2.

Nonlinear Material Analysis

2.1. Introduction
In elastic solid mechanics, linearity is implied in:
a) a linear form of strain–displacement relationships, i.e., ε = [Link], where ε is strain, ue is
element displacement components and B is the strain-displacement relationship.
b) a linear form of stress–strain relationships, i.e., τ = E.ε, where E is an elasticity matrix
containing the appropriate material properties, and τ and ε are stress and strain tensor
respectively.
Many problems of practical consequence exist in which such linearity is not preserved as
plasticity, creep, or other complex constitutive relations. These classes of problems can be simply dealt
with without reformulation of the complete problem. If a solution to the linear problem can be reached
by some iterative process in which, at the final stage, the material constants are so adjusted that the
appropriate new constitutive law is satisfied, a solution is achieved. However, if the strain-
displacement relationship is nonlinear, then reformulation is necessary.
It's worth noting that, while in linear problems the solution is always unique, this no longer is
the case in many nonlinear situations. Thus, if a solution is achieved it may not necessarily be the
solution sought. Physical insight into the nature of the problem and, on occasion, small-step,
incremental approaches are essential to obtain the significant answers. The iterative approaches used
can, on occasion, be interpreted as purely 'numerical analysis' processes such as Newton-Raphson
methods, etc.

2.2 General Physical Approach

In small strain linear elastic problems formulated by the
displacement approach, a final answer is achieved by solving the
assembled (stiffness) equations:

KU – R = 0 …………… (1)

in which the vector R lists all the "forces" due to external loads, initial
stresses and strains, etc. In eq. (1) validity of a linear elastic constitutive
law is assumed, i.e.

τ = E (ε -𝜀0 ) + 𝝉0 ……. (2)

in addition to a linear strain-displacement relation, continuity of displacement,

and approximate satisfaction of equilibrium are also assumed
In small displacements analysis of problems in which a nonlinear
constitutive relationship applies, displacement continuity and equilibrium
still have to be satisfied. The stress-strain relationship (Eq.2) has to be
replaced by:

F (τ, ε) = 0 ……………… (3)

If a solution of Eq.(1) can be achieved in which, by adjustment of
one or more of the parameters E, εo, or τo in Eq.(2), this and Eq.(3) are
made to yield the same stress and strain values, then a solution is found.
An iterative approach is obviously essential, which of the three quantities:
E, εo, or τo, is to be adjusted in the iteration process depends on:
a) The solution method used in the equivalent linear elastic problem,

Nonlinear analysis 13
 b) The nature of the physical law defining the stress-strain relation.
If the iteration is conducted by adjustment of the E matrix the process will be known as one of variable
stiffness. If 𝜀0 or 𝝉𝑜 are adjusted, so-called initial strain or initial stress approaches will be obtained.
In many real situations, relations of the type of Eq.3 cannot be written in terms of total strains
and stresses, but nevertheless can be established for increments ∆τ and ∆ε. In such cases the same
processes can be applied for an increment of load (or time in creep situation). Such incremental
processes can be combined with any of the previous methods. The specification of E, 𝜀0 and 𝝉0 forms
the essential data required for any linear elastic analysis program. Thus such programs will form the
essential core of any non-linear analysis solution.

2.2.1 Variable stiffness method

If the stress-strain relation Eq.3 for the behavior of a particular material can be written in the
form of Eq.2 but with the elasticity matrix being now a function of the strain level reached, i.e.

E = E (ε) = E (U) …………. (4)

Then, the process of variable stiffness can be applied.

As the elasticity matrix influences the final stiffness matrix of the assembly, the solution of a
problem:
r = R – [𝑲(𝑈)]𝑼 = 0 …………… (5)

is required, which can be accomplished iteratively in different ways.

A simple iterative process, in which 𝑼0 = 0 is an initial estimate in order to evaluate 𝑲(𝑈)𝑜 =𝑲0 ,
and solve 𝑼1 =𝑲−1 𝑜 R repeating the process with

𝑼𝑖 =𝑲−1
𝑖−1 R …………… (6)

Where i is the number of iteration and the process is repeated until the displacement changes can be
neglected or less than certain tolerance.
If the constitutive law is such that a relation of type of Eq.(4) can only be written for increments
of stress and strain, then the process has to be adopted for small load increments starting from
previously arrived conditions. In either case the standard linear elasticity program can be used,
providing the form of E arrived at is symmetric. One serious disadvantage of the variable stiffness
approaches is the fact that at each step the stiffness matrices have to be reformulated and a new solution
of equations obtained.

2.3. Solution of Nonlinear Equilibrium Equation

In an incremental procedure or in step–by-step analysis, the relations in Eqs.(5 and 6) can be
employed to calculate an increment in the displacements, which is then used to evaluate approximation
to the displacement, strain, and stresses corresponding to a load step n. Assuming that the approximate
displacements, strains, and stresses have been obtained, the equilibrium condition can be expressed as
the difference between the external applied loads and the nodal point forces that correspond to the
elements stresses in this configuration , i.e., .

r  Rn  [ K (U )]n1U n (7)

Or,
r  Rn  [ K (U )]n1 U n  F n1  (8)

Where n is the load increment and F is the contribution of internal forces to equilibrium.

Nonlinear analysis 14
 Therefore, the solution of nonlinear problem reduces to that of tracing a nonlinear load
displacement path by solving a system of nonlinear algebraic equations. The solution procedures for
the equilibrium equations are numerous, and the most popular procedures are given briefly here.

The first class of solution procedures is that which satisfies or at least attempts to satisfy the
equations of equilibrium exactly .i.e.:

{r} = 0 (9)

The first method which satisfies this condition is the method of successive approximations, in
which the initial stiffness matrix [K] 0 is employed and the iterative technique is characterized by the
recursion relation:

[𝐾]0 {𝑑}𝑖 = {𝑅}𝑛 − {𝐹}𝑖−1 (10)

Where,
{𝐹}𝑖−1 = {𝐹}𝑛−1 + [𝐾]0 {𝛥𝑈}𝑖−1 (11)

, 𝜟{𝑼}𝒊 = 𝜟{𝑼}𝒊−𝟏 + {𝒅}𝒊 (𝟏𝟐)

And,
{𝑟}𝑖 = {𝑅}𝑛 − {𝐹}𝑖 (13)

Where the subscript, i, is the iteration number. The iterative

technique of this method is one of the easiest to apply, but the
method generally exhibits a very slow rate of convergence
for problems of significant nonlinearity. Also, the possibility
of obtaining a divergent solution may occur.

The second method which satisfies the condition, {r} = {0}, is the Newton–Raphson
method. For approximate displacement vector, {u}, a Taylor's series expansion of the force
residuals about the point {u} yields:
𝜕𝑟𝑖 (𝑈)
𝑟𝑖 (𝑈 + 𝛥𝑈) = 𝑟𝑖 (𝑈) + . 𝛥𝑈𝑖 + ⋯ (14)
𝜕𝑈𝑖

The conventional Newton–Raphson procedure retains only the first two terms in the
Taylor's expansion. In addition, it is assumed that the unbalance in the nodal forces corresponding to
(U+ ΔU) is zero. These assumptions allow equation (14) to be written as:
−𝜕𝑟𝑖 (𝑈)
𝑟𝑖 (𝑈) = . 𝛥𝑈𝑖 (15)
𝜕𝑈𝑖
Or,
[𝐾]𝑛−1
𝑖−1 {𝑑}𝑖 = {𝑟(𝑈)}𝑖−1 (16)
And,
𝜟{𝑼}𝒊 = 𝜟{𝑼}𝒊−𝟏 + {𝒅}𝒊 (𝟏𝟐)

Beginning with an initial estimate of the

displacement vector, Δ{U}i, equations (16) and (12) are
successively applied to yield better and better
approximations. This iteration is continued until either the
value of the displacement increment or the force unbalance
is sufficiently small. In nonlinear analyses, the Newton–
Raphson method has proved itself as one of the best methods of solution available. This is because the
method possesses some excellent characteristics such as the ability to converge for highly nonlinear

Nonlinear analysis 15
behavior, the accurate and the rapid convergence for a realistic initial estimate of the displacement
vector, and finally, the possibility to control the error and estimate the rate of convergence since the
iteration can be continued until a specified degree of accuracy is obtained.

The requirement that the user must make an initial

estimate of the displacement vector is considered a drawback
of this technique. As for the method of successive
approximations, the accuracy of this initial guess determines,
to a large extent, the number of cycles required for
convergence. The most significant drawback associated
with the Newton–Raphson procedure is the large amount of
computational effort required to compute and invert at each
cycle the coefficient matrix in equation (16). In order to
reduce the amount of computational effort, a modified
Newton–Raphson procedure can be employed wherein the
coefficient matrix is held constant for several iterations or
possibly for several load increments. The possibility of slow
convergence or even divergent solution may occur with the
modified Newton–Raphson method when the coefficient
matrix is not updated as often as necessary.

The second class of solution procedures is that which

.
satisfies the condition {r (U )}  {0} , where the dot means
differentiation with respect to some load parameter thus,

[𝐾]𝑛−1 𝛥{𝑈} = 𝛥{𝑅} (17)

Where,
𝛥{𝑅} = {𝑅}𝑛+1 − {𝑅}𝑛 (18)

Where, in equations (17) and (18), the coefficient matrix is calculated at the beginning of each
increment. In purely incremental method, (Eqs.17, 18), the equilibrium equation is generally not
satisfied because the procedures do not involve equilibrium iterations, and the solution tends to diverge
from the true deformation path. Therefore, the Newton–Raphson method is the best technique to apply
and the modified Newton–Raphson method has the advantage that the stiffness matrix needs to be
formed and factorized once per increment or per several
increments, but the modified method suffers the disadvantage
of slower rate of convergence.
As an alternative to the Newton–Raphson method, a class
of method known as matrix update method or Quasi–Newton
method has been developed. These methods provide a secant
approximation to the matrix from iteration i-1 to i. That is if:

{𝑑}𝑖 = {𝑈}𝑖 − {𝑈}𝑖−1 (19)

and,
{ 𝑟 }𝑖 = {𝑟(𝑈)}𝑖−1 − {𝑟(𝑈)}𝑖 (20)

Then, the updated matrix

_
should satisfy the Quasi–Newton equation:
[𝐾]𝑖 {𝑑}𝑖 = {𝑟}𝑖 (21)

Nonlinear analysis 16
Amongst the Quasi–Newton methods available, the BFGS (Broyden–Fletcher–Goldfarb–Shanno)
method, in which the updated inverse [K]-1 is given by:
[ K ]1 i  [ I ]  { y}i .{Z}Ti  [ K ]1 i1 ([I ]  {Z}i .{ y}Ti )
_ _ _ _
(22)
 
Where,
{𝑑}𝑖
{ 𝑦 }𝑖 = (23)
{𝑑}𝑇
𝑖 { 𝑟 }𝑖
And,
1

 _
2
 {d }i {r}i 
__ T __
{ z }i   T  .[ K ] i 1 {d }i  { r }i (24)
{d }i [ K ]i 1{d }i 
In which the superscript, T, means transpose. It should be noted
that the updating process was not performed if the value of
square root in equation (24) is greater than 105.

2.3.1. Convergence criteria

In nonlinear analysis of a structure it is necessary to set convergence criteria which if achieved
during iteration of an applied increment of load or displacement will indicate a state of equilibrium
on the equilibrium path and the iteration process is terminated. The convergence criteria used for
nonlinear structural problems can be classified as follows:

(1) Force criteria (2) Stress criteria (3) Displacement criteria

These three types of convergence criteria will be discussed next.

1. Force Criteria
The force criteria are normally based on a comparison between the computed internal forces
Pint within the structure and the external applied loads P. This difference being the out of balance
force vector Pob. However, making use of such a comparison does not always make sense because
the force quantities to be compared may be of a completely different dimensions. So, to make
moments and forces of a comparable order so that a single preset convergence limit can be used for
both of them, the values of the moments at the ends of each member have been summed and divided
by the member length to convert the moment into an equivalent (shear) force. Another problem
associated with the use of force as a convergence criteria is that it is not easy to decide an appropriate
convergence limit for Pob because this depends on the type , the stiffness , and the geometry of
structure being considered. A stiff structure may be unable to converge within a small Pob limit,
while softer structures with a big Pob limit may give solutions that do not lie on the equilibrium path.

2. Stress criteria
This type of convergence criteria involves a check on the changes in stress values during an
iteration cycle which can be compared with prescribed stress levels. This type of criterion is suitable
for structure where only one type of stress dominates the behavior such as trusses or cable and
membrane structures.
3. Displacement criteria
As has been noted above, the preset limiting value for Pob in force convergence criteria should
depend on the stiffness properties and the nature of the structure. However, it is often easier to place

Nonlinear analysis 17
the convergence criteria directly on increments in displacements. Consider a nondimensional vector
based on displacements and defined as
∆𝑢1 ∆𝑢2 ∆𝑢𝑘 ∆𝑢𝑛
{𝜀𝑗 } = {∆𝑢 … … … … … }
1.𝑟𝑒𝑓 ∆𝑢2.𝑟𝑒𝑓 ∆𝑢𝑘.𝑟𝑒𝑓 ∆𝑢𝑛.𝑟𝑒𝑓

Where, n is the total number of degrees of freedom and Δuk is the change in the displacement
component k during iteration cycle j. Every such component is scaled by a reference displacement
quantity uk,ref. These reference quantities are, in general, not equal to the corresponding total
components because if uk is close to zero, the ratio Δuk / uk could be a large number even after
convergence has occurred. Every Δuk is therefore scaled by the largest displacement component of
the corresponding type. For example, the changes in deflections can be scaled by the largest absolute
deflection component, and the changes in the rotations can be scaled by the largest absolute rotation
component.

Three alternative norms can be used for measuring the size of the vector {ε} for the
displacement criteria:
n
Modified absolute norm = ║ε║1 =1/n ∑ │ Δuk / uk,ref │
k=1

n
Modified Euclidean norm = ║ε║2 = [1/n ∑ │Δuk / uk,ref │2 ]1/2
k=1
Maximum norm = ║ε║max =max k │ Δuk / uk ref │

The absolute and the Euclidean norms have been modified by the division by n to obtain quantities
that are independent of the total number of degrees of freedom. For any of these norms, the
following convergence criteria can be used

║ε║ < tol

where the value of tol is the required tolerance , usually of the order of 10-2 to 10-6 , depending on
the desired accuracy.

2.4. Example
A bar rigidly supported at both ends is subjected to an axial load as
shown in Figure (1-a). The stress-strain relation and the load versus time
curve relation are given in Fig. (1-b&c). Assuming that the displacements
and strains are small, and that the load is applied slowly, calculate the
displacement at the point of load application using the Successive Iterations
method, the Newton-Raphson (N.R.) method and its modification (M.N.R.),
and B.F.G.S. method. Compare the results with each other and with the exact
solution.
 Exact Solution
- Elastic behavior
R= Fa + Fb
𝑢 𝑢 3
=𝐸𝐴(10 + 5)= 𝑢𝐸𝐴(10)
10
Then, 𝑢 = 𝑅 ∗ 3𝐸𝐴
Substitute for u in Fa and Fb gives:

Nonlinear analysis 18
 1 2
𝐹𝑎 = 𝑅 ∗ 3 𝐹𝑏 = 𝑅 ∗ 3
i.e. Part (b) reaches yield before Part (a), and the corresponding load
is:
3
u1=εy*Lb=0.002*5=0.01 cm, and 𝑅 = 𝑢 ∗ 𝐸𝐴 ∗ 10
= 3 ∗ 104 𝑁
The equilibrium equation after yielding of Part (b) becomes:
1 1
1 ∗ 104 = 𝐴(𝐸 10 + 𝐸𝑇 5)∆𝑢= ∆𝑢𝐴(1.02 ∗ 106 )
Which gives ∆𝑢=0.98 *10-2 and u=1.98 *10-2 cm

Using FEM

The iterative equations, in this analysis, are:

K an1  K bn1  d i  R n  (Fan1 )i1  (Fbn1 )i1
uin1  uin11  d i
2.4.1. The Successive Iteration Method
In the first step, n = 1, i=1 and the above iteration technique gives :
 
K a0  K b0 d1  R1  Fa 0  Fb
0

10 (  ) d1  2  10 4  0
7 1
10
1
5

2  10 4
d1   6.667  10 3 cm
0.3  10 7

u10  0  6.667  10 3  6.667  10 3

u10
( a1 )1   6.667  10  4   y part " a" is elastic
La
u10
( b1 )1   1.333  10 3   y part " b" is elastic
Lb
( Fa1 )1  EA ( a1 )1  10 7  6.667  10 7  6.667  10 3 N
( F )  EA ( )
1
b 1
1
b 1  10  1.333  10
7 3
 13.333  10 3
N
The out of balance loads  R1 - ( Fa1 )1  ( Fb1 )1  2 10 4  0.667 10 4  1.333 1  0
i.e. Convergence is achieved in one iteration, and: u 1  6.667  10 3 cm  u02
n = 2, i= 1
K a  K b  d1  R 2  ( Fa1 )1  ( Fb1 )
0 0

2  10 4
10 7 101  15 d1  4  10 4  2  10 4 , d1   6.667  10 3. cm
0.3  10 7

u1  u 0  d1  6.667  10  6.667  10 3  1.333  10 2

2 2 3
cm
1.333  10  2
( 2a )1   1.333  10 3  y  elastic
10
1.333  10  2
( b2 )1   2.66  10 3  y  plastic
5
( Fa )1  10 1.333 10 3  1.333 10 4 N
2 7

 
( Fb2 )1  A y  ET A ( b1 )1   y )  2 10 4  10 5 (2.667  2.) 10 3  2.0067 10 4 N
The out – of – balance
Nonlinear analysis 19
 r12  R 2 ( Fa2 )1  ( Fb2 )1  4  10 4  3.34  10 4  0.67  10 4 N
then, i=i+1=2
K 0
a 
 K b0 d 2  r12
0.3 10 7  d 2  0.67 10 4 , d 2  2.2 10 3 cm
u 22  u12  d 2  12.333 10 3  2.2 10 3  15.53 10 3 cm
3
15.53 10
( 2a ) 2   1.553 10 3   y  Elastic
10
15.53 10 3
( b2 ) 2   3.106 10 3   y  plastic
5
( Fa2 ) 2  10 7  1.553  10 3  1.553 10 4 N
 
( Fb2 ) 2  A y  ET A ( b2 ) 2   y )  2 10 4  10 5 (3.106  2.)  10 3  2.011110 4 N
The out-of balance: r22  R 2 ( Fa2 ) 2  ( Fb2 ) 2  4  10 4  3.564  10 4  0.436  10 4 N
then, i=i+1=3
K 0
a 
 K b0 d 3  r22
0.3 10 7  d 3  0.436 10 4 , d 3  1.452 10 3 cm

u32  u 22  d 3  15.53  10 3  1.452  10 3  16.985  10 3 cm

The procedure is repeated and the results of Successive Iterations are given next:
Iteration i di (cm) u i2 (cm)
3 1.452 * 10-3 1.6985 * 10-2
4 0.9583 * 10-3 1.7944 * 10-2
5 0.633 * 10-3 1.858 * 10-2
6 0.617 * 10-3 1.899 * 10-2
7 0.276 * 10-3 1.927 * 10-2
i.e., after 7 iterations, the displacements
u2 = u 72 = 1.927 * 10 -2 cm,
and the out of – balance
r72  4  10 4  10 71.927  10 3  2  10 4
 10 5 (3.854  2)  10 3
 0.055  10 4 N  0.0136 2R
 1.36 % 2R
and convergence tolerance in displacements is :
0.276 103
  0.0143  1.43 %
1.927 102
which can be ignored. –Load –displacement response.

Nonlinear analysis 20
 The convergence of M.N.R. method is slow and this is due to two reasons:
1- Large load increment. (Two load increments only)
2- Use of initial stiffness of K a0 and K b0 .
Therefore, the solution can be enhanced if a small increment of loads is used and the
stiffness matrix is updated frequently as in Newton–Raphson method or its modified. For example,
if 4 load increments are used instead of 2, the iterative equations become:
n= 1, i=1
( K a0  K b0 ) d1  R1  Fa0  Fbo
0.3 10 7  d1  110 4 , d1  3.333 10 3 cm

u11  0  3.333  10 3  3..333  10 3 cm

(ε1a )1  3.333 10 4  εy  Elastic
(ε1b )1  6.667 10 4  εy  Elastic

(Fa1 )1  10 7  3.333  10 4  3.333  10 3 N

(Fb1 )1  6.667  10 3 N
r11  1  10  4  1  10 4  o
i.e. Convergence is achieved in one iteration, and: u1 = 3.333 × 10-3 cm = u 02
n = 2, i=1
0.3 10 7  d1  2 10 4  110 4 , d1  3.333 10 3 cm
u12  (3.333  3.333) 10 3  6.667 10 3 cm

(ε a2 )1 , (ε 2b )1  εy
(Fa2 )1  6.667  10 3 N
(Fb1 )1  1.333  10  4 N
r12  2  10 4  (.667  1.333)  10 4  o

i.e. Convergence is achieved in one iteration, and: u2 =6.667 × 10-3 cm = u 03

n = 3, i =1
0.3 10 7  d1  3 10 4  2 10 4 , d1  3.333 10 3 cm
u13  (6.667  3.333) 10 3  1.0 10 2 cm

10 2
(ε 3a )1  10 3  ε y  Elastic
10
10  2
(ε b )1 
3
 2  10 3  ε y  Elastic
5
(Fa )1  1.0  10 4
3
N
(Fb3 )1  2.0  10 4 N
r13  3  10 4  (1.0  2.0)  10 4  o

Nonlinear analysis 21
 i.e. Convergence is achieved in one iteration, and u3 =1.0 × 10-2 cm
n = 4, i=1
Part "a" remains elastic while part "b" enters the plastic zone, the updated stiffnesses are used:
(𝐾𝑎3 + 𝐾𝑏3 )1 𝑑1 = 𝑅 4 − 𝐹𝑎3 − 𝐹𝑏3
1 1
(107 × 10 + 105 × 5 ) 𝑑1 = 4 × 104 − 3 × 104
1.02 × 106 𝑑1 = 1 × 104 , 𝑑1 = 9.804 × 10−3
𝑢14 = 1 × 10−2 + 0.9804 × 10−2 =1.9804x 10-2 cm
(ε a4 )1  1.9804  10 3  ε y  Elastic
(ε 4b )1  3.9608  10 3  εy  Plastic

(Fa41 )1  1.9804  10 4 N
(Fb4 )1  2  10 4  10 5  (3.9608  2)  10 3  2.0196 N
r14  4  10 4  4  10 4  o
i.e. Convergence is achieved in one iteration, and: u4 =1.9804 × 10-2 cm
2.4.2. Using BFGS method
BFGS is employed using two increments as in Successive Iteration method above, the solution
proceeds as follows:
For n =1 is same as in above, u1 =0.667× 10-2 cm, r1=0
For n = 2, i=1
K 0 d1  R 2  Fa1  Fb1
0.3  10 7  d1  4  10 4  2  10 4 ,
2  10 4
d1   6.667  10 3 cm
0.3  10 7
u12  u02  d1  1.333  10 2 cm

(ε a2 )1  1.333  10 3  ε y  Elastic
(ε 2b )1  2.667  10 3  ε y  plastic
7 3
(F )  10 1.333 10  1.333 10 4
2
a 1 N
4 5 3
(F )  2 10  0 (2.667  2) 10
2
b 1

 2.0067 10 4 N
r  R  F  F  4 10 4  3.34 10 4
1
2 2 2 1
a
2 1
b

 0.669 10 4 N -Load-displ. response

then, i = i + 1 = 2
2
r 1  2  10 4  0.669  10 4  1.333  10 4 N , d1  6.667  10 3 cm
2
K1 d1  r , 1 K1  0.2  10 7 , K11  5  10 7

Nonlinear analysis 22
 And using Eqs.22→24
6.667  10 3
y1  3 4
 0.75  10 4
6.667  10  1.333  10
__  6.667  10 3  1.333  10 4 

z1   3 3
  0.3  10 7  6.667  10 3  1.333  10 4
 6.667  10  0.3  10  6.667  10
7

 0.3  10 4

K   1  0.75 10
1
1
4
 
 0.3  10 4  3.333  10 7  1  0.75  10 4  0.3  10 4 
 5  10 7 as above

d 2  5  10 7  0.67  10 4  3.351  10 3 cm
u 22  (13.333  3.351)  10 3  16.684  10 3 cm

16.684  10 3
(ε a2 ) 2   1.6684  10 3  ε y  Elastic
10
16.684  10 3
(ε 2b ) 2   3.3368  10 3  ε y  plastic
5
(Fa2 ) 2  10 7  1.6684  10 3  1.6684  10 4 N
(Fb2 ) 2  2  10 4  10 5  (3.3368  2)  10 3  2.0134  10 4 N
r22  4  10 4  3.3818  10 3  0.3182  10 4 N
i = i + 1=3
2
r 2  0.67  10 4  0.3182  10 4  0.3518  10 4 N
3
d 2  3.351  10 cm
0.3518  10 4
K2  3
 0.105  10 7 ( K 1 ) 2  9.525  10 7
3.351  10

Using Eqs.22→24
__
3.351  10 3
y2  3 4
 2.8425  10 4
3.351  10  0.3518  10
1
__  3.351  10 3 2
z 2   3
  0.2  10 7  3.351  10 3  0.3518  10 4
3 
 3.351  10  0.2  10 7
 3.351  10 
 0.1338  10 4

K  1
2  (1  2.8425  10 4  0.1338  10 4 )  (5  10 7 ) (1  2.8425  0.1338)
 9.5265  10 7 as above
d 3  9.5265  10 7  0.3182  10 4  3.031  10 3 cm
u 32  (16.684  3.031)  10 3  19.715  10 3 cm

Nonlinear analysis 23
 19.715  10 3
(ε ) 2
a 3  1.9715  10 3  ε y  Elastic
10
19.715  10 3
(ε 3b ) 3   3.943  10 3  ε y  PElastic
5
( Fa2 ) 3  10 7  1.9715  10 3  1.9715  10 4 N
3
( F )  2  10  10 (3.943  2)  10
2
b 3
4 5
 2.01943  10 4
N
r  4  10  3.991  10  0.009  10
3
2 4 4 4
N
i.e. after 3 iteration of BFGS method the residual is 0.225 % of R2 , compared with
M.N.R. of 1.36 % after 7 iterations.

Problems to be solved:
1- For the assemblage shown 1n figure E-1-a and
stress –strain curve shown in b, find the response
of the assemblage under the load R shown in c,
using different iteration techniques.

Figure E-1

Nonlinear analysis 24
For the assemblage shown 1n figure E-2-a, Find the
collapse load if:
a- Members cd and Ef are made of steel with
stress –strain curve shown in b.
b- Members cd is made of steel with stress –
strain curve shown in b, and member ef is
made of concrete with stress- strain curve
shown in c.

For Steel:
In tension: τ= E.ε ε ≤ εy
τ= τy +ET (ε - εy) ε ˃ εy a) Geometry
In Compression: τ= E.ε τ ≤ τpb
τ= τpb τ ˃ τpb

For Concrete:
2𝜀 𝜀 2
τ = fco [ −( ) ]
𝜀𝑜 𝑜 𝜀
𝑑𝜏 𝟐𝐟𝑐𝑜 𝜀
E= = [1 − ]
𝑑𝜀 𝜺𝒐 𝜀𝑜

b) Stress-strain curve for steel

c) Stress-strain curve for concrete

Figure E-2

Nonlinear analysis 25