Vous êtes sur la page 1sur 4

Strength of Materials, Vol. 27, No.

7, 1995

INDUSTRIAL SECTION

OPTIMUM DESIGN OF CYLINDRICAL STORAGE TANKS WITH


ALLOWANCE FOR THE UNIFORMITY OF THE RELIABILTY

OF THE CORRODING SHELL

Yu. M. Pochtman and M. M. Fridman UDC 624.134

An examination is made of the optimization of thin-walled structures, such as vertical cylindrical tanks, that
are subject to pitting corrosion. The utility function o f the structure is taken as the optimization criterion. The
choice of the optimum plating over the height o f the tank is determined on the basis of the condition of its
uniform reliability.

The surfaces of tanks that must withstand fluid pressure and the effects of aggressive media in service are often subject
to pitting corrosion. In this case, it is important to evaluate the strength and reliability of the structure, since the growth of even
one pit into a hoIe can make the tank unfit for further use and cause accidents. As was noted in [1], storage tanks are among
the largest structures made and rank near the top in terms of the metal consamed in their construction. This lends importance
to problems relating to improving the strength of tanks, reducing their metal content, and making their construction easier.
A probabilistic model of the fracture of thin-walled structures due to pitting corrosion was presented in [2], where it
was assumed that the initial size distribution of the pits the surface of the structure and the time until the appearance of the first
hole are random variables.
In this investigation, we propose taking a comprehensive approach to optimizing the design of thin-walled structures
from the viewpoint of their utility. As a criterion of optimality of a cylindrical shell (the shell of the tank), we take a function
expressing the mean expected utility [3]

U(T, X) = B(T, X) - H,(X) - L(T, X), (1)

which includes the average income B(T, X) expected from the use of the structure over the course of its rated service life T,
with allowance for possible failure at the moment of time trail < T. Equation (1) also includes the initial cost Hi(X) and the
loss from failure of the structure L(T, X).
The goal of the optimization is to f'md the vector of the optimum parameters of the structure Xopt that maximizes utility
function (1) with the following restriction on reliability:
[U(T, X) = B(T, 20 -- HtCX') - L(T, X)] "* max;
(2)
[ e(~ -_.P.,
where P(T) is the reliability function; P, is the prescribed reliability.
The vector of the variable parameters is taken in the following form:

X={x,,x.AT=ln, T},

Dnepropetrovsk University, Dnepropetrovsk, Ukraine. Translated from Problemy Prochnosti, No. 7, pp. 90-94, July,
1995. Original article submitted February 18, 1994.

422 0039-2~16/95/2707-0422512.50 9 Plenum Publishing Corporation


where n is the number of segments of variable thickness (plates) comprising the tank; T is the rated service life of the tank.
The tank is subjected to internal hydrostatic pressure by the aggressive medium it contains. We assume that the medium is in
the nature of petroleum products and has the density O. The main dimensions of the tank, radius R and height H, are known.
We determine the quantities entering into (2). We find the income and the loss from failure of the structure in accordance with
[31:
r
B(T, X) = of l~(t)P/ "ail(t)dt, (3)

where the income B~ = 7b (1 - e - rt) ", b is the annual income in the absence of failure; r = ln(1 + r'); r' is the percent

of the capital; Pfail is the probability density function for failure of the structure during its service life,

P/aa P) "aa(t)[1 - / ' ( t ) ] ' .


= (4)

We similarly establish the mean loss (or the loss from failure) adjusted for the present time:
T
L(T, X) = f L~ (5)
o

where the loss L~ = Lte-rt; Lt the is the sum of the losses expected before use of the structure.
We determine the initial cost of the structure as [1]

H,(J0 = Cf, (6)

where Cf is the factory production cost,

C f = Cd. m -4- C m - F C s ;

Cd. m is the cost of the direct materials,

Ca. m = 1 , 0 7 C,~ikrpiGi "t" 1,5 ;


i

Crpi is the wholesale cost of the flat-rolled product for the i-th part of the structure; G i is the weight of the i-th part,
/l
at = Z ~ g h ~ r . , ~ = y~ ~ ;
/l i=1

C m are the manufacturing costs,

Cra = C pc + C as'Wd + C c - - 0,62 GV~-n'n;

Cpc, C aswd, and C e are the costs of processing, assembly and welding, and coiling; C s is the cost of the structure (free on
board),

Cs = 1,14[(C.d.m+ Cm)'l,005 + 2,661;

C e is the cost of erecting the tank,

Ce =-0,641 GC-~'n.

Let us now proceed to determination of the shell reliability function entering into (2). By the reliability of the shell
structure, we mean the probability of the following random event: pitting does not exceed an acceptable level for a certain

423
period of service 0 _< t < T. In the present case, this level is the thickness of one plate. We take the corrosion equation in
the form

d/~ = a + flt:ri, (7)


dt

where l i is the running depth of the pits; t~ and/3 are constant coefficients; a i is the effective stress in the i-th plate.
Equation (7) has the following solution under the condition t = 0, l i = /0i:

ti = loi + (a + fltri)t = 1,i + bit.

If we regard the initial pit depth/0i as a random variable with a normal probability density fimction
r' . 2

q(zo,) = - )

with allowance for the fact that b i is constant, then Ii is a fixed random quantity. In this case, the probability that a pit will not
exceed the thickness of the i-th plate of the shell h i (taken as a constant) is determined as follows [4]:

P~-
]
where

hi - bit - l~i
a i =
%,
= l
~ f e -=2/z is the Laplace integral.

If we take N as the total number of pits on the entire surface of the tank and assign thicknesses over the height of the
shell so as to make the reliability of each segment constant (of equal reliability), then the reliability of the tank as a whole is
described by the expression

e = (eT) ".

Knowing the acceptable level of reliability P* and the total number of pits, we can easily calculate the required
thicknesses h i while ensuring that the segments are equally reliable. In this case, Pi* = (p,)l/N and a = (h i - - bit - - l~i)/ato i
= const, where the value of a is found from the condition (p,)l/N = ~(a).
Taking /~ = c~h/, at~ = cz~/, we obtain

hi - c,hi -- bit
a = c2h~ ; bi = a + / k r i. (8)

H R
Since the effective stress in the i-th plate tri = p n " i 9 ~ (where Hi/n is the distance from the surface of the cylinder

to the bottom of the i-th segment), we can use (8) to determine the sought quantities hi as follows:

h / = k + v r ~ + p; i = 1 . . . . . n,

where

at
k=2(1 --
c, --
cza); p = n ( 1 ffHipRtc,
-- -- c2a~ ) "

424
After determining the reliability function and inserting the value of Pfaii found from Eq. (4) into Eqs. (3) and (5), we
write new expressions for the income and the loss from failure of the structure

B(T, 20 = r @(a)] e -~ at12[0,5- @(a - at)] }; (9)

(10)
z r, 20 = [0,s or
- -

where

rh~(l - c,) . rc2hi


b, = bi , a, = bi

(It follows from Eq. (8) that a 1 = const and b 1 = const for any value of i).
After calculating the running values of B(T, X), Hi(X), and L(T, X), we proceed directly to optimization of the tank.
Complex nonlinear multi-extremum problems such as (2) are best solved by one of the efficient algorithms empIoying the
method of random search [5].
As a numerical illustration, we will examine the optimization of a storage tank with the following initial data: R =
2 m; H = 4 m; N = 48; c~ = 0.06 cm/g; p = 0.8-103 kg/m3; c I = 0.2; c 2 = 0.01; P. = 0.99; b = 104; L T = 104; r' =
10%; B = 0.085 cm/ton.yr.
The optimum tank parameters obtained from the calculation were: time of service T = 28 yrs; number of segments
of variable thickness n = 10; other indices: h = 2.19; B(T, X) = 969.42; H(X) = 217.24; L(T, X) = 7.11; U(T, X) =
745.07. As can be seen, the change in thickness h over the height established on the basis of the principle of equal reliability
of the shell of the tank is almost negligible. In conclusion, we note that the use of this principle made it possible to perform
the optimization without the restriction on reliability in the form of (2).

REFERENCES

. Ya. M. Likhtarnikov, Variant Design and Optimization of Steel Structures [in Russian], Stroiizdat, Moscow (1979).
2. R. A. Arutyunyan, "Probabilistic model of fracture due to pitting corrosion," Probl. Prochn., No. 12, 106-108 (1989).
3. G. Augusti, A. Baratta, and F. Kashiati, Probability Methods in Structural Design, Stroiizdat, Moscow (1988).
4. K. Kaput and L. Lamberson, Reliability and the Design of Systems [Russian translation], Mir, Moscow (1980).
5. I. B. Gurvich, V. G. Zakharchenko, and Yu. M. Pochtman, "Randomized algorithm for solving nonlinear
programming problems," Izv. Akad. Nauk SSSR, Tekh. Kibern., No. 5, 30-34 (1979).

425

Vous aimerez peut-être aussi