Professional Documents
Culture Documents
Aalborg University
School of Civil Engineering and Science
MSc Structural and Civil Engineering, 1st semester Group B219 Fall 2013
School of Engineering & Science Sohngrdsholmsvej 57 DK-9000 Aalborg st Telephone (+45) 9940 8530 http://www.ses.aau.dk
Title: Analysis of a steel beam with holes Theme: Analysis and design of load-bearing structures Project period: MSc 1st Semester Fall 2013 Project group: Group B219 Members: Liuba Agapii Jonas Sneideris Paulius Bucinskas Nina Korshunova Alexandru-Iulian Radu Nicolas Martinez Almario Supervisor: Christian Frier No. printed copies: 6 No. of pages: 104 Completed: 16 Dec 2013
Synopsis:
Before a structure can be manufactured engineer has to evaluate the results received from different types of calculations. This report consists of analytical, numerical and experimental analysis of a beam perforated with holes. The main focus of this paper is to assess the behavior of the beam in linearelastic area in two different loading cases. In the analytical part the beam is analyzed in accordance to the following beam theories: Bernoulli-Euler and Timoshenko. These two approaches are compared and the result of it is the basis for calculation of the necessary values. The experimental analysis gives real results of the test object. Firstly the material properties are tested and the obtained results are used in analytical and numerical parts. Secondly the beam is tested and after processesing the received measurements, the results are compared with calculations. The numerical analisys is performed by using our own written code in Matlab R2013b and by commercial software ABAQUS CAE v.6.12. The following models were created: two dimensional (2D) shell and three dimensional (3D) shell and solid. 2D-shell model analisys was performed both in Matlab and Abaqus, also 3D-shell and solid models were analysed in Abaqus. The conclusion is made out of the comparison of three different approaches in analyzing the beam.
The contents of this report are freely accessible, however publication (with source references) is only allowed upon agreement with the authors.
Preface
The report was prepared by the group B219 at 1st semester of MSc program Structural and Civil Engineering at Aalborg University. The project was completed in the module Analysis and design of load-bearing structures under supervision of Christian Frier. The paper was handed in December 2013. This report is based on the following courses: Material Modeling in Civil Engineering and Structural Mechanics and Dynamics. The main aim is to learn how to apply different approaches in solving the problem and comparing the obtained results. The group would like to thank the supervisor of the project Christian Frier and all members of the group for hard work during the semester.
Reading guide The Harvard system of referencing is used in this report. Through the report in brackets the name of the author and year of publication with page is sited (f. x. Lars Damkilde, 2013, p.5). The material that was read and not quoted can be found after the report in the bibliography. Here the references are written in the following order with full information: author (in case of more than one author it is stated the rst name and et al.), year of publication, title, the edition. For a paper from the Internet it is written as following: author, year of publication, title, designation, name of institution submitted, the source which was the basis for writing the chapter of the project is stated before this chapter. The numeration of tables, gures and equations starts with the number of the chapter and continue through the appropriate part. The appendix is numbered in the same manner. All calculations presented in the appendix are done by our group.
Notations
Symbol A Av b E F G I Kt M N Q q S t u w z Description Area of cross section Shear area Width of the cross section Elastic modulus Concentrated force Shear modulus Moment of inertia Stress consentration factor Moment Normal force Shear force Distributed load Static moment of cross section Thickness Projection of displacement to x-direction Projection of displacement to y-direction Distance from neutral axis Rotation Shear correction factor Shear strain Normal strain Total cross section rotation Curvature Possions ratio Normal stress Shear stress Shear angle
Table Of Contents
Table Of Contents Chapter 1 Introduction 1.1 Analitycal Analysis . . . . . . . . 1.2 Experimental Analisys . . . . . 1.3 Numerical Analisys . . . . . . . 1.4 Comparisons and Conclusions 5 9 10 10 10 10 11 11 11 12 17 17 18 19 20 23 23 25 25 25 26 27 30 31 32 32 38 39 39 41 43 43 46
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
Chapter 2 Beam characteristics and Model 2.1 Beam description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Beam Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Cross section characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Chapter 3 Material Properties 3.1 Specimen set-up . . . . . . . . . . . . . 3.2 Strain gauge set-up . . . . . . . . . . . 3.3 Spicemen test . . . . . . . . . . . . . . 3.4 Material properties and data analysis
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
Chapter 4 Analytical part 4.1 Simplications and boundary conditions . . . . . . . . . . . . . . . . . 4.2 Bernoulli-Euler beam theory . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2 Constitutive equation and section integration . . . . . . . . . . 4.2.3 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.4 Differential Equation . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Timoshenko Beam Theory . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Kinematics conditions . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2 Constitutive equation . . . . . . . . . . . . . . . . . . . . . . . . 4.3.3 Equlibrium condition . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Comparison between Bernoulli-Euler and Timoshenko beam theories 4.5 Stresses related to the cross section . . . . . . . . . . . . . . . . . . . . . 4.5.1 Stress concentration factor . . . . . . . . . . . . . . . . . . . . . 4.5.2 Calculation of stresses . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
Chapter 5 Experimental Part 5.1 Beam set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Rosette strain gauges set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
5.3 5.4
49 51 53 53 53 53 53 54 56 57 58 59 60 61 63 66 69 73 73 75 80 83 85 85 86 88 88 89 92 94 94 103
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.1 6.2.2 6.2.3 6.2.4 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Applied load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Supports of the beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Symmetry conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2D Matlab model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3D Abaqus models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2D Matlab model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3D Abaqus shell model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3D Abaqus solid model . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.3
6.4
6.5
Bibliography A Appendix A.1 A.2 A.3 Diagrams for the shear force and the moment for both loading cases . . . . . . . Calculation of moment of inertia and area . . . . . . . . . . . . . . . . . . . . . . . Finite element theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.3.1 A.3.2 A.3.3 A.3.4 A.4 FEM - Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . FEM - Types of elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . Finite element calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . FEM - Plane strain and plane stress . . . . . . . . . . . . . . . . . . . . . .
B Digital Appendix B.1 B.2 B.3 B.4 B.5 B.6 B.7 B.8 B.9
Calculation of moment of inertia and shear area . . . . . . . . . . . . . . . . . . . 103 Calculation of coefcients from material test . . . . . . . . . . . . . . . . . . . . . 103 Calculation of displacement of the beam using Bernoulli-Euler theory . . . . . . 103 Calculation of displacement of the beam using Timoshenko theory . . . . . . . . 103 Comparison of beam theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 Calculation of normal stresses and shear stresses . . . . . . . . . . . . . . . . . . 103 Main test calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 2D MatLab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 2D Abaqus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Introduction
In building construction beams with perforated holes are used often. The main advantage of this sort of beams is that ventilation, heating, plumbing, electicity and other communications can be installed through the holes of the beams, saving usefull space (Figure 1.1)
For this type of beams a simple analytical solution to nd deection does not exist. Beam theories are used for calculations and analysis of the beam behaviour. Later more advanced approach are inplemented in numerical models. Finally, an experiment is performed and obtained results are compared to the calculations. The main aim of the project is to nd out how much the holes in the web affect the overall behaviour of the beam. Also to compare different calculation methods with the results from experimental and numerical parts. 9
10
Figure 2.1. Analysed beam elevation (HEA-140 steel prole) with the 82mm diameter holes.
Case 1: The beam is loaded with two forces (F) applied symmetrically and with a span of 1000 mm between them (Figure 2.2). The F forces are both equal to 18 kN ( this value is determined in chapter 4.5.1 ) and represented by two concentrated forces.
Case 2: The beam is loaded with two forces (F) with intensity of 11 kN (this value is determined in chapter 4.5.1) and applied symmetrically and with a span of 240 mm between them Figure 2.3.
12
Table 2.2. Data and characteristics of a HEA-140 steel prole (prole without interior curves).
Figure 2.6 represents the cross section of an HEA-140 without curvatures and entire hole. It is placed on the beam elevation, in the center of each hole and practically could represent a critical section in the calculation model.
13
Figure 2.5. The assumed cross section of a HEA-140 steel prole (prole without interior curves).
Figure 2.6. Cross section of a HEA-140 steel prole (prole without interior curves and with the 82 mm diameter hole).
Table 2.3 show data and other features of the above pictured cross section, with the entire hole. The average moment of inertia along the beam is 1, 4639 107 mm 4 . This value was calculated using Matlab function which nds value of moment of inertia for every x (xdistance from the start of the beam till the considered cross section) and average all obtained numbers(Figure 2.7). All calculations can be found in Appendix A.2 and Digital Appendix B.1.
14
Table 2.3. Data and characteristics of a HEA-140 steel prole (prole without interior curves and with the entire hole).
Moreover, the same calculations were made for the whole and shear area of the cross section.
15
Material Properties
In order to obtain better results from analytical, numerical and experimental model, it is important to use real material properties, determined via an experiment. The purpose of this chapter is to determine Youngs modulus, Shear modulus and Poisson coefcient. Tensile test is required to nd these material properties.
Figure 3.1. Photo of how acetone is used to remove the residual dust
Figure 3.2. Photo of how cyanoacrylate glue is used to mount one of the strain gauges
17
Strain gauges are now mounted with a cyanoacrylate glue. The contact surface between the strain gauges and steel should be without any air pockets, otherwise the results would be affected.
Figure 3.3. Photo of how the cables are attached to the strain gauges
18
released. New internal structure of particles creates some extra strength in the material. Load is further increased. Finally test specimen breaks with a ductile fracture, and the the collapsing force is registered. This type of fracture is very well known for steel structures and is also called the warning breaking due to the visible deformations.
20
After data interpretation, an average maximum normal stress was determined x = 321.705[N /mm 2 ] which is close to table value. x = N A (3.4.0.1)
where x is the normal stress, N is the normal force and A is the area of the cross section.
E=
x x
(3.4.0.2)
y x
(3.4.0.3)
G=
E 2 (1 )
(3.4.0.4)
Table 3.2. The average value for Youngs module and Poissons ratio for performed tests
21
22
Analytical part
We neglect the weight of the beam in analytical part calculations. The real cross section has curvatures. We simplify the geometry by changing the curvatures into right angles. We model the beam with simple supports and two loads in two different loading cases. It is impossible to take into account all possible loading cases which can occur in actual conditions as we have limited laboratory equipment. In static system the overhanging parts of the beam are neglected. There is no moment and no shear force in these ends. We also consider only linear elastic behaviour of the beam. It was done to avoid yielding. Dynamic behaviour is not analyzed in this project. Buckling is not considered. Temperature and atmospheric conditions are neglected. Moreover, we simplify the model and calculate strain in x-direction and displacement in z-direction. In real conditions loads and supports are distributed, in our case it is assume that they are point loads and point supports. The assumptions applied for each beam theory will be described later.
To nd constants in differential equations in beam theories, boundary conditions needs to determine. The static model and equilibrium equations are stated in 4.2 and 4.3 of this report. The dened kinematic boundary conditions are based on continuous deformations. The beam is divided in 3 elements: 1. 0 x a mm 2. a x b mm 3. b x L mm 23
To the simply supported beam (Figure 4.1) the kinematic boundary conditions are dened as:
w 1 (x ) = 0, if x = 0 w 3 (x ) = 0, if x = L w 1 (x ) = w 2 (x ), if x = a w 2 (x ) = w 3 (x ), if x = L - b 1 (x ) = 2 (x ), if x = a 2 (x ) = 3 (x ), if x = L - b
(4.1.0.1)
where w is deection in z-direction; is rotation of the corresponding cross section. The direction of the displacement and rotation is shown in Figure 4.2.
24
In order to be able to solve the given problem we have to set the static boundary conditions. We have pinned support at x=0 and roller support at x=L . The considered problem is statically determined. The diagrams for moment and shear force distribution for both loading cases are described in chapter 2.2 can be found in Appendix A.1. The chosen signs direction are presented in Figure 4.3
Figure 4.3. Equilibrium for innitesimally small beam element with distributed load, q Haukaas [2013a].
4.2.1 Equilibrium
The equilibrium equations are based on the following considerations: 1) equilibrium in the xdirection for the innitesimal beam element; 2) distributed load, q, acts on opposite direction to z-axis Figure 4.3. According to Figure 4.3 the equilibrium equations can be written down. Vertical equilibrium yields: Q qd x (Q + dQ ) = 0 dQ = q dx (4.2.1.1)
where normal stress; E Youngs module; strain. Notice that Eq 4.2.2.1 is based on two-dimensional elasticitys theory and is called plane stress material law.Haukaas [2013a]. Plane stress material law in x-direction can be written down:
x x = E x x = E
d2 w z dx 2
(4.2.2.2)
Axial stresses over the cross-section are integrated using formula below: M= z d A (4.2.2.3)
4.2.3 Kinematics
Strains in the cross-sections can be computed using Naviers hypothesis for beam bending(Figure 4.4).
du dx
(4.2.3.1)
where u is displacement. Eq. 4.2.3.1 means that displacement u (elongation or shorten of upper or lower layers of crosssection) is divided by original length x. According Figure 4.4 it is obvious that displacement u is related with the rotation of cross-section d . Under the assumption of Bernoulli-Euler beam theory (during the bending the plane sections and cross sections remain plane and perpendicular to the neutral axis) the axial displacement of each layout in the cross-section is expressed in formula: d u = z d 26 (4.2.3.2)
Also according Figure 4.4 the rotation of the cross-section is dened as:
dw dx
(4.2.3.3)
Combining Eq. 4.2.3.2 and Eq. 4.2.3.3 the axial displacement is obtained: u = z dw dx (4.2.3.4)
Finally, using equations mentioned above in sub-article 4.2.3 Kinematics the kinematic equation for beam members is obtained:
du d d2 w = z = z dx dx dx 2
(4.2.3.5)
Innitesimal beam part in the Figure 4.4 is curved because of the bending moment effect. The curvature in Bernoulli-Euler beam theory is dened as:
d d2 w dx dx 2
(4.2.3.6)
(4.2.3.7)
q =
d2 M dQ d2 = = dx dx 2 dx 2
z d A =
d2 dx 2
E z d A =
d2 dx 2
d2 w 2 d4 w z d A = E I dx 2 dx 4 (4.2.4.1)
where E - elasticity modulus is assumed constant over the cross-section; Moment of inertia is dened:
Iy =
z 2d A
A
(4.2.4.2) 27
Since the boundary conditions are specied it is more convenient to use the following formulas. Shear force: Q =E I d3 w dx 3 (4.2.4.3)
Bending moment:
M =E I
d2 w = E I dx 2
(4.2.4.4)
dw dx
(4.2.4.5)
d2 w M = 2 dx EI Integrating Eq. 4.2.4.6 yields to Eq. 4.2.4.7, Eq. 4.2.4.8 and Eq. 4.2.4.9:
(4.2.4.6)
dw 1 = dx
M1 dx + C 1 EI
(4.2.4.7)
dw 2 = dx
M2 dx + C 2 EI
(4.2.4.8)
dw 3 = dx
M3 dx + C 3 EI
(4.2.4.9)
The displacements of the neutral axis of the beam are dened using Eq. 4.2.4.7, Eq. 4.2.4.8 and Eq. 4.2.4.9:
w 1 (x ) =
M1 dx dx + C 1 x + C 4 EI
(4.2.4.10)
w 2 (x ) =
M2 dx dx + C 2 x + C 5 EI 28
(4.2.4.11)
w 3 (x ) =
M3 dx dx + C 3 x + C 6 EI
(4.2.4.12)
Constants C 1 , C 2 , C 3 , C 4 , C 5 , C 6 are determined by using boundary conditions described in Eq.( 4.1) and using computer program Matlab. The calculations and results are presented in digital appendix B.3. On the following charts (Figure 4.5, Figure 4.6) the resulting curve of the displacement is shown for different loading cases.
In Table 4.1 it is shown the maximum value of the displacement using different values of moment of inertia It is clear that the difference between the received results is relatively small. Between the average moment of inertia and the minimum value it is 1.96% . So, it can be concluded that there is no need to use the moment of inertia like a function, because it makes calculations too complicated. Also, in comparison with the rst loading case, the second way of loading decreases the value of the moment and also leads to 7,84% lower value for the displacement.
29
According Figure 4.7, the relation between displacement and rotation is dened in Eq.( 4.3.1.1).
dw dx
(4.3.1.1)
Also, based on Figure 4.7 the relation between the shear angle and the rotation can be written as:
= + =
dw + dx
(4.3.1.2)
Normal strain:
x x =
du dx
(4.3.1.3)
x x =
d2 w d + z dx dx
(4.3.1.6)
x z =
(4.3.1.7)
Substitution of the normal strain and shear strain equations in Hookes law yields stress x x and shear stress x z equations.
x x = E x x = E
d2 w d + z dx 2 dx
(4.3.2.2)
(4.3.2.3)
(4.3.2.4)
x x z d A =
d2 w d + E z2 d A dx 2 dx
(4.3.3.1)
Q=
x z d A =
G d A
(4.3.3.2)
Shear is dened area as Av =v A , where A is the actual area and v is an shear correction factor.
d w d A = d wv Q
(4.3.3.3)
( d w ) t d A = ( d x ) Q 32
(4.3.3.4)
v dx d A = Q dx G G
(4.3.3.5)
Furthermore, on the right-hand side the average shear stress is written in terms of the shear Q force on the cross-section, i.e., v = Av , where A v is an auxiliary shear area that is dened as:
1 Q dx d A = Q dx G G Av
(4.3.3.6)
QS , Ib
(4.3.3.7)
1 QS dx G Ib
d A =
1 Q Q dx G v A
(4.3.3.8)
v =
I2 A
S 2 A b dA
(4.3.3.9)
Aw e b A
(4.3.3.10)
33
According to the Figure 4.8 it is assumed that A w e b is shear area for the I-beam prole. The bending moment M(x) and the shear force Q(x) can be written following:
M=
d2 w d + EI dx 2 dx
(4.3.3.11)
Q=
dM d3 w d2 = + EI dx dx 3 dx 2
(4.3.3.12)
Q = G Av Knowing the shear force in the beam Q(x), the distributed load can be found q(x):
(4.3.3.13)
q =
d d 1 dQ = G A v = q dx dx dx G Av
(4.3.3.14)
q=
d4 w d3 EI + dx 4 dx 3
(4.3.3.15)
The equation for the quasi static bending of a linear elastic, isotropic, homogeneous beam of constant cross-section beam is:
d4 w 1 d2 q q = + dx 4 G A v dx 2 E I
(4.3.3.16)
34
The bending moment, M and the shear force, Q can be identied through equilibrium conditions:
M d2 w d = + EI dx 2 dx
(4.3.3.17)
Q = G Av
(4.3.3.18)
(4.3.3.19)
d2 w M dQ 1 = 2 dx E I dx G A v Integrating Eq. 4.3.3.20 yields to Eq. 4.3.3.21, Eq. 4.3.3.22 and Eq. 4.3.3.23:
(4.3.3.20)
1 dw 1 = Q1 d x + dx G Av dw 2 1 = Q2 d x + dx G Av
M1 d x + C1 EI
(4.3.3.21)
M2 d x + C2 EI
(4.3.3.22)
dw 3 1 = Q3 d x + dx G Av
M3 d x + C3 EI
(4.3.3.23)
The displacements of the neutral axis of the beam, w(x) are dened as following:
w 1 (x ) =
Q1 d xd x + G Av
M1 d xd x + C 1 x + C 4 EI
(4.3.3.24)
w 2 (x ) =
Q2 d xd x + G Av
M2 d xd x + C 2 x + C 5 EI
(4.3.3.25)
w 3 (x ) =
Q3 d xd x + G Av
M3 d xd x + C 3 x + C 6 EI 35
(4.3.3.26)
1 ( x ) =
M1 d x + C1 EI
(4.3.3.27)
2 ( x ) =
M2 d x + C2 EI
(4.3.3.28)
3 ( x ) =
M3 d x + C3 EI
(4.3.3.29)
The constants C 1 , C 2 , C 3 , C 4 , C 5 , C 6 can be determined using boundary conditions (Chapter 4.1, Eq. 4.1, Eq. 4.3.3.27, Eq. 4.3.3.28 and Eq. 4.3.3.29). It was done using Matlab. The calculations and results are presented in the digital Appendix B.4. On the following chart (Figure 4.9 and Figure ??) the resulting curve of the displacement is shown.
36
In the Table 4.2 it is shown the maximum value of the displacement using different values of moment of inertia.
It can be concluded that the difference between minimum and average value becomes signicant in this theory (16.82%). Therefore it was decided to use minimum values for nal results to be on the safe side. Also, the displacement is lower in the second case of 2.37% than in the rst one.
37
Figure 4.11. The displacement of the beam calculated by Timoshenko and Bernoulli-Euler theory for both of the loading cases(minimum value of cross section properties is used).
In the Figure 4.11. it is ploted charts by which the results can be compared. In the Table 4.3 it is shown the value of the displacement for minimum value of moment of inertia (the worst case) and for different cases of loading.
The difference in two theories is presented in both loading cases. It can be explained as following. In Timoshenko beam theory the shear force is taken in to consideration. Timoshenko beam theory will provide bigger displacement for short and thick beams, where the shear force will have an important inuence. In our case, the Timoshenko beam theory has signicant contribution in rst loading case on the displacement. To prove the conclusion that 38
difference between these two theories becomes bigger because of the decreasing of the length the following chart was done. (Digital Appendix B.5)
Figure 4.12. The dependence of the difference between theories from the length of the beam.
For convenience it was considered to make the comparison between two theories with the loading case when the span between the forces is 1 mm. All calculations were made with the minimum value of moment of inertia for the beam. The length of the beam was varied from 0 to 1870 mm.
39
Calculations of the stress concentration factor are very complicated. But if some assumptions are made an appropriate formula for calculating the stress concentration can be found. The choice of the formula depends on the loading case, the type and dimensions of the hole. However, it is important to note that K t is an ideal value based on linear elastic behaviour and it only depends on the proportions of the dimensions of the stress raiser and the notched part. To calculate the stress concentration factor in our project, the formula from the table is used. Pilkey & Pilkey [2008]. Two types of calculation is done. The rst one is for a single circular hole in an innite plate (Figure 4.14)
Figure 4.14. Single circular hole in a innite plate in bending, m 1 and m 2 - bending moments Pilkey & Pilkey [2008].
In our case the element is under the simple bending, so the following formula can be used:
K t = 3.000 0.947
d d d + 0.129 , 0 7.0 t t t
(4.5.1.1)
where d - the diameter of the holes, t - thickness of the plate. Calculating with the dimensions the result for stress concentration factor is 1.46. The second case is for single row for circular holes in innite plate (Figure 4.15)
K t = 1.787 0.060
d d 0.785 L L
+ 0.217
d L
,0
d 1.0 L
(4.5.1.2)
where L - the distance between the center of the holes. So, K t = 1.746 The different approaches to calculate the stress concentration factor was used in order to compare these results with the results from the numerical part. After determining the stress concentration factor the maximum value of the force without having the yielding in the cross section was dened. The maximum allowed stress for the weakened cross section: y i e l d Kt
(4.5.1.3)
Figure 4.15. Single row for circular holes in innite plate in bending Pilkey & Pilkey [2008]
M m a x = m a x Ww e a k e n e d = 191.87 145859 = 27.99kN m The maximum value of allowed force can be dened from the following formula:
(4.5.1.4)
1 27.99 1 a2 Mm a x M m a x = F (l a )+ F = = 55.77kN (4.5.1.5) Fm a x = a2 12 2 8 l 0.5(l a ) + 8l 0.5(1.87 1) + 81.87 For the second loading case: Fm a x = Mm a x 0.5(l a ) +
a2 l
27.99
0.24 0.5(1.87 0.24) + 8 1.87
2
= 34.18kN
(4.5.1.6)
Both values were divided by the safe factor of 3. The force for rst loading case in our project was determined 18kN , for the second one - 11kN .
M z I
(4.5.2.1)
minus sign appears because it is compressive (negative) stresses in the positive z -axis domain that gives a positive bending moment, i.e., bending moment with tension at the bottom. Shear stresses were calculated using Juravski formula (Eq. 4.3.3.7). All the calculations can be found in Digital Appendix B.6. 41
Experimental Part
In order to compare results with the analytical and numerical models, an experimental test has been performed. Displacements and strains are subjects ware measured at specic and precisely positioned points. This chapter refers to effects induced by loads applied to a HEA-140 prole beam. Holes are a subject to a special stress distribution case; thats why, unidirectional and rosette strain gauges measure the strains in the inuenced area, around a hole.
Figure 5.1. Beam set-up - position of loading cases (it was considered one case at a time)
The Figure 5.1 shows both loading cases, loading case number 1 is dotted in the drawing. The tested beam has a cross section HEA-140, an original length of L o r g = 1940[mm ] and a length between the supports of L = 1870[mm ]. Holes are cut using the same technology of CNC water-jet cutting machine. Supports are made of a steel roller with a diameter of 15[mm ]and a length of 140[mm ], beneath this roller a steel plate of 30 25[mm ] by 140[mm ] is placed. Supports base is a steel plate, with the following dimensions: 80 25 140[mm ]. All three components are clamped together to form the roller support. The supports are placed 35[mm ] from the each end of the beam. In terms of structural engineering, a roller support means a free rotation around its axis and restrained displacement on x-direction (x-x direction represents 43
the axis which is orthogonal to the deformed beam axis). Both ends of the model have the same supports as illustrated in the picture below (see Figure 5.3)
Two applied loads, which basically represent two concentrated forces for each loading case, are the only external factors for which model is analysed. Two loading cases are analysed and two tests have been conducted for each of loading cases (Figure 5.1). In both loading cases, vertical forces are applied via manual hydraulic jacks. In the analytical calculations, forces are considered to be concentrated. But in reality a force can not be applied in a point, it always is distributed to a surface, which can only simulate a concentrated force. Each force is applied by the help of a sandwich made of two steel plates, two plastic plates and a grease core. The resulting surface between steel plate and the top ange is 50[mm ] by 140[mm ].
44
The entire sub assembly can be seen in the picture above (Figure 5.3). Each loading case contains two forces [F ] witch are equal to each other, rst case has F equal to 18[kN ], while in the second- F is equal to 11[kN ]. Both cases have the forces applied symmetrical from the beam center as follows: 500[mm ] between the forces in rst the situation and 120[mm ] in the second.
45
Figure 5.6. Beam set-up - B219 group strain gauges and rosettes placed around 4th hole (beam segment is positioned up side down for a better illustration of Ch 14 and Ch 15 - Segment C-C)
46
Leng t h (a b ) =
2 2 d x + 2 u3 + u3 + u4
Leng t h (a b ) d x + u 3
(5.2.0.2)
The normal strain in the x direction of the rectangular element is dened as: x = ext ensi on l eng ht (a b ) l eng ht (ab ) u x = = or i g i nal l eng ht l eng ht (ab ) x (5.2.0.3)
y =
u y y
(5.2.0.4)
The shear strain is dened as the change of the angle between and :
x y = +
(5.2.0.5)
tan =
u4 d x + u3
(5.2.0.6)
tan =
u2 d x + u1
(5.2.0.7)
x y =
u y x
u x y
(5.2.0.8)
Wikipedia [2013]
47
Normal strains are dened in the material properties part, stresses in x-x, y-y and x-y directions, are evaluated using Hookes law (for plane stress).
E x 2 1 y = x y 0
E 12
0 x 0 y G x y
(5.2.0.11)
(5.2.0.12)
x = [(x + y ) E ]/(1 2 ) 48
(5.2.0.13)
(5.2.0.14)
(5.2.0.15)
previous chapters. Section 2-2 also shows the effects of the hole on the stress distribution as it is located 10[mm ] from the edge of the hole as illustrated in the picture ( Figure 5.6). Rosettes are placed along Section 2-2, they measure strains in three directions. Section 22 has the most inuential position and by the help of rosettes and close experimental data analysis interpretations can be easly performed. Moreover, this section is different from the other sections because, it is the only section in which measurements are made by the help of rosette strain gauges, thus allowing us to determine the full stress states. Channels 17, 20 and 23 are orientated in the x-x direction in the section 2-2. For the same section, strains on y-y axis are captured by Ch18, Ch21 and Ch24. Previous table (Table 5.1) shows the channels that correspond to each strain gauge plus strain gauge orientation. Conventional axis for all three directions have been chosen showed in the illustrations and those are:
50
x-x is the axis parallel to the symmetry axis it goes along the models length and is perpendicular to the loading orientation; y-y axis is orthogonal to x-x direction and parallel to the loading orientation; x-y or y-x axis is the 45 vector resulting from the composition of the rst two principal directions.
51
52
Numerical Analysis
6.1 Introduction
This chapter covers numerical analysis of the beam. Four different models were created using a code written in MatLAB and commercial nite element program Abaqus CAE. The main purpose of the numerical part is to investigate the behavior of the beam in two different loading cases using Finite Element method. The created models are compared with each other. The results from the most appropriate model are then compared with the values obtained from analytical and experimental parts. The following models are presented in this chapter: a 2D model which is created using an assumption of plane stress and plain strain; a 3D solid model and a 3D shell model. The magnitudes of forces were calculated to avoid yielding and exceeding linear elastic state in the beam (see Chapter 4.5.1). The theory about nite element method can be found in Appendix A.3
the main test (see Chapter 5.1). Forces in both loading cases are applied in the negative ydirection (see Figure 6.1 and Figure 6.2).
Figure 6.1. Full beam model with applied forces in the 1st loading case.
Figure 6.2. Full beam model with applied forces in the 1st loading case.
54
It constrains displacement in y-direction in the applied section and rigid body motion in zdirection (see Figure 6.5). In both loading cases, the beam is equally supported. In contrast, point supports are used in two dimensional models with the same movement restriction. Point and line supports causes stress concentration in the areas where they are applied. To avoid this phenomenon plate supports could be modelled. It would help to distribute stresses in the wider area. In this project it was decided not to use plates like supports because the stress concentration in supports areas does not have big inuence on the nal results obtained from the models. Also, point/line supports were chosen to create similar support conditions like in the laboratory-roller supports (see Chapter 5.1).
Figure 6.5. Full beam model supported by roller at one end and pinned support at another.
55
To obtain the same results from the quarter and the full beam models the following boundary conditions have to be specied. A roller support is applied on the cross section in the middle of the beam to keep the model from rigid body motion on z-axis (see Figure 6.7). It also restricts rotations around x and y-axes. Another roller support is designed on the second symmetry plane which splits the cross section in to two parts (see Figure 6.8). In this case displacement on x-direction and rotations around y and z-axes are constrained. Just a quarter of the total force in both loading cases is taken into consideration because the full model is split in to four parts. For the same reason load distribution area is four times smaller than in the full model.
56
Figure 6.7. The support applied on the cross section in the middle of the beam illustrates 1st symmetry plane.
Figure 6.8. Roller assigned along beam model interprets 2nd symmetry plane
Symmetry conditions specied above are used in both loading cases for all 3D models. It was decided not to use the symmetry conditions in the Abaqus 2D plane stress and plane strain model and in 2D model created in MatLAB because a ne mesh was generated for the full model and accurate results were obtained. Also the main purpose of the creation of 2D model in Abaqus is to compare it with the 2D MatLAB model.
The mesh density can be increased or decreased by changing variables cc, gg, nn in the MatLAB script input. The variable cc describes the mesh density around the holes, while gg and nn describe the mesh density between the holes and at the beam ends respectively (see Figure 6.10 and Figure 6.11).
Figure 6.10. Generated mesh on the part of the beam. In this case cc = 4, gg = 4 and nn = 4.
58
Figure 6.11. Generated mesh on the part of the beam. In this case cc = 8, gg = 8 and nn = 8.
around the holes; in the top ange where the force is applied; in the web section where the strain gauges are installed; in the bottom ange at the one end where the beam is supported; in the bottom ange at half the beams length (see Figure 6.14).
Figure 6.12. Side one of varying mesh density through the model. In this case mesh consists of 86 490 elements and 158 810 nodes.
59
Figure 6.13. Side two of varying mesh density through the model. In this case mesh consists of 86 490 elements and 158 810 nodes.
For von Mises stress, the point located on the top ange, 850 mm from the left-beam end in Abaqus and the middle point in the bottom ange in MatLAB are chosen (see Figure 6.15 ). For deection, the point is established in the middle of the beam on the bottom ange.
Because of the lack of time the different point for von Mises stresses convergence test in MatLAB was chosen to avoid rewriting script for convergence analysis. Von Mises stresses were determined in both MatLAB and Abaqus because they represent real stress state more general (the formula for calculations includes all stress components). The results for other stress components are presented in the Appendix A.4.
60
Figure 6.14. Marked points for convergence test for 3D Abaqus models
Figure 6.15. Marked point for both von Mises stress and displacement convergence tests for 2D Matlab model
61
Table 6.1. Steps for increasing the mesh density in convergence analysis.
The following gures (see Figure 6.16 and Figure 6.17) show the results from performed convergence analysis for LST elements (see Appendix A.3). It can be noticed that the convergence of stress is slower than that of displacement (see Table 6.2).
62
Figure 6.17. Convergence analysis in terms of von Mises stresses for MatLAB 2D model.
Table 6.2. The results from convergence test for von Mises stresses and displacement for 2D MatLAB model.
63
The following gure (see Figure 6.18) and table (see Table 6.4) show the results from performed convergence analysis.
Figure 6.18. Convergence analysis in terms of displacement for Abaqus 3D shell model.
An approximation of using a shell in modelling of a structure is based on the advantage of the dimensions of the shell (thickness is small compared to the other dimensions) Simulia [2010]. Three-dimensional shell elements are 4- to 8-node isoparametric quadrilaterals or 3- to 6-node triangular elements in any 3-D orientation. Each shell element node has 5 degrees of freedom (DOF) - three translations and two rotations. The translational DOF are in the global Cartesian coordinate system. Mirza & Smell [2011] There are two types of shell elements in Abaqus: conventional and continuum shell elements. In conventional shell elements the thickness of material is dened through section properties. On the other hand, continuum shell elements resemble three-dimensional solid elements. Simulia [2010] Shell formulation assumes thin shell problems and thick shell problems. Thick shell problem takes in consideration transverse shear deformation and thin shell problem neglects transverse 64
Table 6.4. The results from convergence test for displacement for 3D Abaqus shell model.
shear deformation. Shell element has both displacement and rotation degrees of freedom. In our case, the beam was modelled like a conventional thin shell model. The displacement is assumed to vary linearly through the thickness. Wemper & Talaslidis [2003] The convergence for Abaqus 3D shell model was performed only in terms of displacement because it is complicated to interpret stresses. Stress components in the shell are interpreted in the local directions, which are dependent on the orientation of each element and not on the orientation of the global axes. Stress components S11 and S22 are stresses in local 1 and 2 directions. Stress component S33 is normal to the surface therefore it is always equal to zero. Local 1- and 2directions lie in plane of the shell. Default local 1-direction is the projection of global 1-axis onto shell surface. If global 1-axis is normal to shell surface, local 1-direction is the projection of the global 3-axis on shell surface. Local 2-direction is perpendicular to local 1-direction on surface of the shell, so that local 1-direction, local 2-direction, and the positive normal to the surface form a right-handed set (see Figure 6.20),Morley [1995]
65
Since the 3D solid Abaqus model consists of more than 87 000 quadratic elements it is not possible to run the model due to the limited virtual memory of the computer. But it is not necessary to continue the test since the convergence is reached (see Figure 6.21, Figure 6.22, Table 6.7 and Table 6.8). Displacement grows slightly with 10 3 [mm ] increment after 87 000 linear elements bound. The same result is obtained using just 9200 quadratic elements. The results (see Figure 6.21 and Figure 6.22) prove that both types of elements can be converged, but ten-node tetrahedrons provide more accurate results. For the above mentioned reasons C3D10 elements are used in the nal mesh. Distorted elements have a deformed geometry, i.e. one point (nodal or mid-side) is displaced far away from the rest of element points. Data 66
regarding this type of elements was collected during the test as well (see Table 6.22). It is very important to avoid distorted elements because when force is applied, strain value in such element and around it becomes unreliable. It can be noticed in Table 6.22 that number of distorted elements decreases with a higher density mesh (there are some inconsistencies because some individual elements get distorted in areas where mesh is rough).
In this case it is not important that number of distorted elements increases at the end of the test (see Table 6.21) because quadratic elements (mesh number 8) are decided to use in the nal 3D solid Abaqus model.
Figure 6.21. Convergence analysis in terms of von Misses stresses for Abaqus 3D solid model.
67
Figure 6.22. Convergence analysis in terms of displacement for Abaqus 3D solid model.
68
Finally, 3D solid model mesh consisting of 87855 ten-node tetrahedron elements is decided to be used for the nal model. Based on the convergence test, mesh with a lower element number could be used to obtain accurate results but since some distorted elements are provided it is better to use ner mesh.
6.5 Results
In this sub-chapter the results obtained from different numerical models for displacement and von Mises stresses for the second loading case are presented. A more detailed discussion is developed in the following chapter (see Chapter 7). All obtained results from numerical part can be found in Appendix ref{ap:A.4 The biggest deection (1,517 mm) was determined in the point located on the bottom ange in the middle of the beam (see Figure 6.23).
As shown in the Figure 6.24 the biggest displacement is in the middle point of the area where the load is applied. The monitor point is in the middle of the beam and the deection obtained from this point is equal to 1,526 mm. The difference between 2D MatLAB and 2D plane stress and plane strain models is very small because the same assumptions were made both for MatLAB model and Abaqus model. The only one difference between these two models is the used element type. The obtained displacements in the 3D models are shown in Figure 6.25 and Figure 3.6.4. More reliable results are determined in the 3D solid model (deection is equal to 1,581 mm at monitor point) than in the 3D shell model (deection is equal to 1,424 mm). That happens because in the 3D solid model the cross section of the beam better resembles the real cross 69
Figure 6.24. Deection of 2D plane stress and plane strain Abaqus model.
section (e.g., curvatures are modelled). Since elements used are different in both models, the resulting deection can also be affected.
70
In the following gure (see Figure 6.27) the distribution of von Mises stresses in 2D MatLAB model is presented. It was decided to show von Mises stresses because it is common used failure criterion which represents the obtained results well. The maximum stresses appear around the hole because of the inuence of stress concentration factor. This phenomenon was also discussed in analytical calculations (see Chapter 4.5.1)
The von Mises stresses of 2D plane stress and plane strain model in Abaqus (see Figure 6.28), (maximum value is equal to 149,1MPa) are very close to results from 2D model in MatLAB (maximal value is at the edge of the fourth hole at 135 angle and the magnitude is 145, 9M P a ). The small difference in results might appear because of different types of elements.
71
The maximum value of von Misses stress of Abaqus 3D solid model (269,4MPa) appears in the area where line support is applied (see Figure 6.29 upper left corner). Distribution of von Mises stresses in pressure load area varies approximately from 67MPa to 110MPa. It is visualized in Figure 6.29 lower right corner. All models can be found in Digital Appendix B.8, B.9, B.10 and B.11.
72
Comparison-Conclusion Chapter
This chapter aims to compare and discuss results gained from the analytical, numerical and experimental analysis. The following part will be divided into three main sub chapters: discussion of displacement, stresses and conclusion. All models uses parameters which were obtained via experimental workshop see Chapter 3.4.
7.1 Displacement
Bar-chart and table below show the obtained displacement for seven types of performed analysis. Under the rst loading case, displacement is bigger in accordance to second loading case. Hence, displacement is directly affected by the increasing of moment and applied force.
In the analytical model, the displacement calculated with Timoshenko beam theory is bigger than the displacement calculated with Euler - Bernoulli because the rst theory considers the 73
shear deformation. The difference between analytical and numerical results was expected because of assumptions made in beam theories regarding cross section characteristics. The 2D Abaqus and 2D MatLAB models provide almost exact results because as mentioned before models were created using plane stress and plane strain (see Table 7.2). Difference can be explained by different element types used. Abaqus 3D solid model compared to 3D shell model gives more realistic results because 3D solid model represents the beam cross section better.
Figure 7.1. Displacement of the bottom ange along the beam from different models for the rst loading case
74
Figure 7.2. Displacement of the bottom ange along the beam from different models for the second loading case.
Experimental results give the biggest displacement due to some error sources. Firstly, the real beam cross sections dimensions vary along the beam which can cause a difference in the measured displacement. In order to check, the dimensions of the real cross section were measured in three different positions, e.g. anges thickness varies up to 0.9[mm ], the diameter of the holes is not constant throughout the entire beam. On the other hand materials proprieties that were obtained during material test can include some uncertainties. Inaccurate strain gauges installation, processing the material properties data using linear regression, the test specimen which might not represent the material of the beam precisely can all be a cause of the difference between calculations and test results. Also, the displacement transducer can cause an error if it is not properly zero-calibrated and/or vertically placed. Furthermore, modelling of loads and boundary conditions can cause some uncertainties because of made assumptions
7.2 Stresses
Besides displacement, stresses were also a point of interest in this project. Three cross-sections (named: 1-1, 2-2, 3-3) were considered for further stress-distribution analysis. Results for stresses from analytical and numerical analysis were obtained in the same positions where the strain gauges were placed during the experimental set up. Stress distribution obtained from all the parts is presented in the following graphs and the values for the interesting points are in the following tables .
75
Normal s tr ess dis tribution Sec tion 1-1 Normal s tr ess dis tribution Sec tion 2-2 She ar s tr ess dis tribution Sec tion 2-2
66,50 66,50 h [mm] h [mm] 63,50 53,50 49,88 49,88 43,50 33,50 43,50 33,50 33,25 33,25 23,50 13,50 [M Pa] 3,50 [M Pa] 0,00 0,00 5,00 [M Pa &! -80,00 -60,00 -40,00 10,00 0 30 60 -60 1,50 2,00 16,63 16,63 63,50 '!"((&! 53,50
She ar s tr ess dis tribution Sec tion 1-1 Normal s tr ess dis tribution Sec tion 3-3
63,50 h [mm]
!
!
80,00
! h [mm&!
53,50
60,00
43,50
33,50
40,00
23,50
23,50
13,50 [M Pa&! 0,00 0,00 -0,20 20,00 40,00 60,00 3,50 0,00 0,20 0,40 0,60 0,80
13,50
20,00
3,50