A method is developed whereby the optimal allotment of manufacturing tolerances to the dimensions of mechanical de signs can be made. In this class of problems, the manufacturing cost is minimized, whilst the design is constrained to meet specified performance requirements and production yields. The technical performance is specified in terms of constraint functions, and the production yield in terms of the percentage adherence within all the constraint functions. The optimization problem is specified in stochastic terms, and truncated Monte-Carlo methods are used to evaluate the performance, and thus the production yield. The total manufacturing cost is written in terms of known, deterministic tolerance cost functions. It is shown that classical gradient methods cannot be used int he solution of the optimization problem. To rectify this,, use is made of Response Surface Methods to approximate the Yield function. Results compare favorably with published results.
展开▼