A generalized approach to determine the optimal manufacturing target for arbitrary loss functions and process models

The deviation of a quality characteristic from its design target is typically modeled using mathematically well-tractable models, such as Taguchi’s quadratic loss function. By incorporating a process distribution model, the optimal manufacturing target can be determined analytically—for a subset of loss functions and process models. In this article, the manufacturing target is instead interpreted as the result of a combination between two signals, the loss function, and the process model. This allows to approximate the optimal manufacturing target for arbitrary loss functions and process models by using signal processing techniques. Numerical results of the proposed method are compared to analytical solutions obtained from literature, demonstrating accurate approximations. The generalizability of the approach to determine the optimal manufacturing target under the presence of multiple quality characteristics is discussed and demonstrated. As an application that can benefit from the model’s flexibility, an application example from the metrology field is discussed.