Unbalanced Tolerance Design and Manufacturing Setting with Asymmetrical Linear Loss Function

Taguchi has developed many loss functions, and many of the situations are approximated by the quadratic function. Using a quadratic loss function when the actual loss function is non-quadratic may cause the incorrect input parameter levels to be chosen. In certain situations, a linear loss function is more appropriate in industrial applications. If tolerance design is unbalanced, setting the process mean at the nominal will not minimise the expected quality loss. Thus, two specific models of linear loss function are discussed for unbalanced tolerance design. One is an asymmetrical linear loss function and the other is a truncated asymmetrical linear loss function. The optimal process means for these two loss functions are also compared with each other. The results show that the process mean should be shifted a little from the nominal such that the expected quality loss is minimised.