A Comparative Study of Different Kernel Functions Applied to LW-KPLS Model for Nonlinear Processes

Soft sensors are inferential estimators when the employment of hardware is inapplicable, expensive, or difficult in industrial plant processes. Currently, a simple soft sensor, namely locally weighted partial least squares (LW-PLS), which can cope with nonlinearity process, has been developed. However, LW-PLS exhibits disadvantages handling strong nonlinear process data. To address this problem, Kernel functions integrated into to form (LW-KPLS). Notice that minimal study was carried out on impact different kernel have not LW-KPLS, model potential be applied chemical-related Thus, investigates predictive performance LW-KPLS several using three case studies. As results, performances Polynomial better than other functions. The values root-mean-square errors (RMSE) and error approximation (Ea) for training testing dataset by utilizing function lowest their respective studies, 34.60% 95.39% lower RMSEs 68.20% 95.49% smaller Ea values.