You are on page 1of 5

004s7949/ss s3.00 + 0.

00
Pngamon Pressplc

TECHNICAL NOTE

A NOTE ON OPTIMUM GEOMETRY OF A PORTAL FRAME

JACEK RETRZAK
Civil Engineering Department, Technical University of Poznan, 60-965 Poznan, Poland

(Received 21 April 1987)

Ahatraet-It is shown that by properly inclining the frame columns one can increase the buckling load
for two-bar and portal frames. This fact is worth considering in frame optimization. The inclusion of the
column inclination parameter into the design variable vector results in a more economic design than that
obtained by optimization of a frame of fixed geometry. Numerical examples that refer to a hinged-based
portal frame are given.

1. INTRODUCI’ION portal frames; (ii) for each frame, an angle P can be found
which maximizes the function @c. In the cases of the
The recent decades have witnessed si~ifi~nt progress in the frames a and b, the ma~mum values of the function differ
development of the theoretical foundations and numericat only slightly from unity; max (J,/&) is 1.001 for the pin-
methods of optimization of structures’ topology, geometry supported bent, and reaches the value of 1.012 for the bent
and the member cross-sectional properties. In structural with clamped supports. By contrast, for both of the portal
design practice, however, optimization is usually confined to frames, max (&/&) differs significantly from unity; for the
searching for the optimum sizes of the member cross- frame c, max A,= 2.14& and for the frame d, max
sections, in spite of the Fact that optimi~tion of topology & = I.43 n;.
or geometry of some structures may be more eKective in Since, for the two-bar frames, a, is rather insensitive to
weight reduction than changes in the member sizes. geometry variation, little benefit can be expected in this case
Most numerical results substantiating the advantages of from geometry optimization. The portal frames c and d are,
including in an optimization process the variation of geom- however, promising in this respect.
etry have been gathered in the field of pin-jointed trusses Confining further attention to frame c, which exhibits the
(see e.g. [l-3]). Less documented are the benefits of opti- strongest sensitivity of A, to 6, we will examine, in the next
mization of geometry of rigidly jointed frames. section, the combined effect of the geometry variation and
The aim of the present note is to show that the frame changes in member sixes on the magnitude of & and the
geometry changes, consisting of in&ring the columns, may feasible design domain.
lead to a remarkabie increase of the frame critical load and
noticeable reduction of the frame volume.
3. OYMMUM PORTAL FRAME
Consider &hehinged-base portal frame shown in Fig. 3(a),
2. FSFRCT OF COLUMN INFLATION with the I-shaped cross-section [Fig. 3(b)]. The section is
ON THE CRITICAL LOAD assumed to be the same for both the columns and beam of
the frame. The only dimension of the section permitted to
Consider the four frames depicted in Fig. I, subject to the vary is its depth, h. The columns of the frame are allowed
one-parameter loading P = PA, where P is a load pattern to be inclined at angle 8, which may assume values between
matrix, and Z a load factor. The load factor may vary in the 40 and 90”.
range (0, ,2,), where & is the value of 1 at which ins~bi~ty For the selected points of the design area defined by the
of the frame occurs. in~u~ties 4OO” Q @ c 90” and 0. I m G h GO.18 m, the val-
The columns of the frame are allowed to deviate from the ues of the critical load factor function, /1,(& h), have been
vertical direction by an angle 6, as shown in Fig. 1. As will calculated and the contour lines of the surface A, construc-
be seen, the column inclinations affect, sometimes ted. The contour lines are shown in Fig. 4. They enable one
si~ifi~ntly, the values of the critical load factor 1,. to quantitatively estimate the advantages of permitting the
In order to examine the effect of L on &, linear buckling geometry of the frame to be varied. For example, it can be
analyses of the four frames were carried out for various seen that the orthogonal frame, subject to loading with
values of L. It was assumed in the analyses that the frames 1 = 0.25, hecomes unstable for h Q 6, where 6 = 0.1623 m.
were prevented from out-of-plane buckling. The calcu- The material volume, p, corresponding to 6 is 0.02968 m’.
lations were performed on a personal computer, using a The frames with inclined columns, subject to the same
standard criterion (based on the value of the stiffness matrix loading, require smafler depth, L, and less material to remain
determinant) for occurrence of bifurcation. stable. For /I equal to 80 and 70”, the minimum allowable
The results obtained are shown in Fig. 2 in the form of web depths, h, are 0.1275 and 0.1127 m, respectively. For
a plot of the four non-dimensional functions L&,, where & B = 62”, which is an angle corresponding to point K on the
denotes the critical load factor for the frame with vertical contour line of A,, minims h is 0.1095 m. The minimum
column(s). (For the frames a, b, c and d, x, is equal to material volumes, V, required for the frames with the
0.15435 MN, 0.29816 MN, 0.0191 I MN and 0.07711 MN, inclined columns are as follows: for p = 80”, V = 0.9156P;
respectively.) for /i = 70”, V = 0.87977; and for B = 62”, V = 0.8721 P,
It can be seen from Fig. 2 that: (i) the critical load factor, which indicate that noticeable material savings accrue from
& is only moderately affected by L for the two-bar frames, inclining the ~01~~s. Similar obsetvations can be made for
and turns out to be quite sensitive to changes in t for the the frames subject to loading of different magnitudes of 1.
683
684 Technical Note

I P,=IX (MN)

A I
I
I L
T 4- T

Frame a Frame c

Frame b Frame d

For aU members :
La3m
A = 1.8 x lC2m2
I * 1.35 x IQ6m4
E=?OGPa

Fig. I. Sampie frames.

Detailed analysis of the cases corresponding to other values


of 1 will follow.
Obviously, whether or not a pair (B, h), pertaining to a
point on the contour line 1, = 1, gives a feasible design, still
depends on the other constraints, of which the constraint
imposed on stresses is most likely to become active. To
examine the role of the stress constraint in forming the
feasible domain, let us adopt the constraint in the form
CT= (jNI/A + [MI/S) 6 R, in which N is the column axial
force, M is the maximum bending moment, S is the section
modulus, and R is the prescribed allowable stress.

Frame c

lb) t

I
-40 -50
I I
-20 -IO
i I
0
I
IO 20
I
x) t
Efd6ggf
I b !
Fig. 2. Diagrams of A/& versus the column inclination
angle, e, for the four sample frames. Fig. 3. Geometry and cross-section of a portal frame.
Technical Note 685

_ii*O.l623 m
tG = Q02968 m3)

0.1 I 0.12 0.13 414 O.I!s 0.16 0.17 0.16


h(m)

Fig. 4. Contour lines of the critical load factor function.

ii=Ql455 m
/
I
I

CXII ai2 o.i3 0. I4 0.15 0. I6


h(m)

90.1773m
9o(b)
1

0.1
h(m)

ii-0.2034m

h(m)

Fig. 5. Design areas for three loading cases: (a) L = 0.2; (b) I, = 0.3; (c) rl = 0.4.
686 Technical Note

R=iOOMFn

(b)

R=iGO MPa
I

R = 150,175, xx) MPo

= 100,125,150,175,200 MPa

(c)

'2:0.9 z IO
> 0

0.0 20

i
90 60 ?D 60 x) 40
j3 (degl

Fig. 6. Diagrams of the weight reduction function, rp, for three loading cases: (a) L = 0.2; (b) I = 0.3; (c)
L = 0.4.

Consider three loading cases that differ in the magnitude assumed as 125 MPa, while the relevant data pertaining to
of the load factor 1. The cases are: i = 0.2,0.3 and 0.4. For other values of R are presented in Fig, 6.
each value of 1, the contour lines of the stress function have For A = 0.2, Fig. S(aj, the optimum-design is represented
been determined and superimposed on the curve A = ACand by ooint K (19 = 62’. h = 0.0976 m’l. The volume of the
on the contour lines of the material volume function (see o&mum frame, V, is b.S789P, where P is the volume of the
Fig. 5). relevant minimum weight orthogonal frame (@= QO”,
Inspection of Fig. 5 enables us to notice that: (i) for a li = 0.1455 m).
given A, the point X that would minimize the frame volume For 1. =0;3, Fig. s(b), the optimum design is also
in the absence of the stress constraint may (for small values snecifled by noint K If3= 62“. li = 0.1201 ml. In this ease
of 0 assumed as R) or may not (for la&‘R) be excluded k = 0.8662p; where VP refers to the orthogonal frame
from the feasible domain; and (ii) for a liven R, the risk that (B = QO”,k = 0,1773 m).
the point K tinds itself outside’the f&ibie area increases For 1 = 0.4, Fig. 5(c), point K is non-feasible. The
with the increase of 1. optimum design is represented by point L @I= 78*,
To quantitatively evaluate the reduction in the frame h = 0.155 m) and V, = 0.8932 p, f’ being the volume of the
volume, resulting from optimization of both the angle of the frame of /I = 90” and li = 0.2034 m.
column inclination and depth of the section, a vafue for R To enable the evaluation of weight reduction for other
has to be adopted. The results quoted below refer to R values of R be made, the diagrams shown in Fig. 6 were
Technical Note 681

constructed. For each of the considered load factors, the increased by properly inclining the frame columns. In
function V/V, corresponding to various R, is plotted. The optimum designing, this modification of the originally or-
diagrams confirm the earlier remarks on the role played by thogonal frames can be used as a means of extending the
the stress constraint. Additionally, it can be seen that for all feasible design domain, because the stability requirement
the Is the function V/V exhibits the steepest descent for usually constitutes an essential constraint on a frame design.
80” g b < 90”, which, simultaneously, is the range of the The percentage of material savings, accruing from the frame
angle /I for which the function is not affected by the stress geometry changes, varies with the intensity of the frame load
constraint. Thus, even a small deviation of the column lines and values of the allowable stress. For the sample hinged-
from the vertical direction (-c not exceeding lo”) may lead base frame, the material savings resulting from the inclina-
to a noticeable reduction of the frame volume with respect tion of the columns by IO” were about S-9%; for the larger
to the volume of the orthogonal frame (in the case consid- inclination angles, the savings exceeded 10%.
ered, 10” deviation gives 8-9% volume reduction). For
--L > lo” (i.e. /I c 80”) the volume reduction rate is smaller.
The material savings function, cp = (V - V)lOO/V, reaches REFERENCES
its maximum at different points, depending on R and ,I.
1. M. P. Saka, Shape optimization of trusses. J. sfruct.
Div., ASCE RK(STS), 1155-1174 (1980).
4. CONCLUSIONS 2. W. R. Spillers, Iterative design for optimal geometry. J.
SITI(CLDiv., AXE lOl(SIY), 14351442 (1975).
The numerical results presented in this note indicate that 3. P. Pedersen, Optimal joint positions for space trusses. J.
the critical load for portal frames can be substantially strucf. Div., ASCE 9!?(ST12), 24592476 (1973).

You might also like