Optimisation of a parametric CAD geometry by means of a genetic algorithm

This thesis has the aim of optimising a three-dimensional geometry whose dimensions have been previously defined by parameters. Such a definition allows the geometry to change its shape by altering the value of a parameter without the need of directly modifying a geometrical feature (extrusion, protrusion, etc.) itself. In this way, a single value may affect more than one characteristic of the design. Then, the task is to find the best combination of parameters which improve the performance of a design. In here, the optimisation is performed by the computer by means of a genetic algorithm to change the values of such parameters. There is extensive work in the subject of optimising the performance of products by means of genetic algorithms (GA). In this regard, much work has been done by using equations which model their behaviour. The widespread use of computerised technology (for design and analysis) eventually allowed researchers to combine GA with computer-aided design (CAD) and engineering (CAE). This resulted in two major branches of GA-based optimisation of geometries: the agglomeration of finite elements and the combination of geometrical features. In here, a GA is used to iteratively modify (in CAD) and evaluate (in finite-element analysis) a parametric geometry. This geometry corresponds to a mechanism recently developed by the author. The device has both dynamic and static characteristics, however, the evaluations (analyses) performed on the design are purely static. An interface between a numerical analysis software (MATLAB1) and a CAD-CAE suite (CATIA1) is developed. The GA is run within the numerical analysis software and is interfaced to the CAD-CAE suite by means of its application programming interface (API). The results show the feasibility of improving a three-dimensional, feature-based and parametric CAD by means of the variations created by a GA.