Design automation of geometric profile tolerance specification and placement of fluid power cartridge valves

This dissertation provides solutions for two problems in design automation. The two solutions are developed as independent essays. The first problem is on generating non-uniform geometric profile tolerance boundaries for implicit/explicit and freeform features. Such boundaries are required when singularities in shapes occur. We show that non-uniform B-spline curves and surfaces, a class of computeraided geometric design elements, are useful to specify non-uniform profile tolerance boundaries. The specification, visualization, and mathematical methods to present deviations are developed in this thesis. For the evaluation of actual measured points of the features defined by such tolerance boundaries, an evaluation procedure called the binary adaptive subdivision method is developed. This solver is applicable for features without multiple stationary points and self-intersections. Finally in this essay, to establish the realization of this technique, a Java ™applet is developed to demonstrate our solution to the unambiguous specification. The second problem deals with the optimal placement of fluid power cartridge valves in a manifold block. The placement is a quadratic assignment problem. To consider optimal placement, an optimal routing scheme to interconnect the ports of the cartridge valves internally is required. The routing is another NP hard problem. In this essay, we develop an evolutionary algorithm to determine the optimal placement considering optimal routing methods. The results from the algorithm provide guidelines for the placement problem to the designers. From the extensive experiments, we extract descriptive statistical data that provides guidelines to determine the conditions for the convergence of the evolutionary algorithms to global optimal solutions.