Product and process improvement using mixture-process variable methods and robust optimization techniques

摘要

Most industrial processes consists of two parts: the product recipe made from raw materials (RMs) and physical process based process variables (PVs). A design of experiment for mixture-process variable (MPV) can be complicated since change of recipe in a chemical process is not easy. This situation leads to split-plot designs. A complete MPV model formed by crossing mixture experiment model with a PV model is fit for the situations. This, along with other variable selection techniques, such as:Best-subset regression,Step-wise regression, and Backward elimination