Development of Non-Dominated Sorting Genetic Quantum Algorithm for Structural and Tool Shape Optimization


Structural optimization is important for product design in order to improve the performance and decrease the production time as well as total cost of a product. In the past, engineers used costly and time consuming trial-and-error methods to find the optimum for their structural design problems. Today, with steadily increasing computational power, computers are increasingly used to design, manufacture, and evaluate the performance of a product without having to physically building it to reduce the cost of development. The finite element analysis (FEA) used in such developmental process are expensive and hence, often hinders the direct use of conventional optimization techniques such as the mathematical programming and most heuristic approaches. When there are multiple objectives and constraints, the difficulty compounds. Efficient optimization of structural design problems with computationally intensive FEA processes has been studied intensely by many researchers for the past decade. Only recently, researchers have started to focus on optimization of multiple objective problems using these expensive FEA. This thesis aims to develop a multiobjective optimization algorithm which can handle multiple design variable, objectives, and constraints of expensive FEA models. A new algorithm, called non-dominated sorting genetic quantum algorithm (NSGQA), is developed by integrating a newly developed evolutionary algorithm, genetic quantum algorithm (GQA), and a multiobjective sorting mechanism. The performance of NSGQA is intensively evaluated using inexpensive optimization problems and compared with state-of-the-art meta-heuristics multi-objective optimization algorithms. Subsequently, NSGQA is applied to a real structural optimization problem with 47 trusses and many constraints. From the tests and application to structural optimization, NSGQA demonstrates promising performance. The NSGQA is integrated with a FEA-based process model for composite processing and is applied to optimize tool shape to reduce the process-induced warpage. This preliminary benchmarking was done to pave the way for further research on multi-objective tool shape optimization for composite processing.