Knowledge engineering applications in automated structural design

The present work focusses on the development of AI based design systems. The work proposes a general methodology for integrating AI methods with algorithmic procedures for the development of preliminary configurations of load bearing truss and frame structures. The proposed framework is an adaptation of a recently proposed model for generic problem solving systems. The methodology is characterized by the use of algorithmic processing to enumerate the design space, and utilizing this information in a design process dominated by heuristics.;The design process deals with the synthesis of preliminary structural configurations for a given problem specification. The approach is based upon the systematic decomposition of the design space, and repeated application of design heuristics to incrementally generate an optimal structural topology to account for the applied loads. The principal shortcomings of a heuristic decomposition approach are underscored, and more rational quasi-procedural methods are proposed.;The research sheds light on the two distinct methods of decomposition for large structural design problems. The first method which has its basis in global sensitivity analysis, proposes an approach to decompose the design space into subproblems for which solutions can be readily obtained, and to integrate these partial solutions into the system solution by consideration of first order sensitivities of subproblem interactions. The second approach employs the use of dynamic programming, wherein the size of the solution space is managed by a decomposition of the design problem into many stages. The use of this strategy generates a number of possible configurations, which are then subjected to a genetic algorithm based stochastic search to yield a near-optimal structure. Several design examples are presented to demonstrate the validity of the design methodology and the decomposition methods.