Вы находитесь на странице: 1из 1

OPTIMAL DESIGN OF STUCTURAL ELEMENTS 1.

1 INTRODUCTION Engineers have become more aware of the use of optimal design, rather than merel y safe design, as a result of the ever-increasing emphasis on cost effectiveness . The structural engineer is confronted with various types and complexity of str uctures, both determinate and indeterminate. The usual method of designing a str ucture consists of a trial-and-error process in which a trial design is assumed and the structure is then analyzed and the results are compared with allowable limits of, perhaps, stress and deformation. The design is revised on the basis o f the results of that analysis. This cycle is repeated until an acceptable desig n is reached, the criterion being based upon man hours spent, money limitations, or complexity. Hope- fully, the final design is near the optimal design. The computer provides the means of obtaining the optimization of problems which formerly were too complicated and/or lengthy to compute by hand. The progress in operations research techniques has given further impetus to design optimization . Mathematical programming treats the problem of maximizing or minimizing a functi on of one or more decision variables that are subject to specified constraints. The optimal solution of this function, or objective function, is the basis for making engineering or managerial decisions. If a policy is defined as a set of d ecision variables, then an optimal policy is one which consists of the optimal s et of decision variables. A critical step in any optimization is that of abstractly transforming the physi cal environment into a mathematical model. The process entails the identificatio n of the decision variables, the determination of the limiting conditions, or co nstraints on the decision variables, and the statement of the desired objective of the problem. There are several mathematical programming techniques available to optimize stru ctural design problems. Three programming methods are Linear Programming, Dynam ic Programming, and Geometric Programming. Dynamic Programming is restricted to problems which can be formulated as recursi ve relationships. Geometric Programming It is a systematic method for formulating o ptimization problems which tend to appear in engineering design. It has the adva ntage of being directly applicable to problems comprised of linear and/or nonlin ear objective and constraint functions. The optimal value of geometric programs is determined by the solution of a system of linear equations. These are easily formulated from the exponents of the design variables. When the system of linear equations forms a square matrix, Geometric Programming is spectacular in the ea se with which it obtains a solution even though the design problem may possess h ighly nonlinear functions. A better method for optimizing nonlinear functions is Linear Programming , in wh ich, minimizes or maximizes linear objective functions subject to linear constra int functions. The simplex algorithm efficiently obtains an optimal solution pro vided the problem contains only linear functions. This method has the disadvanta ge that non- linear functions must be approximated by linear functions prior to the use of the simplex algorithm. But, it is quiet easy and provides nearly exac t result. In most structural design problems involve highly nonlinear objective function a nd constraint relations. These structural design problems can only be solved ite ratively. Nevertheless, the category of LP is of interest to us because of sever al reasons. First of all, many nonlinear constrained problems can be approximate d by linear ones which can be solved efficiently by using standard LP algorithms . Using such approximations opens up a possibility for solving NLP problems. Tha t is, almost all NLP problems can be solved as a sequence of repetitive approxim ate LP problems which converge to the exact solution of the original NLP problem provided that the procedure is repeated enough number of times. This powerful p rocedure is called sequential linear programming (SLP).

Вам также может понравиться