A novel approach for measuring the maximum process-loss information in multiple production line conditions

作者：Wu Chien Wei^*; Shu Ming Hung

来源：International Journal of Production Research, 2012, 50(14): 3809-3820.

DOI：10.1080/00207543.2011.588626

• coverage rate
• generalised confidence intervals
• generalised pivotal quantities
• quadratic loss function

The process-loss index L-e, the expected value of the ratio of the quadratic loss function to the square value of half specification width, proposed by Johnson (1992, The relationship of C-pm to squared error loss. Journal of Quality Technology, 24 (4), 211-215), has been widely used in a variety of industries to provide a numerical loss measure for assessing the performance of their production processes. However, the sampling distribution of its uniformly minimum variance unbiased estimator (UMVUE), obtained from traditional approaches involving unknown parameters, is able neither to form classical confidence intervals (CCIs) nor to provide justifiable process-loss information in practice. To tackle this disadvantage, in this paper a novel approach known as generalised confidence intervals (GCIs) is adopted. Instead of Monte Carlo simulations that were popularly utilised in implementing the GCIs method for assessing production process performance, we theoretically derive analytical forms of upper confidence bounds (L-e-GUCB) for L-e and program them to provide the maximum process-loss information for the manufacturing processes. Two common manufacturing scenarios are presented in order to work out: (1) whether the underlying production process loss is capable (or whether products received from one supplier are acceptable); and (2) whether the maximum process-loss information existing in multiple production line conditions is acceptable (or whether products received from several suppliers are acceptable). The applicability of the results is demonstrated by two examples.

• 出版日期2012