.526 0.536
0.524 0.534
0 1 2 3 4 5 0 1 2 3 4 5
Lc / L b Lc / L b
Figure 4.5: Sample line fitting for coefficient (d), at 3 storeys high and various number
of bays.
21
4.3.2 Multidimensional Curve Fitting
The final step was to relate variables P1 and P2 for each of the coefficients to
the number of bays and storeys in the frame geometry. As stated previously 30
different regular uniform frames were analysed, and as such 30 different P1 and
P2 values for each coefficient and frames were obtained.
It was found that P1 and P2 for the majority the coefficients (a, c and d) could
be represented by bivariate polynomials of varying degrees. In two instances this
was not the case, and alternative equations were created, see (4.12) and (4.15).
For coefficient a, P1 and P2 are represented by,
1
P1a =(−0.009216Ns3 − 0.003361) · (4.12)
eNs + eNb + 9.023Ns2
−8 2 2
− 4 × 10 · 3310000Ns − (801Nb + 63550Nb + 8050000)Ns − 1862750 ,
and,
−8
P2a = −1 × 10 · 167100Ns4 − (4231Nb + 2440000)Ns3 (4.13)
for coefficient (a). Fig. 4.6 show how equations (4.12) and (4.13) fit the data
points for both P1 and P2 to the number of bay’s and storeys in a frame.
22
Figure 4.6: Three dimensional surface fit, relating P1a and P2a to the number of bays
and storeys in a frame.
23
In a similar manner, P1 and P2 were developed for coefficients c and d. These
are represented by equations (4.15 - 4.18) below2 ,
e(0.3941Ns −0.2578Nb )
P1c = + 3.151 × 10−5 Nb2 − 8.82 × 10−4 , (4.15)
300.2
−8
P2c = −5 × 10 · 194600Ns3 − (5590Nb + 2242000)Ns2 (4.16)
−9
P1d = −5 × 10 · 1561Ns3 + (1876Nb − 46000)Ns2 (4.17)
−8
P2d = −1 × 10 · 688000Ns3 + (15140Nb − 8025000)Ns2 (4.18)
and,
Lc
d = P1d · + P2d . (4.20)
Lb
2
See Appendix B for Fig. B.1 and B.2, showing the fit of these equations to the data
points.
24
4.4 Individual Storey Approach
As previously discussed in chapter 3, Irvine [2] attempted to approximate the
response of the whole structure using one hyperbolic tangent curve once the dis-
placement at maximum stiffness was calculated. The limitations of this were
determined to be that only regular frame with identical storeys could be anal-
ysed.
For this study an individual storey approach was adopted, where the Tanh
curve is calculated to represent each storey. The response of the piperack can
then be build up with each storey, to find the maximum displacement at the top
floor. This method was used to include irregular structures such as the piperack
show in Fig. 2.1.
To implement this approach, firstly the ISM (4.7) is used to calculate the
storey drift3 for the upper limit of stiffness. The next step is to determine the
manipulation parameters used in the Tanh function for the storey. To do this,
the number of bays (Nb ) in the current storey along with the total number of
storeys (Ns ) are used. Furthermore, each storey is analysed separately using its
own α or β and its own Lc /Lb ratio, independent of the storeys above and below.
Essentially, each storey is analysed as an individual Nb × Ns frame of regular
uniform geometry. Once the response of each storey has been determined these
can be summed together to calculate the response of the whole structure. The
partially worked example below will be used to help illustrate this methodology.
3
Storey drift is defined as the displacement of a storey independent of the storeys
below.
25
Figure 4.7: Irregular non-uniform frame used for the illustrative example shown below.
Storey 1 - Nb = 5
Initially, the ISM is used to calculate the displacement at the upper limit
of stiffness.
40003
∆max,1 = 4000
12 × 210 × 103 × 6 × 3 × 107
40003 40002
+ + (13000 − 4000) 1000 = 0.5665mm
3 × 210 × 103 × IA,1 2 × 210 × 103 × IA,1
Following this, the coefficients for the Tanh curve are determined using the
storeys individual parameters.
Ib /Lb 1 Lc 4000
α= = 1.0667 β= = 0.9375 = = 0.8
Ic,i /Lc,i α Lb 5000
P1a = − 0.026, P2a = 0.3840, P1c = 0.0043, P2c = 0.0106, P1d = −0.0015, P2d = 0.5207
y =a · tanh(0.3 · (α − 1) + c) + d = 0.5325
∆1 =∆max,1 /y = 1.0638mm
This process is repeated for all the storeys until the response of each has
been calculated
26
Storey 2 - Nb = 5
Ac,2 D2 1000 × 250002
IA,2 = = = 1.0417 × 1011
Nc,2 6
30003
∆max,2 = 3000
12 × 210 × 103 × 6 × 1 × 107
7000 3 70002
+ 3
+ 3
(13000 − 7000) 1000 = 0.5477mm
3 × 210 × 10 × IA,2 2 × 210 × 10 × IA,2
Ib /Lb 1 Lc 3000
α= = 2.4 β= = 0.4167 = = 0.6
Ic,i /Lc,i α Lb 5000
P1a = − 0.0026, P2a = 0.3840, P1c = 0.0043, P2c = 0.0106, P1d = −0.0015, P2d = 0.5207
y =a · tanh(0.3 · (α − 1) + c) + d = 0.6758
∆2 =∆max,2 /y = 0.8104mm
Storey 3 - Nb = 5
Ac,3 D2 1000 × 250002
IA,3 = = = 1.0417 × 1011
Nc,3 6
20003
∆max,3 = 2000
12 × 210 × 103 × 6 × 1 × 107
90003 90002
+ + (13000 − 9000) 1000 = 0.1243mm
3 × 210 × 103 × IA,3 2 × 210 × 103 × IA,3
Ib /Lb 1 Lc 2000
α= = 1.6 β= = 0.625 = = 0.4
Ic,i /Lc,i α Lb 5000
P1a = − 0.0026, P2a = 0.3829, P1c = 0.0043, P2c = −0.0106, P1d = −0.0015, P2d = 0.5207
y =a · tanh(0.3 · (α − 1) + c) + d = 0.5929
∆3 =∆max,3 /y = 0.2097mm
27
Storey 4 - Nb = 2
Ac,4 D2 1000 × 100002
IA,4 = = = 3.333 × 1010
Nc,4 3
40003
∆max,4 = 1000
12 × 210 × 103 × 3 × 1 × 107
130003 130002
+ + (13000 − 13000) 1000 = 0.9512mm
3 × 210 × 103 × IA,4 2 × 210 × 103 × IA,4
Ib /Lb 1 Lc 4000
α= = 3.2 β= = 0.3125 = = 0.8
Ic,i /Lc,i α Lb 5000
P1a = − 3.07 × 10−4 , P2a = 0.3803 P1c = 0.0089, P2c = −0.0668, P1d = 0.0033, P2d = 0.4909
y =a · tanh(0.3 · (α − 1) + c) + d = 0.6978
∆4 =∆max,4 /y = 1.3632mm
Total Displacement
Once all the storeys responses have been calculated, they are summed to
obtain the maximum displacement for the whole frame (∆).
∆ = ∆1 + ∆2 + ∆3 + ∆4 = 3.447mm
Table 4.3: Comparison between frame analysis package “MASTAN2 ” and Individual
Storey Approach.
Maximum
Percentage
Displacement,
Difference %
∆ (mm)
Approximate Method 3.447
4.1
MASTAN2 3.596
28
Chapter 5
5.1 Introduction
This chapter will present the results and evaluation of the proposed method.
Firstly, the storey drift approximation will be investigated to ascertain the
accuracy that can be expected. From this analyses it was noted that the
highest errors occurred in the β range. The discussion in Section 5.3 con-
siders this effect in previous work, and its limited impact on the storey drift
approximation. The analysis of the complete method for regular and irregu-
lar non-uniform frames (Section 5.4) shows the effectiveness of this method.
The investigation begins by exploring the first floor correction, which was
outlined by Schultz [7] and found here to be necessary. This correction
stops the methods tenancy to underestimate the frame stiffness. Regular
and irregular frames were rigorously tested and it was determined that the
method was accurate and robust. Finally, bracing was investigated with it
inclusion in the ISM found to need a further correction. The inclusion of
bracing in the final Tanh model provides an interesting avenue for future
research.
29
frames are presented alongside a comparison with Irvine’s method to help
justify the considerably more complex equations used in this method.
Figs. (5.1 - 5.3) shows the ratio between the exact stiffness (Ke ) to the
approximate frame stiffness (Ka ) for various one, three, and five storey reg-
ular frames at different α and Lb /Lc values. It can be seen from Fig. 5.1
that the approximated stiffness using the Tanh coefficients presented here
is within c.5% of the actual frame stiffness. The accuracy of the approx-
imation seems to decrease as the number of storeys increases. The worst
approximation is in a five storey, two bay frame with a Lb /Lc of 0.5 showing
an overestimation of c.12%. The parameters with the most noticeable influ-
ence on the accuracy are the number on storeys and the α value. It is clear
to see in the figures that as α is decreased (or β increased) the approxima-
tions accuracy decreases. For α values above one, the approximations never
exceed c.5% of the actual stiffness. This trend will be discussed further in
the next section. On the other hand, the Lb /Lc ratio along with the number
of bays in the structure appears to have very little effect on the accuracy.
Figure 5.1: Showing Ke /Ka ratio for a number of one storey, multi-bay frames at
Lb /Lc ratio’s of 0.5, 1, and 2. Along with comparison Irvine’s method, Data points in
red are calculated using Irvine’s Tanh coefficients.
30
Figure 5.2: Showing Ke /Ka ratio for a number of three storey, multi-bay frames at
Lb /Lc ratio’s of 0.5, 1, and 2. Along with comparison Irvine’s method, Data points in
red are calculated using Irvine’s Tanh coefficients.
Figure 5.3: Showing Ke /Ka ratio for a number of five storey, multi-bay frames at
Lb /Lc ratio’s of 0.5, 1, and 2. Along with comparison Irvine’s method, Data points in
red are calculated using Irvine’s Tanh coefficients.
31
Comparing the results of this study for regular uniform frames to the
results using Irvine’s, the improvements in accuracy obtained by using equa-
tions (4.12 - 4.20), rather than (3.8 - 3.11) are clear. The worst approxi-
mation seen in this study is a stiffness overestimation c.12%, with the vast
majority of the results within 5% of the actual stiffness. Using Irvine’s
method, for α less than one, shows the error in many of the approximations
exceeds 30% and is out with the range of the graphs. The author believes
that Figs. (5.1 - 5.3) provide enough justification for the use of the longer
and more complex equations presented in this study.
32
This can be attributed to the accuracy of the Tanh curve achieved within
this study significantly reduces the β error. While the β error has not been
completely eradicated, it has been more than half when compared to Irvine
as shown in Figs. (5.1 - 5.3). The reduction of the β error allows this
method to analyse a wider range of frame structures to a reasonable degree
of accuracy.
33
Table 5.1: Showing non-dimensional parameters for nine-bay, five-storey non-uniform
frame.
Figure 5.4: a) Three storey, three bay frame modelled using original approach. b)
Three storey, three bay frame modelled using Ns equal to one for first floor.
Continuing in this vein, the value of Ns for the first storey was experi-
mented with for many frames to determine the following,
2
Determined by variable R is Table 5.1.
34
Figure 5.5: Showing three storey, three bay frame modelled using Ns = 1 × ζf s for the
first storey.
Fig. (5.6 - 5.8) shows the results of this study. In a similar manner
to previous sections the results are presented as a ratio between the exact
stiffness and the approximate lateral stiffness. From the graphs presented
35
Table 5.3: Showing non-dimensional parameters for nine-bay, two-storey non-uniform
frame.
it is clear to see how well the individual storey approach approximates the
lateral stiffness, with the majority of the approximations within 5% of the
actual stiffness. This is consistent with what can be seen in estimations in
Section 5.2. The least accurate result of c.10% can be seen in Fig. 5.7, at an
α of 1/6 and Lb /Lc ratio of 2. This lower limit of accuracy is again in keep-
ing with that predicted from the story drift approximation. As expected,
the Lb /Lc ratio seems to have very little effect on the approximation, only
slightly increasing the overall approximated stiffness for the frame.
The main result of increasing the base column stiffness is an increase the
β value in the first storey. This in turn increases the error in the approxi-
mation as discussed in Sections 5.2 and 5.3. This can be seen particularly
well in Fig. 5.7 and 5.8 for already high β values. However, the effects of
changing the base columns stiffness seems to have less effect as the number
of storeys is increased, this makes sense when considering the nature of the
approach used. The individual storey approach sums the response of each
storey and hence the largest response will have the greatest influence on the
error. Increasing the column stiffness will reduce the displacement of this
storey. Therefore, for higher storey structures, as the column stiffness is in-
creased the error due to the β error in the first floor will have less effect on
the overall response as the displacement of the upper floors will correct this
error. This error also seems to decrease with the number of bays, although
to a much less noticeable degree. This can be seen when comparing Fig.
5.7 and 5.8. Furthermore, by comparing these two frames it can be seen
that reducing the number of bays also seems to slightly increase the overall
36
approximated stiffness for all α values.
Figure 5.6: Showing normalised stiffness against Ic ratio for nine-bay, five-storey frame.
Where, a) shows Lb /Lc = 0.5, b) shows Lb /Lc = 1 , c) shows Lb /Lc = 1.5, d) shows
Lb /Lc = 2
Figure 5.7: Showing normalised stiffness against Ic ratio for nine-bay, two-storey frame.
Where, a) shows Lb /Lc = 0.5, b) shows Lb /Lc = 1 , c) shows Lb /Lc = 1.5, d) shows
Lb /Lc = 2
37
Figure 5.8: Showing normalised stiffness against Ic ratio for two-bay, two-storey frame.
Where, a) shows Lb /Lc = 0.5, b) shows Lb /Lc = 1 , c) shows Lb /Lc = 1.5, d) shows
Lb /Lc = 2
The first frame used to conduct this investigation, described in Table 5.5
is based upon the piperack shown in Fig. 2.1. Table 5.5 shows the non-
dimensional parameters used in this frame. Investigations will be carried
out in the same way as previously shown by changing the variables ‘R’, ‘S’
and α.
From Fig. 5.9 it can be seen that the model predicts the lateral stiffness
for this irregular frame well, within c.5% for all but one case. This trend
follow that of the previous section with the α frames providing the most
conservative approximation. For this frame the β values actually give the
most accurate estimation with none exceeding c.5%. The largest error can
be seen in the Fig. 5.9 a), with α equalling 3 presenting an underestimation
of 10%. However, this underestimation of stiffness will give a conservative
38
estimation which is preferable. This four storey structure also exhibits a
similar trend to the five storey, nine bay regular frame in Fig. 5.6 in regards
to the increasing Ic ratio. As the Ic ratio is increased the approximate
stiffness improves, this can be explained in the same way. However, the
effects will be greater due to the setback in the top floor causing the top
displacement to increase and suppress the β error in the first floor further.
Table 5.5: Showing non-dimensional parameters for five-bay, four-storey irregular frame
shown in Fig. 4.7.
Figure 5.9: Showing normalised stiffness against Ic ratio for five-bay, four-storey
irregular frame. Where, a) shows Lb /Lc = 0.5, b) shows Lb /Lc = 1 , c) shows
Lb /Lc = 1.5, d) shows Lb /Lc = 2
39
dimensional parameters for this frame are shown below in Table 5.6.
Figure 5.10: Showing normalised stiffness against Ic ratio for nine-bay, three-storey
irregular frame. Where, a) shows Lb /Lc = 0.5, b) shows Lb /Lc = 1 , c) shows
Lb /Lc = 1.5, d) shows Lb /Lc = 2
From inspection of Fig. 5.10 the approximate method estimates the re-
sponse of the frame well, all but one of the parameters tested shows less
than 10% error. That being said, this frame also presents the worst approx-
imation across all the frames tested here, with an overestimation in stiffness
of c.14%. This overestimation occurs when the beams are twice the length
of the columns, the Ic ratio is at 3.5, and the β value is 6. Based upon
what has previously been seen for low storey frames, at high Ic ratio and β
values, this type of error is to be expected.
40
5.4.4 Summary of Model
The majority of the result underestimate the stiffness to within 5% of the
actual value. This underestimation is preferable as it provides a more con-
servative approximation. The β error influence has been minimised, but not
neglected with the worse error of c.14% being at a β of 6.
The testing conditions used in this section were selected to provide a
more realistic evaluation of the method. The simplification of the beams
parameters remaining constant in every floor has been made. The beam
parameters are however represented in the α and β formulas. Therefore,
from Section 5.2 it can be concluded this method could easily handle beam
parameter varying between storeys.
Overall the Individual Storey Approach developed here has proven to
be a accurate and robust method for approximating the lateral stiffness in
non-uniform and irregular frames.
5.5 Bracing
A further deliverable of this study is to include braced frames into the
method. The preliminary analysis for this is presented here, with the suc-
cessful inclusion of bracing into the ISM and a more problematic inclusion
into the individual storey approach. The inclusion of bracing into the in-
dividual storey approach raised several challenges and avenues to further
research.
41
Figure 5.11: Ke /Ka ratio for various Infinitely stiff regular uniform braced frames,
Lb /Lc equal to one.
Figure 5.12: Ke /Ka ratio for various nine-bay Infinitely stiff regular uniform braced
frames, at different Lb /Lc ratios
42
effectively. The following term is presented as a IA bracing correction factor,
Lc 1 − Ns,i
µb = − , (5.2)
Lb Ns,i
where Ns,i is the current storey. This is to be applied to the (4.2) in the
following manner,
µb · Ac,i · D2
IA,i = . (5.3)
Nc,i
Fig. 5.13 and 5.14 show the results of applying µb to the structures
moment of inertia. The error in the ISM for bracing has been reduced to a
maximum of c.7% from c.30% which could be seen in Fig. 5.12.
Figure 5.13: Ke /Ka ratio for various Infinitely stiff regular uniform braced frames, at
Lb /Lc ratio equal to one, modelled using corrected ISM.
43
Figure 5.14: Ke /Ka ratio for various nine-bay Infinitely stiff regular uniform braced
frames, at different Lb /Lc ratios, modelled using corrected ISM.
these numbers are purely experimental. That being said, Fig. 5.16 shows
the results of these alterations. The results shown in this figure provide proof
that if new a, c, and d coefficients, or correction factors for the current ones
where developed, this method could be applied to braced frame.
44
Figure 5.15: Showing ratio Ke /Ka for braced nine-bay, two-storey frame, modelled
using Tanh curve developed in Section 4.3.
Figure 5.16: Showing ratio Ke /Ka for braced nine-bay, two-storey frame, modelled
using experimentally altered Tanh curve.
45
Chapter 6
The study presented here attempted to take the initial preliminary work of
Irvine [2] further and develop a more generally applicable design algorithm
for approximating lateral stiffness in piperack support systems. This work
had two main deliverables, which were the ability to include non-uniform
and irregular frames. To achieve these objectives, further work was devel-
oped from Irvine’s Tanh function approximation to develop an approach
that could be implemented on a wide range of frame geometries. This fur-
ther work resulted in the generalisation of the ISM and the development
of the Individual Storey Approach. The developed Individual Storey Ap-
proach uses the storey drift of the ISM and the specific storey parameters of
the floor to determine the actual storey drift. This approach works on the
basic principle of calculating the storey drift for each floor before summing
them up to get the overall response. After initial testing it was evident that
the first floor stiffness was not being modelled correctly and underestimat-
ing the stiffness. Further research enabled the development of a modelling
formula for the first storey. The results of this saw these main deliverables
met with great success, for non-uniform and irregular frames. The majority
of approximations were found to be within 5% of the actual stiffness, with
only a few results over 10%.
An additional deliverable of this study was to include bracing in the
approximate algorithm. This was implemented by including stiffness terms
for various bracing types into the ISM. This would set a new upper limit and
the Tanh function could be used to approximate a response. The inclusion
of bracing was only met with partial success. Initially it was found that the
ISM tended to overestimate the response of the infinitely stiff braced frame.
This was determined to be due to the µ term and a correction to this was
46
presented. The inclusion of bracing into the Individual Storey Approach
was less successful. The method was only found to work in the high α
range and collapsed in the β range.
Overall, the main deliverables outlined in the objectives were achieved.
The work conducted by Irvine provided the foundations for the approach
taken. Key to this was the identification of the Tanh like nature of the ratio
between the actual and infinitely stiff response of elastic frames.
47
References
[2] G. Irvine, “Developing a design guide for lateral stiffness of pipe rack
support structures,” April 2018, unpublished Dissertation, University
of Aberdeen.
48
[10] J. R. Benjamin, Statically Indeterminate Structures: Approximate
Analysis by Deflected Structures and Lateral Load Analysis. New York,
N.Y: McGraw-Hill, 1959.
[12] T. Eroğlu and S. Akkar, “Lateral stiffness estimation in frames and its
implementation to continuum models for linear and nonlinear static
analysis,” Bulletin of Earthquake Engineering, vol. 9, no. 4, pp. 1097–
1114, 2011.
49
Appendix A
50
34 %__________________Nodes and Elements_________________
35
36 [nx,ny,Ele,e,f,rf] = nxnyEle(Lb,Lc1,Nb,Ns,G_Floor,F_x);
37
38
39 %____________________________________________________________
40 % Infinitely Stiff Beam Analysis
41 %____________________________________________________________
42
43 [L,Theta,X,Y] = ELe_length(nx,ny,Ele);
44
45 for i = 1:length(Ele)
46 if Theta(i) == deg2rad(90)
47 A(i) = Ac; I1(i) = Ic;
48 elseif Theta(i) == 0
49 A(i) = Ab; I1(i) = Ib1;
50 end
51 end
52
53 GK = Global_K(Theta,Ele,nx,E,A,L,I1);
54 [Re,df,Kff]=solveK(GK,rf,f,e);
55 Max_d = df(length(df) − 2)';
56
57
58 %____________________________________________________________
59 % Max Displacement Analysis
60 %____________________________________________________________
61
62 for alpha = 0.1:0.1:10
63 in = round(alpha*10);
64
65 % Beam Moment of Inertia for Alpha Range
66 Ib2(in) = alpha*(Ic/Lc)*Lb;
67 Alpha(in) = alpha;
68
69 for i = 1:length(Ele)
70 if Theta(i) == deg2rad(90)
71 A(i) = Ac;
72 I2(i) = Ic;
73 elseif Theta(i) == 0
74 A(i) = Ab;
75 I2(i) = Ib2(in);
76 else
77 A(i) = A_br;
78 I2(i) = I_br;
79 end
80 end
51
81
82 % Generates the Global Stiffness Matrix
83 GK = Global_K(Theta,Ele,nx,E,A,L,I2);
84
85 %__________________Displacements (Alpha)__________________
86
87 [Re,df2,Kff] = solveK(GK,rf,f,e);
88 Tip_displacementA(in,t) = df2(length(df2) − 2);
89 Tip_d_ratio_A(in,t) = Max_d/Tip_displacementA(in,t);
90 end
91
92 for beta = 0.1:0.1:10
93 r = round(beta*10);
94
95 % Beam Moment of Inertia for Beta Range
96 Ib3(r) = (1/beta)*(Ic/Lc)*Lb;
97 Beta(r) = beta;
98
99 for i = 1:length(Ele)
100 if Theta(i) == deg2rad(90)
101 A(i) = Ac;
102 I3(i) = Ic;
103 elseif Theta(i) == 0
104 A(i) = Ab;
105 I3(i) = Ib3(r);
106 else
107 A(i) = A_br;
108 I3(i) = I_br;
109 end
110 end
111
112 % Generates the Global Stiffness Matrix
113 GK = Global_K(Theta,Ele,nx,E,A,L,I3);
114
115 %__________________Displacements (Beta)__________________
116
117 [Re,df3,Kff]=solveK(GK,rf,f,e);
118 Tip_displacementB(r,t) = df3(length(df3) − 2);
119 Tip_d_ratio_B(r,t) = Max_d/Tip_displacementB(r,t);
120
121 end
122 end
123
124 % Creates Augmented Alpha/Beta X axis and Y axis
125 [x_AB, y_AB] = CoordAB(Alpha,Beta,Tip_d_ratio_A,Tip_d_ratio_B);
126
127
52
128
129 %____________________________________________________________
130 % Output Analysis
131 %____________________________________________________________
132
133 coef = zeros(t,3);
134 for q = 1:t
135 % Tanh Curve Fitting
136 y = fittype('a*tanh(0.3*(x)+c)+d','dependent',{'y'},'independent',{
'x'},'coefficients',{'a','c','d'});
137 myfit = fit(x_AB',y_AB(:,q),y,'StartPoint',[0.3 0.184720424
0.499244691]);
138
139 % Tanh Manipulation Parameters
140 coef(q,:) = coeffvalues(myfit);
141 end
53
Appendix B
Multi
Figure B.1: Three dimensional surface fit, relating P1c and P2c to the number of bays
and storeys in a frame.
54
Figure B.2: Three dimensional surface fit, relating P1d and P2d to the number of bays
and storeys in a frame.
55
Appendix C
56
57