Professional Documents
Culture Documents
Structural Analysis Techniques - Assignment 3
Structural Analysis Techniques - Assignment 3
Structural Analysis Techniques - Assignment 3
Unit: SA2
TABLE OF CONTENTS
1. PROBLEM 01 2
1.1. Introduction 3
1.2. Explicit method (or “Euler forward”) 8
1.3. Explicit method (or “Euler forward”) - Using Internal Force 9
1.4. Full Newton-Raphson method 10
1.5. Modified Newton-Raphson method 11
1.6. Comparison 12
2. PROBLEM 02 14
2.1. Von Mises 14
2.2. Tresca 16
3. PROBLEM 03 19
3.1. Introduction 19
3.2. Direct Method (“Euler forward”) 21
3.3. Direct Method (“Euler forward”) - Plastic strain divided 23
3.4. Implicit Method (“Euler backward”) 27
3.5. Comparison 29
1. PROBLEM 01
1.1. Introduction
The given problem is a column under a compressive load. The first assumption to
solve the problem is that the displacement in the concrete is the same as the steel, so
the strain is the same for both. In this simples case, the strain is direct the ratio
between the displacement and the initial length of the column.
Before to start the calculation is important to state that all units are in Meters and
Newton.
The constitutive relation of the materials (steel and concrete) can be represented in
the functions below:
The internal forces of the structure, for this case, is the sum of the stress multiply by
the area for each element (material) in the axial direction:
The tangent function of the constitutive relation of the materials (steel and concrete)
can be represented in the functions below:
The stiffness of the structure, for this case, is the sum of each element (material) in
the axial direction:
1.6. Comparison
For all of the cases, the calculations were very fast, finding no difference between
them. It is because it is a simple problem and the number of steps was low.
For all the explicit method, the values were always higher than the real value. For the
implicit method, we can see that the values are always the “exactly” one, but we still
need more steps to have a smooth graph.
For the explicit method, we can see in the graph below that the fact to use the internal
force, we get more accurate values. With enough number of steps, the graph can
reach the real one.
For the implicit method, we can see that both of them found the “exactly” value.
However, the modified Newton-Raphson needed to do many steps to converge. For
simples problems, like this one, Full Newton-Raphson is shown to be a good tool, but
the same cannot be said for larger problems, which to calculate the stiffness matrix
consume much time.
2. PROBLEM 02
To find the minimum (or maximum) values of , we need to reach the board of the
yielding function, it means that we need to find the root of the equation.
For our case that σ 2 = 500 , we can find the value of σ 1 in the graph
2.2. Tresca
To find the minimum (or maximum) values of σ 1 , we need to reach the board of the
plasticity function, it means that we need to find the root of the equation.
The Von Mises and Tresca yielding function was plotted in the graph below. The
coincidence of the values found for σ 1 in both functions can be noticed in the graph.
3. PROBLEM 03
3.1. Introduction
The given problem is a steel plate subjected to two strain increments. Initially, the
plate has no stress. We are assuming that its behavior can be simplified as plane
stress. All the units are in Newton and millimeter.
● The yield stress of the steel is:
To simplify the plane stress problem, we are ignoring the shear stress from the
equation, so the matrix has been “correct” to represent this simplification.
● The derivative of the Von Mises equation for a plane stress problem:
● This equation has been created with a purpose to calculate the alfa, through
the root of the equation.
● Initial state
● First increment
● Second increment
● Results
The stress state for each increment:
The one-step routine for the return mapping was not enough to come back inside
(below zero) the yield function.
The results that we need for the next question are: the tension applied to achieve the
limit of the elastic part and the plastic strain
Now we are going to do a different procedure, instead of to use the total plastic strain
in just one step, let us divided into many steps. For the first case, we divided it into
two steps and for the second case, we divided it into many steps as necessary to
have an acceptable error.
● First case
The initial state is the previously elastic stress.
● Results
The stress state for each step:
The two-step routine for the return mapping was not enough to come back inside
(below zero) the yield function, the error still not acceptable.
● Second case
For the second case, we want to find an error of less than 10-4 x the initial yield
function value (one division). So we used the value from the question 3.1.
The number of divisions was done incrementally until to find an error of less than 10-4 x
the initial yield function value or a number considered enough to find a good solution.
It was necessary to divide the plastic strain in 2297 to find an acceptable error.
Counting with all the calculations that had to be done to find the convergence, the total
number of calculations were:
It is a high number of calculations and the computation time consumption was more
than expected. To fix this problem, the number of divisions was done by multiplying the
previous one for two.
Doing that, the number of calculations decreased considerably, even the number of
divisions increased. Because of this, we found the solution in a faster and more
accurate way.
We are assuming the initial condition is one with the elastic part and then we are
increasing the strain.
In this method, to find the new stress state, we need the previous one and the elastic
prediction-plastic correction of the next step.
To find the elastic prediction-plastic correction of the next step, we can use the
approximation.
Where
And
3.5. Comparison
For the Euler forward method, using a one-step-back, we observed that it was not
enough to return to the yield surface. So, it was necessary to do sub-divisions of the
load step to reduce the error, but the number of sub-divisions to find an acceptable
error led to a very large number.
The implicit Euler backward method showed to be a good solver for the incremental
relations between stresses and strains. Only with six steps, we could achieve
convergence with a negligent error.