Multi-disciplinary design optimisation of mechanical parts with freeform surfaces: Case study encompassing manufacturing criteria

Abstract

Designing parts with freeform surfaces, as typically applied in dies and moulds, is currently dealt with through designer experience, if design intent is to be maintained and if, at the same time, the manufacturing process is to be facilitated. This work puts forward a solution to these issues which is based on computational intelligence. A library of freeform surface morphological features is defined using parametric wireframe models that include constraints. The part is constructed using wireframe features from this library and these are subsequently converted to solid models. The effect of changes of feature parameter values is linked to part functional characteristics in standard design environment. Regarding the effect on manufacturing process characteristics, various models may be employed. As an example, a fuzzy system that decides tool diameter and the necessity of a semi-finishing operation is employed in this work. Artificial neural networks are trained with a number of workpiece variations corresponding to different feature parameter values and the pertinent outputs from functional and manufacturing assessment. Next, a standard genetic algorithm is set up to find the best values of the feature parameters based on both functionality and manufacturing criteria with suitable weighing. The evaluation function of the genetic algorithm employs the artificial neural networks constructed as metamodels. The methodology is demonstrated through an illustrative case study, but may encompass further Design-for-X disciplines.