Knowledge-Based Discrete Optimalization of Truss Structures

Abstract

The knowledge-based approach to discrete optimization is presented in the paper. The minimization problem characterized by linear objective function and arbitrary constraints is considered when design variables have to be chosen from a set of discrete values avail­able. The controlled enumeration algorithm according to the non-decreasing values of the objective function is supplied with an additional module manipulating the information represented symbolically. This module contains the domain-oriented knowledge expressed in the form of heuristic rules and is used to eliminate the useless constraints verification for the propositions considered to be "non-promising". The approach coupling the symbolic and numerical computations enables a significant reduction in the number of design variables variants that must be checked for feasibility in order to find the optimum. The numerical examples for the minimum weight optimization of a cantilever truss structure and the corresponding simple heuristic rules are presented.