Determining the optimum manufacturing target using the inverted normal loss function

摘要

Spiring and Yeung (1998) introduced the concept of inverting a normal probability density function to provide a more realistic loss function. Numerous loss functions have been proposed that use various distributions to depict loss. In this research, the concept of the inverted normal loss function is furthered to accurately model losses in a product engineering context. Expected loss can be computed by numerical integration, the integral of the product of the loss function and the probability density function. If the actual process parameter distribution and a realistic loss function are given, expected loss can be determined numerically. A case study involving a shaft bearing for a microcontroller product is given to illustrate the inverted loss function. Two experiments were performed to determine the process variables having the strongest effect on the product's yield and the ideal process target and the specification limits.