A Micro-Computer Program For The Elastic-Plastic Analysis and Optimum Design of Plane Frames

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

A MICRO-COMPUTER PROGRAM FOR THE

ELASTIC-PLASTIC ANALYSIS AND


OPTIMUM DESIGN OF PLANE FRAMES

H, SCHOLZand C. FALLER
~par~ent of Civil ~ngin~~~g, Unive~ity of the W~twatersrand, Johann~burg, South Africa

(Received 28 January 1986)

Abstract-A computerized version of an approximate second-order, elastic-plastic analysis of multi-storey


sway frames is pmsented. The computer program which applies the method to steel frames on a
story-b~sto~ey basis is described, The program is used to d~o~st~~e bv way of an example the
appli~bility of the method for the analysis and design of large, multi-stomy frames. It is argued that the
tecbniaue can meat& reduce analysis time as well as simplify the optimi~ti~n of members in the design
of suci fram&

1. l[NTRODUCTION The increasing availability of high-speed digitial


computers has encouraged the development of com-
The analysis of ~multi-storey frameworks is compli- puter techniques which analyse the whole structure
cated by the need to take second-order effects into through the plastic range to failure (3,4$ These are
account. The most sibilant and-order effect, the rigorous analyses which can take most secondary
P-Delta effect, causes additional moments when the effects into consideration. The fact that the entire
axial forces on the columns are laterally displaced structure is analysed as a whole makes this approach
through storey sway. cumbersome for the design of large multi-storey
The conventional way of ~rn~u~ting for the frames. Changing one member in the framework may
P-Delta effect in the design of multi-storey sway necessitate the re-anaiysis of the whole frame and a
frames is through the “effective length” technique. large number of analyses would have to be performed
The frame is analysed using a first-order elastic to op~~ze all the members.
analysis and the effective column lengths are used Approximate second-order analyses have been de-
with these forces in the design of members to ensure veloped which are applied to ~am~olumn sub-
frame stability, i.e. to account for the P-Delta effect. assemblages[5~. Tbese methods are time consuming
It is well known that greater design anomies can in that they require the full plot of the load-displace-
be achieved by taking full plasticity into account in ment history and are not very accurate&], Cheong-
the design of frames. The “effective length” concept Siat-Moy[q developed a storey stiffness method
combined with a first-order elastic analysis does not which can be applied to the design of multi-storey
enable the designer to utilize fully the load carrying frames. This method is used on a sto~~by-story
capacity of the st~ctu~ in the plastic range. Recently basis by plotting the load vs storey stiffness curve for
it has been shown by Scholz[I) and Cheon~Siat- each storey. The storey-by-storey approach adopted
Moy[2], that the “effective length” technique has by this method greatly simplifies the design of multi-
more serious shortcomings. They have demonstrate storey frames but the plotting of each storey load vs
that designing structural members based on the storey stiffness curve is still tedious.
“effective length” concept does not always ensure Interaction methods make use of the rigid-plastic
overall frame stability. collapse load and the elastic buckling load of the
It seems, therefore, that a more realistic design structure to predict the elastic-plastic failure load.
procedure should be based on a second-order The Me~hant-Rankine method is the most widely
elastic-plastic analysis. A considerable amount of used interaction method. The shortcoming of this
work has been done in trying to develop suitable method are its i~accu~cy for slender frameworks and
second-order a~lysis t~h~ques which can be incor- for cases of pure vertical loading. Furthe~o~, the
porated in the design of multi-storey sway frames. calculation of the rigid plastic collapse load for large
Three different approaches to the problem can be frameworks is very time consuming. Scholz[8] has
identified: developed an improved version of the Merchant-
Rankine interaction method which can be applied to
(I) Rigorous elastic-plastic analysis of the entire the structure as a whole or on a story-by-story
framework. basis. This method is not subject to the limitations
(2) Approximate second-order analysis applied to of the Merchant-Rankine method and by applying
suitable su~as~mblag~. the method on a storey-by-stony basis the analysis
(3) Application of interaction formulae. computations and design optimi~tions are greatly
941
942 H. SCHOU and G. FALLEX

simplified. A brief summary of this method is them to form the storey sub-structure stiffness matrix.
presented later on. The load-displacement relationship is used to calcu-
This paper describes the computerization of the late the member forces and moments on a first-order
interaction method developed by Scholz [8] applied basis, i.e. ignoring the effect of displacements on axial
on a storey-by-storey basis. The program is then used forces. When the fully-plastic moment capacity of a
to demonstrate the interaction method applied to the section is reached at a certain point in a member, a
design of a multi-storey sway frame. ~roughout this plastic hinge is inserted at this discrete point while the
study the objective has been to provide an approxi- rest of the member remains elastic. Loading is applied
mate elastic-plastic analysis technique utilizing all incrementally and a first-order elastic analysis is
possible simplifications in order to minimize com- performed at loads between the formation of con-
puter and design time without loss in accuracy. secutive hinges. With the formation of each con-
secutive hinge the stiffness of the structure is reduced
2. COMPUTER PROGRAM
to account for the hinges and the stiffness matrix
adjusted accordingly. The rigid-plastic collapse load
The program analyses the structure one storey at of the structure is the load at which a hinge forms,
a time, from the top storey down. A typical internal causing the determinant of the structure stiffness
storey model is shown in Fig. 1 with points of matrix to become negative or zero.
inflection assumed at mid-column height. The calcu- In the analysis of the storey sub-structure (Fig. I),
lation of the elastic-plastic failure load factor for the effect of axial load on the stiffness of the members
each storey is performed in three stages: is ignored. It has been assumed that points of
inflection occur at mid-column height, except in the
(I) Calculation of the storey rigid-plastic collapse bottom storey. As a result of these simplifications, the
load factor, yP. only unknown defo~ations which have to be solved
(2) Calculation of the storey elastic buckling load are the rotations of the beam-column intersection
factor, yP joints and the lateral deflections of the columns.
(3) Using rP and yOin the interaction method [8] All axial deformations have been ignored so that
the elastic-plastic storey failure load, rr, is the lateral deflection, AZ, of all the columns at a floor
obtained. level are equal. Similarly, the lateral deflection, A,, of
all the columns at mid-column height of the upper
The lowest elastic-plastic failure load factor ob- coiumns are assumed to be equal. A storey sub-
tained for any one stony is then taken as the failure structure having n bays (i.e. n + I columns) thus has
load factor of the structure. A flow-chart outlining only (n + 1) + 2 = n + 3 unknown defo~ations to
the program structure is shown in Fig. 2. The estab- be solved. The size of the accompanying stiffness
lishment of the required parameters is outlined be- matrix is (n + 3) x (n + 3).
low. A load factor equal to unity is assigned to an When a plastic hinge forms in a member, the
arbitrary load level, usually the working load. technique presented by Liang and Yuan [4) for ad-
justing the structure stiffness matrix is used. In this
Rigid-plastic collapse load faetor method, no additional unknowns are introduced into
The rigid-plastic collapse load factor is calculated the computation when the stiffness matrix is adjusted
by utilizing the sio~e~~tion equation, without to accommodate plastic hinges. The size of the matri-
stability functions, for each member and combining ces involved in the computations is thus minimized

p, W-J) P, (W P, / (W

I i

w1 (kN/m)

PI I .
ez +
I

Fig. 1. Typical internal storey model.


Elastic-plastic analysis and optima design of plane frames 943

Next
Stosey

Fig. 2. Flow chart of computer program.

and the analysis can be performed easily with any from the relationship:
modem desk-top computer.
A!fP= l.lWf,(l - P/PJ, (2)
The load at which each plastic hinge forms is found
using the iteration procedure outlined by Jennings where MP = plastic moment capacity ignoring axial
and Majid [3]. Consider the bending moment at one loads, P = axial load on the member under consid-
end of a member to be m, for the first iteration and eration and P, = axial yield load of the member, i.e.
m, for the second iteration. If the corresponding load area times yield stress. The relationship for predict-
factors are yi and or, we can predict the load factor ing the load at fo~ation of a plastic hinge has to be
at the formation of a plastic hinge at this section by adjusted to include the reduction of MP under axial
using the relationship: loads when apphed to the column members.
The calculation of the rigid-plastic collapse load of
_ Yz-71 MP+YIzI;mt, the storey can thus be summarized as follows:
Y (1)
I m2-4 I 2 I
l Set up the stiffness matrix [K’J and the force
This relationship is used for beam members where matrix {F) for the structure under the applied
axial toads are ignored. For column members, the loading. Defme a unit load factor.
plastic moment capacity Icr, decreases with in- 0 Calculate the unknown defo~ations (D,f)
creasing axial load. This is taken into consideration from the relationship (Of) = (a-r(F). Use
in the program by calculating the column plastic these values to calculate the resulting forces
moment capacity for bending about the major axis and moments, m, .
944 H. SCHOLZ and G. FALLER

l Increase the load factor and repeat the calcu- yF *FAILURE LOAD FACTOR

lation, obtaining m, . yp m PLASTIC COLLAPSE LOAD FACTOR

l Using the moments M, and m, from the steps y. l ELASTIC BUCKLING LOAO FACIOR

tx *EON (1.2)
above, interpolate using eqn (1) to identify the
load factor at which the next plastic hinge
forms.
0 Adjust the stiffness matrix to include the new
plastic hinge and repeat the process.
0 When the inclusion of the new plastic hinge
results in the formation of a collapse mech- 0.6

anism it is registered by a change in sign of the


dete~inant of the stiffness matrix and the t+y,
0,4
rigid-plastic collapse load factor has been
found.

A graphic representation of the plastic collapse


mechanism is then offered by the program. This
picture shows the sequence of formation and posi- 0
0 0.2 0.L 0.6 0.6 1.0
tions of all the plastic hinges. When using the pro-
gram for design, the hinge locations will show the YF
“YO
designer in which members the plastic hinges causing
the collapse mechanism formed. By adjusting these Fig. 3. Basic interaction curves [8].
members the designer can then increase or decrease
the elastic-plastic failure load of the structure. rigid-plastic collapse load and elastic buckling load
of the given structure and the parameters of a so-
Elastic buckling load
called “limiting frame”. These are entered in frame
The elastic buckling load factor y,, is calculated interaction curves similar to single column curves.
using Cheong-Sat-Moy’s storey stiffness method [‘I]. The “limiting frame” is described as a more slender
The prediction of the critical load is made simple by version of the actual frame which attains first yield as
considering buckling instability as a function of its failure load (which is closely related to its elastic
storey stiffness. The elastic critical load of a typical buckling load) is applied i.e. failure and first yield
internal storey such as that shown in Fig. 1 is coincide for the “limiting frame”. It has been shown
obtained from the relationship (for a storey with m that each frame belongs to a certain frame family
columns): which is characterized by a unique “limiting frame”
and curve in the multi-curve frame interaction
graph [S]. The multi-curve interaction graph of Fig. 3
is reproduced from this reference. A brief summa~
of the required computation is listed below and
in which E = Young’s modulus, h = storey height, further elaborated in the Appendix.
Z, = moment of area of the column, U = the ratio of
the sum of the axial loads in the column above and 0 An approximate second-order elastic analysis
below floor level to the load in the column below is performed on the storey sub-structure with
floor level and II/ = the ratio of the column stiffness loading related to the elastic buckling load.
to the sum of the stiffnesses of the adjoining beams. Elastic second-order effects are taken into ac-
The term [Z,/( 1 + V$)] is calculated for each column- count by ma~if~ng the horizontal load. In the
restraining beam su~as~mblage and then summed presented program this step is performed by
over the m columns in the storey. The left-hand side factoring the elastic moments previously calcu-
of the equation, y&Z’, gives the total load that will be lated in the computation of the rigid-plastic
applied to the storey when elastic buckling occurs. collapse load, but prior to any hinge for-
The buckling load factor, yO, is then obtained by mation.
dividing yJP by CP, the load acting on the storey l The slenderness of the real frame and of the so
factored by a load factor equal to unity. called ‘“limiting frame” are calculated and the
lowest real f~me~limiting frame” slenderness
E~~t~c-~~~tic failure load ratio is identified.
The interaction method proposed by Scholz[SJ is @ The elastic buckling load/~~d plastic collapse
now used with y,, and yOto calculate the elastic-plastic load ratio is calculated for the “limiting frame”
failure load factor for the storey. This method is fully as well as for the actual structure.
documented in the literature[S] and will not be @ The elastic buckling load/plastic collapse load
presented in detail here. The method predicts the ratios of the real frame and of the “limiting
elastic-plastic failure load by making reference to the frame” are entered in the frame interaction
Elastic-plastic analysis and optimum design of plane frames 945

13.8kN Columns

Level Int. Ext.

l-3 8W13 8W13


3-5 8W20 8W20
S-7 8W28 8W28
7-9 8W35 8W35
9-11 8W40 8K40

Level Beams

1-2 12w22
3-4 14W22
5-8 14W26
9-10 16W26

Roof load = 20.43 kN/m E = 200.8 GPa


Floor load = 23.9& kN/m f = 250.0 MPa
Yb
Wind load = 5.84 M/m height f = 345.0 MPa
YC
Exterior column
wall load = 42.25 kN/storey
Storey height = 2.896 m
Bay width = 6.096m

Fig. 4. Frame designed by Cheong-Siat-Moy [7].

curves to read off the storey failure load. For analysis would identify which storeys have an
the purpose of this study the curve values have elastic-plastic failure load factor lower than or exces-
been stored in the computer. sively higher than the desired design load factor. Each
storey would then by taken in turn and a member or
A considerable saving in computer time is achieved members adjusted until the storey failure load factor
over rigorous elastic-plastic methods due to the fact is within acceptable limits above the design load
that no axial forces in the form of stability functions factor.
appear in the stiffness matrix. Because of this, rig A ten-storey high, three-bay wide frame, previously
orous elastic-plastic analysis requires approximately designed by Cheong-Siat-Moy [7l, is taken to demon-
five to seven iterations within each load increment strate the design of multi-storey sway frames by the
compared to a single matrix evaluation for each load presented interaction technique. The example frame
level in the presented technique. Thus, assuming, for is shown in Fig. 4. The frame members were first
instance, that 10 load increments are investigated proportioned for a load factor of 1.7 under pure
between the commencing load level and failure, 40-60 gravity loads and then certain members were adjusted
matrix evaluations are avoided, i.e. a saving of at to satisfy a load factor of 1.3 under combined gravity
least 80%. In addition, since the presented method is and wind loading. Using the computer program
performed on one storey at a time, a much smaller presented in this paper, this frame was t-e-analysed on
matrix is involved compared with the complex matrix
of the entire structure. Table I. Storey-by-storey analysis on frame
designed by Cheong-Siat-Moy [7]

3. APPLICATION TO ANALYSIS Storey failure load factors


AND DESIGN storey Plastic Elastic Elastic-plastic
I 2.17 32.12 2.11
The first step in the design process would be to use 2 1.95 13.27 1.81
some simplified method to obtain trial member sizes 3 2.03 12.93 I .86
for the whole structure. A suggested procedure would 4 1.83 9.36 I .63
be to use an elastic analysis and a plastic sectional 5 1.79 9.69 1.58
6 1.72 7.% I .49
design such as is embodied in modem load and 7 1.64 7.60 1.41
resistance factor approaches (limit state design) or a 8 1.56 6.59 1.32
rigid-plastic analysis and design approach. The struc- 9 1.50 6.95 I .28t
ture would then be analysed on a storey-by-storey 10 1.44 11.11 1.28
basis using the programmed interaction method. This t Structure failure load factor = 1.28.
H. SCHOUand G. FALLF.R

13.8kN All columns are


- identical to
those shown in
Fig.4

Level Beams

: lOB15
12B16.5
3 8W24
4 12319
5 lOM22.9
6 12w22
7 14W22
8 16W26
9 161326
10 16W50

Roof load = 20.43 kNfm E = 200.8 GPa


Floor load = 23.94 M/m = 250.0 MPa
Wind load = 5.84 kN[m height = 345.0 MPa
Exterior column
wall load = 84.50 kN/storey
Storey height = 2.896 m
Bay width = 6.096 m

Fig. 5. Frame optimized for design load factor = 1.26.

a storey-by-storey basis. The results of this analysis computer program. The accuracy of the design by
are shown in Table 1. the storey-by-storey interaction method was sub-
Assuming for the purpose of this example one were sequently tested by performing a rigorous second-
aiming for a design load factor of 1.26 for the com- order, elastic-plastic analysis on the full frame shown
bined loading case only, it is evident from Table 1 in Fig. 5. The result of this analysis was a structure
that the members in the upper storeys could be con- failure load factor of 1.27. This is in excellent agree-
siderably reduced. Each storey was therefore taken in ment with the storey-by-storey approach of the
turn and optimized to the design load factor of 1.26, present method.
using American standard sections. For simplicity all It shouid be pointed out that the program can also
the beam members in any one storey were kept the handle the load case of pure gravity load. In this
same while the column members were not adjusted at instance the resulting failure load is an approxi-
all. The optimized member sizes are shown in Fig. 5 mation of the inelastic buckling load of the frame-
while the corresponding failure load factors for each work.
storey are given in Table 2.
The lowest storey failure load factor in the opti-
mized frame is 1.26 and thus this is also taken as 4. SUMMARYAND CONCLUSIONS
the structure failure load factor identified by the
A computerized version of the interaction method
Table 2. Storey-by-storcy analysis of opti- developed by Scholz [8] applied on a storey-by-storey
mized frame basis has been presented. The computer program
Storey failure load factors provides a simple approach to the second-order,
storcy Plastic Elastic Elastic-plastic elastic-plastic analysis of sway frames subjected to
1.29
gravity or combined loading. By using this method
I 1.33 21.98
2 I.39 10.42 1.29 many of the shortcomings of alternative methods are
3 1.45 7.26 1.30 overcome and analysis time is greatly reduced. The
4 l-47 6.92 1.29 ease with which the method can be applied to the
5 1.48 5.81 1.26t design of large multi-storey frames has been demon-
6 1.53 5.70 1.28
7 I.51 6.16 1.29 strated. A frame designed on a storey-by-storey basis
8 1.53 6.17 1.30 with the presented interaction method was analysed
9 1.50 6.38 1.27 using a rigorous second-order analysis and found to
IO 1.39 II.68 1.28 be marginally on the conservative side. It is believed
t Structure failure load factor = 1.26. that this method can provide the designer with a
Elastic-plastic analysis and optimum design of plane frames 947

powerful and simple method of optimizing the design APPENDIX

of large multi-storey frames. Fur&mental concept of interaction method


In its present format the program only covers
The characteristic features of the curves of Fig. 3 can be
two-dimensional plane frames. Furthermore, it is summarized as follows:
assumed that the frame is subjected to proportionally
(1) Each curve of Fig. 3 belongs to a specific family of
increasing loads. Local member aspects and lateral frames.
torsional buckling are excluded. (2) The ordinate section on the right vertical axis belongs
Discrete studies have shown that the procedure can to frames within a frame family that fail completely
also be applied to three-dimensional frameworks elastically.
made up of frames on an intersecting rectangular grid (3) The curved portion of the interaction arauh encom-
passes frames failing inelastically. - .
and subjected to torsion as well as to load cases
(4) The curve intersection point at iF/yp = 1 on the left
involving non-proportional loading. Further research vertical axis of Fig. 3 signifies a frame that is not
is being conducted to incorporate local member subjected to P-Delta effects, i.e. rigid-plastic analysis
aspects and lateral torsional member buckling. is appropriate.
(5) The transition between elastic and inelastic failure is
marked by the so-called “limiting frame” of the
frame family. This unique frame for each family of
frames is identifiable by equating elastic failure with
first-yield. This leads to the identification parameter,
Acknowledgement-Thanks are due to Professor B. van A,, given in eqn (A.]):
Rensburg for his help in applying his elastic-plastic com-
puter program to the numerical example. i,=$;[;+;{l +&r2], (A.])

where

1,= limiting slenderness ratio of limiting frame


REFERENCES y = distance from centroid of section to extreme
fibre
1. H. E. Scholx, Calculation of effective length of columns r = radius of gyration
in u&raced frames using storey stiffness. Civ. &gr S. A4 = elastic second-order bending moments
Afr. 24, 137-142 (1982). P = axial member force
2. F. Cheong-Siat-Moy, Frame design without using L = member length
dfective column length. J. Srrucr. Div., ASCE !STll, I = second moment of area
23-33 (1978). fp = stress at onset of yield.
3. A. Jennings and K. I. Majid, An elasticplastic analysis
by computer for framed structures loaded up to col- The largest ratio 1,/1 for any one member is
lapse. Srrucf. Engr 43, 407-412 (1965). significant, where 1 = slenderness of the actual
4. Q. Liang and S. Yuan, A new method of second-order member. Simplifications for eqn (A.1) have been
elastic-plastic analysis of steel rigid frames in tall build- suggested elsewhere 191.
ings. Third International Conference on Tall Buildings, (6) The factor a in Fig. 3 is equal to I.0 for completely
Hong Kong and Guangxhou, IO-15 December 1984. symmetrical structures and loading conditions and
5. Lehigh University Lecture Notes (1965) Plastic design reduces to less than 1.0 in accordance with eqn (A.2)
of multi-storey frames. Fritz Engineering Laboratory for any deviation from full symmetry:
Report. No. 273.20. Lehiah Universitv. Bethlehem. PA
(1965): - -’ 0.4
6. J. H. Davison and P. F. Adams, Stability of braced and a= (A.21
I - 0.q700 - 1,,,/700)3
unbraced frames. J. Srruc~ Div., ASCE ST2, 319-334
(1974). The term A,,, is representative of the non-
7. F. Cheong-Siat-Moy, Multi-storey frame design using symmetrical conditions and takes the form of a
storey stiffness concept. J. Struct. Div., ASCE 103; slenderness ratio which becomes greater than zero
1197-1212 (1976). for unsymmetrical loading conditions or geometry. It
8. H. E. Schol& A new multi-curve interaction method for is evaluated from eqn (A.1) using elastic second-
the plastic analysis and design of sway frames. Third order moments and axial forces from loading corre-
International Colloquium-Stability of Metal Struc- sponding to 0.4~~. In addition, / is set equal to
tures, Toronto, 1983. 250 N/mm2 and y/r qua1 to 1.0. fhe smallest ratio
9. H. E. Scholx, Simplified Interaction Method for Sway ~I~,., found for the structure is significant. Further
Frames. J. Srrucl. Engng, ASCE 110,992-1007 (1984). simplifications are suggested in [9].

You might also like