An optimal tolerance allocation model for assemblies with consideration of manufacturing cost, quality loss and reliability index

Purpose - The purpose of this paper is to provide an optimal tolerance allocation model for assemblies with consideration of the manufacturing cost, the quality loss, the design reliability index with various distributions to enhance existing models. Results of two case studies are presented. Design/methodology/approach - The paper develops a model with consideration of the manufacturing cost, the Taguchi's asymmetric quadratic quality loss and the design reliability index for the optimal tolerance allocation of assemblies. The dimensional variables in normal distributions are initially used as testing and compared with the data from the prior researches. Then, the dimensional variables in lognormal distributions with the mean shift and the correlation are applied and investigated. Findings - The results obtained based on a lognormal distribution and a normal distribution of the dimension are similar, but the tolerance with a lognormal distribution is little smaller than that with a normal distribution. The result of the reliability with the lognormal distribution obtained by the Monte-Carlo is higher than that with a normal distribution. This paper shows that effects of the mean shift, the correlation coefficient and the replacement cost on the cost are significant and designers should pay attention to them during the tolerance optimization. The optimum tolerances of components of a compressor are recommended. Research limitations/implications - The model is limited to the dimensions of components with the normal distribution and lognormal distributions. The implication should be enhanced with more data of dimension distributions and cost of assembly components. Practical implications - Two case studies are presented. One is an assembly of two pieces and another is a compressor with many components. Originality/value - This model provides an optimal tolerance allocation method for assemblies with the lowest manufacturing cost, the minimum quality loss, and the required reliability index for the normal distribution and lognormal distribution.