Concurrent optimization of parameter and tolerance design based on the two-stage Bayesian sampling method

Researchers usually expect to reduce costs while improving product robustness in product quality design. The concurrent optimization of parameter and tolerance design assumes that the output response follows a normal distribution. However, non-normal responses are also common in product quality design. As for the concurrent optimization of parameter and tolerance design with a non-normal response, a novel total cost function based on the two-stage Bayesian sampling method is proposed in this paper. First, the hierarchical Bayesian model constructs the functional relationship between output response, input factors, and tolerance variables. Secondly, a two-stage Bayesian sampling method is used to obtain the simulated values of the output responses. The simulated response values are used to build the rejection cost and quality loss functions. Then, the genetic algorithm is used to optimize the constructed total cost model, including the tolerance cost, rejection cost, and quality loss. Finally, the effectiveness of the proposed method is demonstrated by two examples. The research results show that the proposed method in this paper can effectively improve product quality and reduce manufacturing costs when considering the uncertainty of model parameters and the variation of the output response.