Documente Academic
Documente Profesional
Documente Cultură
COLLAR
Bellows is the flexible element of an expansion joint w
used to absorb dimensional changes, such as those caused nt
tc
design for a particular application involves a compromise Figure 1 Structure of unreinforced U-shaped bellows.
of conflicting requirements[1,2]. Optimization techniques Bellows circumferential membrane stress due to
are widely used to achieve design goals of the complex pressure S 2 is:
design problems with high quality and efficiency[3]. PDm K r q
Bellows may be either U-shaped or -shaped (toroidal) S2 = (1)
2 Ac
in cross section. The U-shaped bellows is superior for great
deflection but has a lower pressure capacity for the same where, P is pressure; Dm is the mean diameter of bellows
material thickness. Conversely, the -shaped bellows is convolutions, and Dm =Db + w + nt ; n is number of bellows
limited to small deflection but has a higher pressure material plies of nominal thickness, t ; K r is circumferential
capacity. The use of external reinforcement of the U-shaped
stress factor; q is convolution pitch; Ac is cross sectional
can provide a combination of great deflection and high
internal pressure capacity, and the pressure capacity can metal area of one convolution.
also be increased by the use of multi-ply construction or by Bellows meridional membrane stress due to pressure
increasing the material thickness of the bellows. The U- S3 is:
shaped bellows is mostly manufactured and used in
*
Correspondence author
Pw where, S y is yield strength at room temperature of the
S3 = (2)
2ntp actual bellows material after completion of bellows forming
where, w is convolution height; tp is bellows material and any applicable heat treatment; is inplane instability
stress interaction factor.
thickness for one ply, corrected for thinning during forming, Bellows theoretical axial elastic spring rate per
and tp =t Db Dm . convolution fiu is:
Bellows meridional bending stress due to pressure S 4 1.7 Dm Eb tp 3 n
is: f iu = (9)
w3Cf
2
P w 3 Optimization model description
S4 = Cp (3)
2n tp
Modeling is of prime importance for optimization of
where, Cp is factor used to relate U-shaped bellows bellows design. The mathematical model usually consists
convolution segment behavior to a simple strip beam. of variables, constraints, and objectives.
Bellows meridional membrane stress due to deflection For a design, all variables form a design domain,
S5 is: denoted by a vector X = [x1 , x2 , " , xn ]T . n is number of
Eb t p 2 e domain dimensions, and X can be considered as a point
S5 = (4)
2 w3 Cf in the n-dimensional design domain. Variables are often
where, Eb is modulus of elasticity at design temperature; e is
subjected to some constraints, denoted as gi (X ) 0 . i is
number of the constraints, and i = 1, 2, " , nc . Objectives
total equivalent axial movement per convolution; Cf is
factor used to relate U-shaped bellows convolution segment can be expressed by the objective function f (X ) based on
behavior to a simple strip beam. certain criteria. Then, the optimal design is a group of
Bellows meridional bending stress due to deflection variables having some values that make f (X ) minimum.
S6 is: 3.1 Variables
5 Eb tp e
S6 = (5) The variables of an optimization model must be the
3w2 Cd independent variables and be determined through
where, Cd is factor used to relate U-shaped bellows optimization. Number of variables need to be minimized on
convolution segment behavior to a simple strip beam. the premise of good optimization performance for purpose
Fatigue life N c , number of cycles to failure, is: of the decrease of model complexity.
a
The structure parameters of an unreinforced U-shaped
c bellows are as follows: mean diameter of bellows
Nc = (6)
St b convolutions Dm , inside diameter of bellows convolutions
where, a , b , and c are material and manufacturing Db , convolution height w , convolution pitch, mean radius
constants; S t is pressure, and S t =0.7 ( S3 + S 4 ) + ( S5 + S6 ) . of bellows convolutions rm , bellows nominal material
Limiting internal design pressure based on column thickness of one ply t , number of bellows material plies n ,
instability for single bellows (both ends rigidly supported) and number of convolutions in one bellows N , etc.
Psc is: Since bellows are compatible with pipeline, duct or
0.34 C f iu vessel linked, inside diameter of bellows convolutions Db is
Psc = (7) always defined according to the diameter of pipeline, duct
N 2q
or vessel. There is determined relationship between Db and
where, C is column instability pressure reduction factor
Dm , and Dm =Db + w + nt . Thus, Db and Dm are not
based on initial angular rotation; fiu is bellows theoretical
independent variables. In most cases, convolution pitch
initial axial elastic spring rate per convolution; N is number q and mean radius of bellows convolutions rm are preset by
of convolutions in one bellows.
manufacturer for simplicity of the molds for bellows
Limiting design pressure based on inplane instability
production. Therefore, the independent variables are
and local plasticity at temperature below the creep
convolution height w , bellows nominal material thickness
range Psi is:
of one ply t , number of bellows material plies n , and
1.3 Ac S y number of convolutions in one bellows N , then the variable
Psi = (8)
K r Dm q vector is:
X = [x1 , x2 , x3 , x4 ]T = [ w, t , n, N ]T (10) where, P is design pressure.
3.2 Constraints 3.3 Objective function
The value of optimization objective depends on Bellows functions as the flexible element of an
variables, and there are some constraints acting on these expansion joint. Total equivalent axial movement per
variables. They can be either boundary constraint or convolution e , an index of flexibility evaluation for a
behavior constraint. bellows, can be selected as an objective for the optimization,
and the bigger flexibility, the better. From Eqs. (4) and (5),
3.2.1 Boundary constraint there is:
The constraint used to define the scope of variable ( S 5 +S 6 ) w 2
values or the relations between or among variables. e= (25)
The ratio w q has much influence on the performance Eb tp 5 ( 3Cd ) + tp ( 2wCf )
of a bellows, and it should be confined in the range of 0.6 to Bellows material is high quality austenitic stainless steel
1.6. Then: with strict requirements. On conditions of performance
g1 = 0.6 w q (11) warranty of bellows, decrease of the material use can result
in good benefit for the bellows production. Then, the total
g 2 = w q 1.6 (12)
mass of a bellows m can be another objective for the
There are some defined specifications for bellows optimization, and it is expressed as:
nominal material thickness of one ply t , and they can be
m=2 NntDm ( w 2rm + rm ) +2 ntL ( Db + nt ) (26)
any value of the following: 0.1, 0.2, 0.3, 0.4, 0.5, 0.8, 1.0,
1.2, 1.5, 2.0, 2.5, and 3.0mm. Then: where, L is bellows tangent length including collar,
g3 = 0.1 t (13) L = Lt + Lc .
g4 = t 3 (14) By means of evaluation function method, objective
function can be expressed as:
Number of bellows material plies n should be confined
in the range of 1 to 5, and number of convolutions in one m1
min f (X ) = 2 (27)
bellows N should be confined in the range of 1 to 15. Then: e
g5 = 1 n (15) where, 1 and 2 are weight factors of mass and flexibility,
g6 = n 5 (16) respectively, and set 1 = 1 , 2 = 1.1 according the
g7 = 1 N (17) importance of objectives.
g8 = N 15 (18) 4 Solving strategies
3.2.2 Behavior constraint Three are 4 independent variables and 14 constraints for
The constraint derived from requirements of design or the optimization problem. Variable of convolution height
performance of a bellows and expressed as function of w is a continuous variable, and variables of bellows
variables. nominal material thickness of one ply t , number of bellows
The stresses should be evaluated for pressure capacity material plies n , and number of convolutions in one
as follows: bellows N are discrete variables. It is a nonlinear
g9 = S2 Cwb Sab (19) optimization problem with mixed discrete and continuous
g10 = S3 + S 4 Cm Sab (below the creep range) (20) variables, and some strategies are employed to solve the
problem[5,6].
g11 = S3 + S 4 1.25 Sab (in the creep range) (21)
4.1 Strategies
where, Cwb is longitudinal weld joint efficiency factor;
Sab is allowable material stress at design temperature; Cm is 1) Initial discrete complex: For the n-dimensional
domain, the initial discrete point X (0) has boundary
material strength factor at temperature below the creep
conditions as follows:
range.
Fatigue life should be evaluated as: xi min xi(0) xi max (28)
g12 = N c N ab nf (22) where, i is the i th component of X (0) , and i =1, 2
", n ;
where, N ab is allowable number of cycles to failure; nf is xi min and xi max are the floor bound and ceiling bound of xi(0) ,
safety factor of fatigue, and nf =10 . respectively.
Instabilities should be evaluated as: Vertices of the initial discrete complex are defined as:
g13 = P Psc (23)
g14 = P Psi (24)
xi(1) = xi( 0 ) RE 0 (33)
(j +1) If stop criterion cannot be met after search of k 1
xi = xi
(0)
(i j ) times, move all vertices except X ( L ) to X ( L ) some distance,
(j +1)
x j = x j min (29) often 1 3 of the distance between each vertex and X ( L ) .
(n + j +1)
xi = xi( 0 ) ( i j ) 5) Verification: The vertex X ( L ) that meets stop criterion
x(n + j +1) = x is an optimal vertex in the domain of neighborhood of X ( L ) ,
j j max which cannot guarantee it is the optimal vertex in the whole
totally, there are k vertices defined, and k = 2n + 1 . domain. Therefore, let X ( L ) be another initial discrete
2) One dimensional discrete search: Suppose X ( H ) is point X (0) . Repeat the optimization process mentioned
a vertex having the maximum value f (X ) among all the above, and get a new vertex X (*) that meets stop criterion
initial discrete complex vertices, and X ( C ) is the geometric until X (*) = X ( L ) . Then, the final optimal vertex X (*) can be
1 k (l ) obtained.
(C )
center of all vertices except X ( H ) , and xi = Xi ,
k 1 l =1 4.2 Program design
l = 1, 2, " , k ; l H . The search direction S is from X ( H ) According to strategies presented, a program is
to X , and S i = x x .
(C ) (C ) (H ) developed for solving the optimization model. The
i i
flowchart of the program is shown in figure 2.
Then, the new vertex after discrete search X ( R ) is:
START
xi(R ) = xi(H ) + Si (i = 1, 2, " , n)
(R ) (30)
xi = xi (i = 1, 2, " , p; p n)
(R )
INITIALIZATION
Optimization method was employed in the design of an [6] Sun Jingmin. Optimal Design of Machine, the 4th
unreinforced U-shaped metal bellows. A 4-variable 14- Edition. Beijing: China Machine Press. 2010. 1-250.