Advanced Structural Optimization Using Genetic Algorithm and Neural Network

The goal of this work is to develop a software package that seamlessly integrates the processes of computer-aided design, computer-aided analysis, and structural optimization. A final implementation of this software will provide a user the capability of finding a minimum weight structure from a given preliminary design, constraints, and objectives. This effort will add significant capability to the current state-of-the-art. Improvements include (1) a genetic algorithm for finding near global-optimum solution and gradient-based method for further refinement and sensitivity analysis; (2) optimization performed directly on the CAD model rather than on a FEM; (3) simultaneous size, shape, and topology optimization; (4) modular development for easy compatibility with existing software; (5) adaptive meshing algorithm to recycle the existing FEMs; (6) variational technique to minimize the number of finite element runs; (7) distributed processing of GA-based search; (8) parametric meshing to minimize user intervention in finite element analysis; (9) platform-independence and use of TCP/IP as the communication protocols for networked computers; and (10) use of APIs and scripting interface for providing higher access to user and other software tools. A successful implementation of these technologies will make this software a very powerful CAD-based optimization tool with efficient computation and requiring minimum user intervention.