Vous êtes sur la page 1sur 17

Plastic Analysis and Design by Linear Preprint

Programming using MATLAB® and Octave

Aamer Haque

Abstract
Plastic analysis and design can be performed using linear programming. The procedure to formulate the linear programming
problem is explained. MATLAB® and Octave are then used to solve the linear programming problem.

Preprint submitted to Elsevier April 28, 2015


1. Introduction For any frame, the number of independent mechanisms can
easily be determined. All possible mechanisms are either one
1.1. Plastic Methods
of the independent mechanisms or a combination of them.
Applied loads Wj = αj W j = 1, . . . , Nloads The procedure to determine the independent mechanisms is
as follows:
Member lengths L k = γk L k = 1, . . . , Nmembers
• Let r be the degree of redundancy of the structure.
Moment capacities Mpk = βk M p k = 1, . . . , Nmembers
• Let N be the number of plastic hinges that can form.
Moment distributions mi i = 1, . . . , N These can only form at the locations of the applied loads
and at joints.
Table 1: Notation for plastic analysis
• The number of independent mechanisms is computed as:
Plastic methods of analysis and design have been used for Nmech = N − r.
several decades. These methods are described in detailed in • Classify each of the Nmech independent mechanisms as:
several classic texts: Baker and Heyman [1], Beedle [2], Hey- beam, sway, or joint.
man [4], Hodge [5], Horne [6], Neal [7]. The method of in-
equalities uses the equations of equilibrium along with the The number of independent mechanisms is Nmech the same
plastic limit states as a means of computing the collapse load as the number of independent equations of equilibrium Neq
of a structure. The method is actually a type of linear pro- for the structure. Using the method of virtual work in con-
gramming problem and can be solved using any linear pro- junction with these mechanisms provides a simple method to
gramming technique. In order formulate the linear program- derive the equations of equilibrium:
ming problem, one should be familiar with some of the basic
N
theorems of plastic analysis and design. The theorems are X
aij mj = bi (1.1)
stated in this paper without proof. Complete explanations
j=1
and derivations are found in Haque [3], Horne [6], Neal [7].
NX
loads
Several items of notation for plastic analysis are required.
The applied loads are denoted as Wj where j = 1, . . . , Nloads . bi = rij αj W L (1.2)
j=1
All loads are assumed to act simultaneously. Proportional
loading is assumed. This means that every load is a scaled where i = 1, . . . , Neq and Neq = Nmech is the number of in-
factor of a single positive load W . Each member of the frame dependent equations of equilibrium. These equations can be
has length Lk and plastic bending moment capacity Mpk where written in matrix form:
k = 1, . . . , Nmembers. These can be expressed by factors of
a basic length L and moment capacity Mp . Members must Am = b, b = W L Ra
be able to form a plastic hinge when the maximum absolute
bending moment reaches Mpk . They must also be designed to A = [aij ], m = [mj ], b = [bi ], R = [rij ], a = [αj ]
resist failure due to buckling and other instabilities at their
The coefficients aij and rij depend on the geometry of the
plastic moment capacity. Plastic hinges can only form at the
frame. These matrices and vectors are summarized in table
locations of applied point loads and at joints (i.e. connec-
2.
tions between members). The moments at these locations
are denoted mi where i = 1, . . . , N . The moments mi may
Name Symbol Size
be elastic or plastic. Since we assumed all applied loads are
point loads, the bending moments will vary linearly between Equilibrium coefficients A Neq × N
the locations of mi . By the above definitions, we clearly have
N ≥ Nloads and N ≥ Nmembers. We also require a map Bending moments m N ×1
K : {1, . . . , N } → {1, . . . , Nmembers} from the location of the Equilibrium loads b Neq × 1
moment mi to its corresponding member k. Thus k = K(i) is
the index of the member which contains the bending moment Geometric factors R Neq × Nloads
mi . Table 1 summarizes the notation used in plastic analysis.
Load factors a Nloads × 1
The notion of a mechanism is fundamental to plastic analy-
sis and design. A mechanism occurs when a sufficient number
Table 2: Matrices for equilibrium equations
of plastic hinges are formed that allow deflections to increase
without increasing applied load. Every mechanism will have
a minimum applied load which must be reached before the
mechanism can form. This load is not necessarily the collapse
load of the structure because another mechanism may form
at a lower value of applied load. The goal of plastic analysis is
to determine the actual mechanism which causes collapse of
the structure. Mechanisms are demonstrated in the examples
described in this paper. Further information on mechanisms
can be found in Haque [3], Hodge [5], Horne [6], Neal [7].

2
2. Plastic Analysis For computational purposes, it is convenient to introduce
dimensionless variables and parameters. Let m̂j = mi /Mp
2.1. Formulation K(j)
and β̂j = βK(j) = Mp /Mp . The applied load is a variable
Definition 2.1. The bending moments mi of a frame are of the problem and is written in the following manner:
statically admissible if they satisfy the equations of equilib-
rium with the applied loads Wj . Mp
W =λ
L
We also must ensure that the bending moments do no ex- The right hand side of the equations of equilibrium, in terms
ceed the plastic moment capacities of the corresponding mem- of dimensionless variables, are written as:
bers. We require a map K : {1, . . . , N } → {1, . . . , Nmembers}  
from the location of the moment mi to its corresponding mem- NX
loads

ber k. Thus k = K(i) is the index of the member which bi =  rij αj  W L


contains the bending moment mi . j=1
 
NX
loads
bi WL
Definition 2.2. The bending moments mi of a frame are =  rij αj 
Mp j=1
Mp
safe if they do not exceed their corresponding plastic moment
capacities.
 
NX
loads
bi
|mi | ≤ MpK(i) , i = 1, . . . , N =  rij αj  λ
Mp j=1
Recalling that all applied loads are proportional to the rep-
resentative load W , we can state the goal of plastic analysis bi
= b̃i λ
in terms of this load. Mp

where
NX
Definition 2.3. The smallest load W which causes the frame bi loads

to form a mechanism is called the collapse load Wc . b̃i = = rij αj


WL j=1

Theorem 2.1 (Static or Lower Bound Theorem). If a set of


loads on a frame produces bending moments which are stati- Theorem 2.4 (Linear Programming for Plastic Analysis).
cally admissible and safe, then the loads are less than or equal The collapse load Wc is the solution of the optimization prob-
to the collapse load. lem: Find bending moments m̂j , satisfying equations (2.2)
W ≤ Wc and inequalities (2.3), which maximize applied load factor λ.

λc = max λ (2.1)
Theorem 2.2 (Kinematic or Upper Bound Theorem). If a N
set of loads on a frame produces a mechanism, then the loads
X
aij m̂j = b̃i λ i = 1, . . . , Neq (2.2)
are greater than or equal to the collapse load. j=1

Wc ≤ W |m̂j | ≤ β̂j j = 1, . . . , N (2.3)

Proof. The equilibrium equations (2.2) and the inequalities


(2.3) constrain the optimization to statically admissible and
Theorem 2.3 (Uniqueness Theorem). Suppose a set of loads safe bending moments. The Lower Bound and Uniqueness
on a frame produces bending moments which are statically Theorems imply that equation (2.1) is the solution to the
admissible, safe, and results in a mechanism. Then the loads collapse load Wc = λc Mp /L.
equal the collapse load.

The Lower Bound Theorem states that every load Ws which


results in a statically admissible and safe distribution of bend-
ing moments is a lower bound for the collapse load. If S is
the set of all safe and statically admissible loads, then the
Uniqueness Theorem implies the largest such load would be
the collapse load:
Wc = max Ws
s∈S

This result is used to formulate plastic analysis as an opti-


mization problem.

3
2.2. MATLAB® /Octave Function This page intentionally left blank
panal.m solves the plastic analysis problem:

λc = max λ
Am = W
|m̂j | ≤ β̂j

where

A = [aij ], m = [m̂j ], W = [b̃j ], Mp = [β̂j ]

i = 1, . . . , Neq , j = 1, . . . , N

[m,Wc] = panal(A,W,Mp,pf)

Output Description

m Bending moments m
(N × 1 vector)
[Units of Mp ]

lambda Collapse load factor λc

Input Description

A Coefficient matrix A
(Neq × N matrix)

W Load factors W
(Neq × 1 vector)
[Units of W L]

Mp Plastic capacities Mp
(1 × N array)
[Units of Mp ]

pf Print flag

1 Print

0 Do not print

4
2.3. Continuous Beam
A two span continuous beam is shown in figure 1. The
plastic moment capacity of the left span is Mp1 = Mp . The
moment capacity of the right span is Mp2 = βMp . Let α ≥
0 be the load factor for the right span. There are N = 4
possible plastic hinges and r = 1 redundancy. The number of
independent mechanisms is Nmech = N − r = 3. The location
of the plastic hinges and independent mechanisms are seen in
figure 1.
The independent equations of equilibrium can be derived
using The Principle of Virtual Work on the independent mech-
anisms. These equations are valid regardless of the occurrence
of an actual plastic mechanism. The equations remain true
for purely elastic analysis. A sign convention for the virtual
angles is necessary. If the member would be in tension on
the side of the dotted line, then the associated rotation at
that location is considered to be positive. We assume that
the virtual angle θ is small and can approximate the virtual
displacement as ∆ = L2 θ. For the beam mechanism on the
Figure 1: Basic mechanisms
left span we have:

W∆ = m1 (2θ) + m2 (−θ)
 

W = (2m1 − m2 )θ
2
WL = 4m1 − 2m2

Another equation of equilibrium is derived in the same man- Figure 2: Collapse mechanism: α = 4, β = 2
ner using the beam mechanism on the right span:
Output from panal.m for α = 4, β = 2:
αW ∆ = m3 (−θ) + m4 (2θ)
  P l a s t i c a n a l y s i s f o r a lpha =4.00000 beta =2.00000

αW = (−m3 + 2m4 )θ Pl a s t i c Analysis
2 Bending Moments :
αW L = −2m3 + 4m4 ID m/Mp m/Mp_i
1 +0.12500 +0.12500
The final equilibrium equation is provided by considering the
2 −1.00000 −1.00000
joint mechanism. A counter-clockwise rotation of the joint at
3 −1.00000 −0.50000
the central support gives:
4 +2.00000 +1.00000
0 = m2 θ + m3 (−θ) C o l l a p s e Load F a c t o r : 2 . 5 0 0 0 0
0 = m2 − m3 The results are:
Mp
Wc = 2.5
The three equations of equilibrium are thus: L
1
4m1 − 2m2 = WL (2.4) m1 = − Mp , m2 = m3 = −Mp , m4 = 2Mp
8
−2m3 + 4m4 = αW L (2.5) Note that plastic hinges are formed at locations 2 and 4. The
m2 − m3 = 0 (2.6) resulting collapse mechanism is a combination of beam mech-
anism 2 and the joint mechanism. The mechanism is shown
The input to panal.m for this problem is as follows: in figure 2.
   
4 −2 0 0 1
A =  0 0 −2 4  , W =  α 
0 1 −1 0 0
 
Mp = 1 1 β β

5
2.4. Portal Frame
A portal frame is illustrated in in figure 3. Each column has
a plastic moment capacity Mp1 = Mp . The moment capacity
of the beam is Mp2 = βMp . The load factors are αV ≥
0, αH ≥ 0. There are N = 7 possible plastic hinges and
r = 3 redundancies. The number of independent mechanisms
is Nmech = N − r = 4.
The equations of equilibrium are derived using The Princi-
ple of Virtual Work. Let ∆ = Lθ be the virtual displacement
for small angles. The beam mechanism leads to the first equi-
librium equation:

αV W ∆ = m3 (−θ) + m4 (2θ) + m5 (−θ)


αV W (Lθ) = (−m3 + 2m4 − m5 )θ
αV W L = −m3 + 2m4 − m5

The second equilibrium equation is derived using the sway


mechanism:

αH W ∆ = m1 (−θ) + m2 θ
+m6 (−θ) + m7 θ
αH W (Lθ) = (−m1 + m2 − m6 + m7 )θ
αH W L = −m1 + m2 − m6 + m7

The first join mechanism results in the following equation:

0 = m2 θ + m3 (−θ)
0 = m2 − m3
Figure 3: Portal frame mechanisms
The second joint mechanism gives:
Output from panal.m for αH = 1×10−7 , αV = 1, β = 1/2:
0 = m5 θ + m6 (−θ)
0 = m5 − m6 Plastic analysis for
H=0.00000 V=1.00000 beta =0.50000
Thus the equations of equilibrium are: Pl a s t i c Analysis
Bending Moments :
−m1 + m2 − m6 + m7 = αH W L (2.7) ID m/Mp m/Mp_i
−m3 + 2m4 − m5 = αV W L (2.8) 1 +1.00000 +1.00000
2 −0.50000 −0.50000
m2 − m3 = 0 (2.9)
3 −0.50000 −1.00000
m5 − m6 = 0 (2.10) 4 +0.50000 +1.00000
5 −0.50000 −1.00000
The input to panal.m for this problem is: 6 −0.50000 −0.50000
    7 +1.00000 +1.00000
−1 1 0 0 0 −1 1 αH
 0 0 −1 2 −1 0 0  C o l l a p s e Load F a c t o r : 2 . 0 0 0 0 0
 αV 
A=  0 1 −1 0 0
, W =  
0 0   0  The results are:
0 0 0 0 1 −1 0 0 Mp
Wc = 2
L
Mp = [ 1 1 β β β 1 1 ] 1 1
m3 = m5 = − M p , m4 = M p
We wish to compute the solution using αH = 0. However, 2 2
numerical instability will occur using a near exact value of The beam mechanism is the resulting collapse mechanism.
zero. A small but nonzero value is required for both αH and Since the mechanism is a partial collapse, the frame continues
αV . Hence αH = 1 × 10−7 is used to approximate a zero to be statically indeterminate at collapse. The remaining mo-
horizontal load on the frame ments cannot be computed using linear optimization. Thus
the results for m1 , m2 , m6 , and m7 are invalid.

6
Output from panal.m for αH = 1, αV = 1 × 10−7 , β = 1:
Plastic analysis for
H=1.00000 V=0.00000 beta =1.00000
Pl a s t i c Analysis
Bending Moments : Figure 4: Collapse mechanism: αH = 2, αV = 3, β = 2
ID m/Mp m/Mp_i
1 −1.00000 −1.00000 Output from panal.m for αH = 2, αV = 3, β = 2:
2 +1.00000 +1.00000
3 +1.00000 +1.00000 Plastic analysis for
4 +0.00000 +0.00000 H=2.00000 V=3.00000 beta =2.00000
5 −1.00000 −1.00000 Pl a s t i c Analysis
6 −1.00000 −1.00000 Bending Moments :
7 +1.00000 +1.00000 ID m/Mp m/Mp_i
C o l l a p s e Load F a c t o r : 4 . 0 0 0 0 0 1 −1.00000 −1.00000
The results are: 2 +0.20000 +0.20000
Mp 3 +0.20000 +0.10000
Wc = 4
L 4 +2.00000 +1.00000
m1 = m5 = m6 = −Mp , m2 = m3 = m7 = Mp , 5 −1.00000 −0.50000
6 −1.00000 −1.00000
m4 ≈ (2 × 10−5 )Mp 7 +1.00000 +1.00000
The collapse is due to the sway mechanism. Notice that β = 1 C o l l a p s e Load F a c t o r : 1 . 6 0 0 0 0
means that every member shares the same plastic moment The results are:
capacity. It is possible that two plastic hinges are formed at Mp
Wc = 1.6
a joint connecting exactly two members of the same strength. L
However, in plastic analysis we usually represent this case as 1
a single hinge (Neal [7]). m1 = m5 = m6 = −Mp , m2 = m3 = Mp ,
5
m4 = 2Mp , m7 = Mp
The mechanism is a combination of sway, beam, and joint and
is illustrated in figure 4.

7
3. Minimum Weight Design Theorem 3.3 (Lower Bound on Minimum Weight). Suppose
a design experiences a weight compatible mechanism at the
3.1. Formulation applied loads. Then the weight of the design is a lower bound
Definition 3.1. A design D = Mp1 , . . . , MpNmembers , for the for the minimum weight.


loads Wj = αj W , is an admissible design if the resulting


W(D) ≤ W(Dmin )
bending moments are statically admissible and safe. The set
of all admissible designs is denoted as S.

Theorem 3.4 (Uniqueness Theorem for Minimum Weight).


Definition 3.2. The minimum weight design Dmin for the Suppose a design at the applied loads results in a weight com-
load W is the admissible design that produces the smallest patible mechanism and has a statically admissible and safe
total weight. bending moment distribution. Then the design is a minimum
weight design.
Nmembers
X
W(Dmin ) = min W(D), W(D) = f (Mpk ) Lk In order to formulate the linear programming problem for
D∈S
k=1
minimum weight design, the issue of design variables must be
f (Mpk ) is the weight per unit length and is a function of addressed. The design variables are the moment capacities of
plastic moment capacity. The minimum weight design will the frame members. However, it is possible that we require
exist but may not be unique. some members to share the same design variable. Let the
number of design variables be Nd . Obviously Nd ≤ Nmembers
and we require a map from members to design variables: V :
Theorem 3.1. If f (Mp ) is linear, then the minimum weight {1, . . . , Nmembers} → {1, . . . , Nd }.
design problem is equivalently stated as: In order to properly formulate weight compatible mecha-
nisms, the weight function for G is modified as follows:
  Nmembers
X
W(Dmin ) = γ min G(D) + B, G(D) = Mpk Lk Nd
X
D∈S
k=1 G(D) = Mpk Lk
k=1
where γ and B are constants.
Nmembers
X
Lk = Lj φ(j)
Theorem 3.2 (Upper Bound on Minimum Weight). Every j=1
admissible design (i.e. statically admissible and safe) provides (
1 if V(j) = k
a weight which is an upper bound for the minimum weight. φ(j) =
0 otherwise
W(Dmin ) ≤ W(D), ∀D ∈ S Mpk = βk W L are the design variables. G is the sum over
design variables rather than members. Lk is the sum of the
lengths of members sharing design variable k. φ is the choice
For this section we shall consider all plastic hinge angles function that ensures that only members with design variable
to be positive. Define θqk = θi be such that k = K(i) is the k are summed to compute Lk .
member that contains the plastic hinge at location i and q = Let m̃j = mj /W L and G̃ = G/W L2 be dimensionless mo-
Q(i, k) be the local hinge location index. Q is the mapping ments and weight respectively. The theorem that allows us
between the global index i to local index q within member k. to use linear programming can now be stated.
Then we can write the actual work done during collapse as:
Theorem 3.5 (Linear Programming for Minimum Weight
Nqk
NXloads Nmembers
X X Design). The minimum design weight problem is the solution
Wj ∆j = Mpk φk , φk = θqk to the optimization problem: Find bending moments m̃j and
j=1 k=1 q=1 design variables βk , satisfying equations (3.2) and inequalities
(3.3), that minimize weight G̃.
where Nqk is the number of plastic hinge locations within
member k. Note that Nqk is at least one but less than or G̃min = min G̃ (3.1)
equal to the number of possible hinge locations in member k. X N

The fact that the displacements are small allows the angles to aij m̃j = b̃j (3.2)
be linearized in terms of member lengths. The requirement j=1

on the global mechanism is that all plastic hinge rotations are |m̃j | ≤ βV(K(j)) (3.3)
a multiple of a single parameter.
Proof. The equilibrium equations (3.2) and the safe condi-
tion (3.3) ensure an admissible design. Note that G was con-
Definition 3.3. A weight compatible mechanism is a mech- structed to include weight compatible mechanisms. Thus the
anism in which φk = ωLk where ω is a positive constant. Upper Bound and Uniqueness Theorems for Minimum Weight
state that the solution exists.

8
3.2. MATLAB® /Octave Function This page intentionally left blank
minwt.m solves the minimum design weight problem:

G̃min = min G̃
Am = W
|m̃j | ≤ βV(K(j))

where
Nd
X
G̃ = βk L k
k=1

A = [aij ], m = [m̃j ], W = [b̃i ]


Mp = [βk ], L = [Lk ]
i = 1, . . . , Neq , j = 1, . . . , N, k = 1, . . . , Nd

[m,Mp,wt] = minwt(A,W,AMp,L,pf)

Output Description

m Bending moments m
(N × 1 vector) [Units of W L]

Mp Moment capacities Mp
(Nd × 1 vector) [Units of W L]

wt Minimum design weight [Units of W L2 ]

Input Description

A Coefficient matrix A
(Neq × N matrix)

W Load factors W
(Neq × 1 vector) [Units of W L]

AMp Map from moments to


design variables
(1 × N array)
AMp(i)= V(K(i))

L Length for each design variable


(1 × Nd array) [Units of L]

pf Print flag

1 Print

0 Do not print

9
3.3. Continuous Beam 3.4. Portal Frame

Figure 5: Collapse mechanism: α = 4 Figure 6: Collapse mechanism

The input to minwt.m is: The input to minwt.m is:


   
4 −2 0 0 1
  
−1 1 0 0 0 −1 1 αH
A =  0 0 −2 4 , W= α   0 0 −1 2 −1 0 0   αV 
0 1 −1 0 0 A=  0 1 −1 0 0
, W= 
0 0   0 
0 0 0 0 1 −1 0 0
AMp = [ 1 1 2 2 ], L=[ 1 1 ]
   
Output from minwt.m for α = 4: AMp = 1 1 2 2 2 1 1 , L= 2 2
Minimum weig ht d e s i g n f o r a lpha =4.00000
Minimum Weight Desig n Output from minwt.m for αH = 2, αV = 3:
Bending Moments :
ID WL WL/Mp_i Minimum weig ht d e s i g n f o r H=2.00000 V=3.00000
1 +0.16667 +1.00000 Minimum Weight Desig n
2 −0.16667 −1.00000 Bending Moments :
3 −0.16667 −0.18182 ID WL WL/Mp_i
4 +0.91667 +1.00000 1 −0.83333 −1.00000
Moment C a p a c i t i e s : 2 −0.50000 −0.60000
ID WL 3 −0.50000 −0.60000
1 +0.16667 4 +0.83333 +1.00000
2 +0.91667 5 −0.83333 −1.00000
Minimum Weight : 1 . 0 8 3 3 3 6 −0.83333 −1.00000
7 +0.83333 +1.00000
The results are: Moment C a p a c i t i e s :
13 1 11 ID WL
Gmin = W L2 , Mp1 = W L, Mp2 = WL 1 +0.83333
12 6 12
2 +0.83333
The weight compatible collapse mechanism is is shown in fig- Minimum Weight : 3 . 3 3 3 3 3
ure 5.
The results are:
10 5
Gmin = W L2 , Mp1 = Mp2 = WL
3 6
The weight compatible collapse mechanism is is shown in fig-
ure 6.

10
4. Variable Repeated Loading Theorem 4.4 (Uniqueness Theorem for Shakedown). Sup-
pose that at the load factor λ, it is possible to find a dis-
4.1. Formulation tribution of residual bending moments m̄∗i that are statically
Definition 4.1. If any members of a frame fail due to alter- admissible and satisfy the inequalities listed below. If a mech-
nating application of applied loads, then the frame has failed anism is produced at this load factor, then it is equal to the
due to alternating plasticity. shakedown limit λs .

m̄∗i + λMmax
i ≤ MpK(i)
Definition 4.2. Suppose that a frame is subjected to variable
repeated loads. If plastic hinge rotations continue to increase m̄∗i + λMmin
i ≥ −MpK(i)
max min
2 MyK(i)

without bound, then the frame will fail due to incremental λ Mi − Mi ≤
collapse.
We shall assume an ideal elastic-perfect-plastic relation
K(i) K(i)
Definition 4.3. Suppose during application of variable re- for the stress-strain curve. Thus we set My = Mp .
peated loads, a frame ceases to produce additional plastic Dimensionless variables and parameters for shakedown are:
hinge rotation and all subsequent changes of bending mo- m̂∗j = m̄∗j /Mp , M̂max
j = Mmax
j /W L, M̂min
j = Mminj /W L,
ments are entirely elastic. Then the frame is said to have and β̂j = βK(j) = MpK(j) /Mp . The units of λ are Mp /W L.
shaken down.
Theorem 4.5 (Linear Programming for Shakedown). The
In order to state the Shakedown Theorem, a few items of value of the shakedown limit λs is determined by the solution
notation are introduced. Let Mmax i and Mmin
i be the max- of the optimization problem: Find residual bending moments
imum and minimum elastic bending moments that occur at m̂∗i , that satisfy equations (4.5) and inequalities (4.6)-(4.8),
location i for any combination of applied loads. These mo- which maximize λ.
ments are computed from an initial stress-free state and can
be positive or negative. The fact that they are elastic bend- λs = max λ (4.4)
ing moments means that their absolute value can exceed the Neq
K(i)
plastic moment capacity Mp . The value at which plastic
X
aij m̂∗j = 0 (4.5)
K(i)
yielding begins is notated as My . For the ideal elastic- i=1
K(i) K(i)
plastic relationship, we set My = Mp . m̂∗j + λM̂max
j ≤ β̂j (4.6)
+ m̂∗j λM̂min
≥ −β̂j
j (4.7)
Theorem 4.1 (Shakedown Theorem). If it is possible to find  
a distribution of residual bending moments m̄∗i that are stati- λ M̂max
j − M̂min
j ≤ 2 β̂j (4.8)
cally admissible and satisfy the following inequalities, then the
frame will eventually shake down. Proof. Equation (4.5) and inequalities (4.6)-(4.7) state that
the residual bending moments are statically admissible. The
m̄∗i + Mmax
i ≤ MpK(i) (4.1) Lower Bound and Uniqueness Theorems for Shakedown com-
m̄∗ + Mmin ≥ −M K(i)
i i p (4.2) plete the proof.
Mmax
i − Mmin
i ≤ 2 MyK(i) (4.3)
It should be noted that the solution to the linear program-
The inequality (4.3) ensures that the frame does not fail ming problem does not indicate which collapse mechanism oc-
due to alternating plasticity. curs at shakedown. Also, the residual bending moments are
not necessarily those at shakedown. More information con-
cerning shakedown can be found in Haque [3], Horne [6], Neal
Theorem 4.2 (Lower Bound Theorem for Shakedown). The
[7].
load factor λ is a lower bound on the shakedown limit if it
is possible to find a distribution of residual bending moments
m̄∗i that are statically admissible and satisfy the following in-
equalities

m̄∗i + λMmax
i ≤ MpK(i)
m̄∗i + λMmin
i ≥ −MpK(i)
λ Mmax − Mmin ≤ 2 MyK(i)

i i

Theorem 4.3 (Upper Bound Theorem for Shakedown). Sup-


pose that incremental collapse occurs due to an assumed col-
lapse mechanism. Then the load factor λ for the assumed
mechanism is an upper bound for the shakedown limit λs .

11
4.2. MATLAB® /Octave Function This page intentionally left blank
shakedown.m solves the shakedown problem:

λs = max λ
Am = 0
m̂∗j + λM̂max
j ≤ β̂j
m̂∗j + λM̂min
j ≥ −β̂j
 
max min
λ M̂j − M̂j ≤ 2 β̂j

where
A = [aij ], m = [m̂j ]
Mmax = [M̂max
j ], Mmin = [M̂min
j ], Mp = [β̂j ]
i = 1, . . . , Neq , j = 1, . . . , N

[m,lambda] = shakedown(A,Mp,Mmax,Mmin,pf)

Output Description

m Bending moments m
(N × 1 vector)
[Units of Mp ]

lambda Shakedown limit λs

Input Description

A Coefficient matrix A
(Neq × N matrix)

Mp Plastic capacities Mp
(1 × N array)
[Units of Mp ]

Mmax Maximum elastic moments


Mmax
(1 × N array)
[Units of W L]

Mmin Minimum elastic moments


Mmin
(1 × N array)
[Units of W L]

pf Print flag

1 Print

0 Do not print

12
4.3. Continuous Beam
The loading pattern for the continuous beam is shown in
figure 7. The maximum and minimum elastic bending mo-
ments are given in table 3. The input to shakedown.m is:
 
4 −2 0 0
A =  0 0 −2 4 
0 1 −1 0
 
Mp = 1 1 1 1
 5 13

Mmax = 32 0 0 64
3 3 3 5
 
Mmin = − 64 − 16 − 16 32
Figure 7: Loading patterns for beam

The output from shakedown.m :


Shakedown f o r a lpha =1 , beta=1
Shakedown i=1 i = 2, 3 i=4
R e s i d u a l Moments :
ID m/Mp m/Mp_i 1st Pattern 5 3
− 16 5
32 32
1 −0.02632 −0.02632
2 −0.05263 −0.05263 3 3 13
3 −0.05263 −0.05263 2nd Pattern − 64 − 32 64
4 −0.02632 −0.02632
Shakedown Limit : 5 . 0 5 2 6 3 Zero Loading 0 0 0
The shakedown limit is:
5 13
Mp Mmax 32 0 64
λs = 5.05263
WL
3 3 5
Mmin − 64 − 16 32

13 3 3
Mmax − Mmin 64 16 64

Table 3: Elastic bending moments (units of W L)

13
4.4. Portal Frame
The loading pattern for the portal frame is shown in figure
8. The maximum and minimum elastic bending moments are
given in table 4. The input to shakedown.m is:
 
−1 1 0 0 0 −1 1
 0 0 −1 2 −1 0 0 
A=  0 1 −1 0 0

0 0 
0 0 0 0 1 −1 0
Mp = [ 1 1 1 1 1 1 1 ]
 
Mmax = 0.3 0.375 0.375 0.9 0 0 0.925
 
Mmin = −0.625 −0.6 −0.6 0 −0.975 −0.975 0

The output from shakedown.m :


Shakedown f o r H=1 , V=1 , beta=1
Shakedown
R e s i d u a l Moments :
ID m/Mp m/Mp_i
1 −0.29245 −0.29245
2 −0.14151 −0.14151
3 −0.14151 −0.14151
4 −0.01887 −0.01887 Figure 8: Loading patterns for frame
5 +0.10377 +0.10377
6 +0.10377 +0.10377
7 −0.04717 −0.04717
i=1 i = 2, 3 i=4 i = 5, 6 i=7
Shakedown Limit : 1 . 1 3 2 0 8
The shakedown limit is:
1st Pattern 0.300 −0.600 0.900 −0.600 0.300
Mp
λs = 1.13208
WL
2nd Pattern −0.625 0.375 0 −0.375 0.625

3rd Pattern −0.375 −0.225 0.900 −0.975 0.925

Zero Loading 0 0 0 0 0

Mmax 0.300 0.375 0.900 0 0.925

Mmin −0.625 −0.600 0 −0.975 0

Mmax − Mmin 0.925 0.975 0.900 0.975 0.925

Table 4: Elastic bending moments (units of W L)

14
5. Linear Programming in MATLAB® and Octave 5.1.5. Continuous Beam

5.1. Functions [lambdac,wt,lambdas] = beam(alpha,beta)


5.1.1. Copyright
Copyright ©2015 by Aamer Haque. All rights reserved.
Output Description
(Copyright does not include internal MATLAB® and Oc-
tave functions such as linprog and glpk)
lambdac Collapse load factor λc
MATLAB® is the intellectual property of MathWorks, Inc.

Octave is distributed under the terms of the GNU General wt Minimum weight G̃min
Public License.
lambdas Shakedown limit λs
5.1.2. Disclaimer and Limits of Liability
All of the code and software used for this paper are pro-
vided on an "as-is" basis. No guarantee is made concerning
the accuracy, reliability, or correctness of the code and soft- Input Description
ware. The user accepts any and all risks associated with the
use of the code and software. The code and software should
not be used in any circumstances where life or property may alpha Load factor α
be at risk. The code and software are not a substitute for
proper engineering analysis and judgement. Proper consulta- beta Moment capacity factor β
tion with a licensed professional engineer is necessary for any
design work. The developer assumes no responsibility and ac-
cepts no liability for the use of the code and software. Use at 5.1.6. Portal Frame
your own risk!
[lambdac,wt,lambdas] = portal(H,V,beta)
5.1.3. User License Agreement
The user must agree to the following user license agreement:
Output Description
• User will use the code for educational or academic pur-
poses. No commercial use is allowed.
lambdac Collapse load factor λc
• User will properly acknowledge and reference the use of
the code in any publication or work that uses the code.
(e.g. journal article, technical report, homework) wt Minimum weight G̃min

• User will not redistribute modified or unauthorized ver-


sions of the code. User is free to distribute the original lambdas Shakedown limit λs
code.

• User will not incorporate the code as part of their own


code or software. Input Description
• User has read, understood, and agreed to the terms out-
lined above (including the disclaimer and limits of liabil- H Load factor αH
ity).
• You will not hold the software developer responsible for V Load factor αV
any use or misuse of the code or software.
beta Moment capacity factor β
5.1.4. Download

Matlab:
www.clearlyimpossible.com/ahaque/matlab_linprog.tar.gz

Octave:
www.clearlyimpossible.com/ahaque/octave_linprog.tar.gz

15
5.2. MATLAB® linprog 5.3. Octave glpk

[xmin,f] = linprog(c,A,b,Aeq,beq,lb,ub) [xopt,f] = glpk(c,A,b,lb,ub,ctype,vartype,s)

Output Description Output Description

xmin Optimal solution xmin xopt Optimal solution xopt


(n × 1 vector) (n × 1 vector)

f Function value fmin = cT xmin f Function value fopt = cT xopt

Input Description Input Description

c Coefficient vector c c Coefficient vector c


(n × 1 vector) (n × 1 vector)

A Matrix A for Ax ≤ b A Matrix A


(m × n matrix) (m × n matrix)

b RHS vector b for Ax ≤ b b RHS vector b


(m × 1 vector) (m × 1 vector)

Aeq Matrix Aeq for Aeq x = beq lb Lower bounds for x


(m × n matrix) (1 × n array)

beq RHS vector beq for ub Upper bounds for x


Aeq x = beq (1 × n array)
(m × 1 vector)
ctype Constraint type
(m character string)
lb Lower bounds for x
(1 × n array) "S" Equality constraint
A(i,:)*x=b(i)
ub Upper bounds for x
(1 × n array) "L" Inequality constraint
A(i,:)*x>=b(i)

"U" Inequality constraint


A(i,:)*x<=b(i)

vartype Variable type


(n character string)

"C" Continuous variable

s Optimization type
(integer)

+1 Minimization

-1 Maximization

16
References

[1] John Baker and Jacques Heyman. Plastic Design of


Frames 1, Fundamentals. Cambridge University Press,
Cambridge, England, 1969.
[2] Lynn S. Beedle. Plastic Design of Steel Frames. Wiley,
London, 1958.
[3] Aamer Haque. Proofs of the theorems of plastic analysis
and design. 2015.
[4] Jacques Heyman. Plastic Design of Frames 2, Applica-
tions. Cambridge University Press, Cambridge, England,
1971.
[5] Philip G. Hodge. Plastic Analysis of Structures. McGraw-
Hill, New York, 1959.
[6] Michael R. Horne. Plastic Theory of Structures. M.I.T.
Press, Cambridge, Massachusetts, 1971.
[7] B.G. Neal. The Plastic Methods of Structural Analysis.
Wiley, New York, second edition, 1963.

17

Vous aimerez peut-être aussi