Optimizing tolerance allocation for normally distributed dimensions including process capability constraints.

摘要

Tolerance allocation, which is a phase of product design, can reduce cost of the product without any investment or improvement in the quality of the product or the process while the product still satisfies all the specified requirements. Optimizing dimensional tolerance allocation was examined for areas of potential improvement. Current research in the field of optimizing tolerance allocation has proposed various models for specific conditions with limitations in application. This research develops a model for optimizing tolerance allocation that can be solved using common off-the-shelf software, and applying design of experiments to analyze the sensitivity of the total cost to cost coefficients and constraints.; The model proposed can be applied to optimizing dimensional tolerance allocation for normal distributions with any combination of three characteristics. They are (1) the process mean either equal or unequal to the nominal size, (2) possible unbalance in the upper and the lower quality loss coefficients, minimum requirements for the process capability indices, specified ranges for semi tolerance zones, and/or specified minimum proportions of conforming units of the product, and (3) ability to select among non-inspection (NI), 100% inspection without reworking (IWR) or 100% inspection with imperfect reworking strategy (IIR).; The proposed model assumes that the process standard deviation for each part has an increasing linear relationship with its tolerance. The minimum total cost is the criterion for the optimization, solved by genetic algorithm in Evolver. The proposed model is subject to constraints associated with (1) the process standard deviation of the gap resulting from the square root of the summation of the parts and the envelope variances, (2) the allowable ranges for the semi-tolerance zones for the parts, (3) the specified minimum requirements for process capability indices, (4) the allowable ranges for process standard deviations of the parts, and (5) the specified minimum proportion of conforming units of the product.; Design of experiments (DOE) is applied to sensitivity analysis to completely finish the process of optimization. This approach can significantly reduce the number of experimental runs while it can analyze the sensitivity of the factors with a specified level of significance.