Statistical Analysis and Optimal Allocation of Geometric Tolerances
Abstract This paper describes a procedure for the statistical analysis and optimization of geometric tolerances. The proposed procedure assumes that a manufactured surface is represented by a set of points, which are assumed to be random variables having a multinormal distribution. Sets of surface points are generated from the multinormal distribution, and the minimum deviation zone for the geometric deviations in each set is compared with the specified tolerances. A parametric surface is interpolated to the generated points representing the manufactured surface. Genetic algorithms and a Monte Carlo simulation routine which incorporates variance reduction techniques are used to evaluate the geometric deviations of the machined surface. A second routine, based on genetic algorithms, is used to allocate the tolerance values which keeps the part’s probability of rejection within a desired value. An example for simulating a cylindrical feature is presented and the results obtained from the algorithms using the proposed variance reduction techniques are compared with those obtained using simple Monte Carlo simulation. In addition the specified tolerance values are reallocated to achieve a desired probability of rejection.