An overview of strategies for identifying manufacturing process window through design of experiments and machine learning techniques while considering the uncertainty associated with.

Abstract

The industry sector has long been seeking methods to enhance its manufacturing system control, production, and monitoring, while maintaining the quality of its products and reducing costs and time. One method for achieving a more comprehensive understanding of the manufacturing processes is the identification of a corresponding process window (PW). However, there is no globally accepted definition of process window, which is principally used in different manufacturing processes. In some cases, it is combined with operating window or process map concepts. In light of the aforementioned consideration, this article puts forth a definition of process window, drawing upon the various notions and aspects related to the optimal process parameters selection as discussed in the literature. Furthermore, the identification of key controllable process parameters and the criteria to delimit boundaries of desired parts are described as aspects to identify the corresponding process window. Moreover, this paper provides an overview of the techniques used to establish a process window, principally through the application of machine learning techniques (ML) and design of experiments (DOE). Furthermore, the uncertainty intrinsic to the manufacturing process and the methodologies employed to identify the process window are examined. In conclusion, this article emphasises the necessity of transferring these models to new materials or machines, which will require further investigation in future research.