Polymer extrusion process monitoring using nonlinear dynamic model-based PCA

Abstract

Polymer extrusion is one of the final forming stages in the production of many polymeric products in a variety of applications. It is also an intermediate processing step in injection moulded, blown film, thermo-formed, and blow moulded products. However, polymer extrusion is a complex process which is difficult to set up, monitor, and control. As a consequence, high levels of off- specification products and long down-times are the problems facing the plastics industry. This paper proposes a new method for fault detection of the polymer extrusion processes, where the nonlinear finite impulse response (NFIR) model and principal component analysis (PCA) are integrated to form a nonlinear dynamic model-based PCA monitoring scheme. Here the NFIR model is used to capture the nonlinearity and dynamics of the extrusion process. The residuals resulting from the difference between the model predicted outputs and process outputs are then analyzed by PCA to detect process faults. The experimental results confirm the efficacy of the proposed model-based PCA approach for fault detection of polymer extrusion processes.