Optimal Tolerance Allocation of Automotive Pneumatic Control Valves Based on Product and Process Simulations
This paper discusses a computational method for optimally allocating dimensional tolerances for an automotive pneumatic control valve. Due to the large production volume, costly tight tolerances should be allocated only to the dimensions that have high influence to the quality. Given a parametric geometry of a valve, the problem is posed as a multi-objective optimization with respect to product quality and production cost. The product quality is defined as 1) the deviation from the nominal valve design in the linearity of valve stroke and fluidic force, and 2) the difference in fluidic force with and without cavitation. These quality measures are estimated by using Monte Carlo simulation on a Radial-Basis Function Network (RBFN) trained with computational fluid dynamics (CFD) simulation of the valve operation. The production cost is estimated by the tolerance-cost relationship obtained from the discrete event simulations of valve production process. A multi-objective genetic algorithm is utilized to generate Pareto optimal tolerance allocations with respect to these objectives, and alternative tolerance allocations are proposed considering the trade-offs among multiple objectives.