关键词: 统计过程调整, 非均衡失效成本, 贝叶斯模型, 预测模型, 态变数, 逼近式搜索算法

Abstract: After deeply studying the problem of Statistical Process Adjustment (SPA), a process adjustment model was set up. Asymmetric off-target cost was used as the performance evaluation index of adjustment methods which was used in tolerance design and parameters design but not in SPA yet. Different Bayesian models were constructed with known and unknown information in the process variance, and then the prediction model of the quality characteristics was constructed. According to the Bayesian models and prediction model, the optimal control targets for each step were defined and the target adjustments were also performed. Finally, by comparing with Grubbs basic rule adjustment approach and Markov Chain Monte Carlo (MCMC) adjustment approach on condition of asymmetric off-target cost, the performance advantages of the proposed approach were confirmed.

Key words: statistical process adjustment, asymmetric offtarget cost, Bayesian model, predictive model, state variable, asymptotic searching algorithm