Locally weighted partial least square (LW-PLS) model has been commonly used to develop adaptive soft sensors and process monitoring for numerous industries which include pharmaceutical, petrochemical, semiconductor, wastewater treatment system and biochemical. The advantages of LW-PLS model are its ability to deal with a large number of input variables, collinearity among the variables and outliers. Nevertheless, since most industrial processes are highly nonlinear, a traditional LW-PLS which is based on a linear model becomes incapable of handling nonlinear processes. Hence, an improved LW-PLS model is required to enhance the adaptive soft sensors in dealing with data nonlinearity. In this work, Kernel function which has nonlinear features was incorporated into LW-PLS model and this proposed model is named locally weighted kernel partial least square (LW-KPLS). Comparisons between LW-PLS and LW-KPLS models in terms of predictive performance and their computational loads were carried out by evaluating both models using data generated from a simulated plant. From the results, it is apparent that in terms of predictive performance LW-KPLS is superior compared to LW-PLS. However, it is found that computational load of LW-KPLS is higher than LW-PLS. After adapting ensemble method with LW-KPLS, computational loads of both models were found to be comparable. These indicate that LW-KPLS performs better than LW-PLS in nonlinear process applications. In addition, evaluation on localization parameter in both LW-PLS and LW-KPLS is also carried out.

Just-in-time (JIT) adaptive soft sensors have been widely used in chemical processes because they can deal with slow-varying processes, abrupt process changes, and outliers. However, these traditional JIT algorithms including locally weighted partial least square (LW-PLS) have limitations in dealing with non-Gaussian distributed and nonlinear data. To address these issues, a modified LW-PLS-based JIT algorithm, namely, ensemble locally weighted independent component kernel partial least square (E-LW-IC-KPLS) is proposed. Its predictive performances were tested using the data generated from a numerical example and two simulated plants. Then, the results were compared to the ones resulted from LW-PLS, locally weighted kernel partial least square (LW-KPLS), and locally weighted independent component kernel partial least square (LW-IC-KPLS) algorithms. From these comparative studies, it is evident that E-LW-IC-KPLS is superior compared to its traditional counterparts concerning predictive performances. The predictive errors for E-LW-IC-KPLS are lower by 8−94% than those of LW-PLS, LW-KPLS, and LW-IC-KPLS.

No abstract

The textile bleaching process that involves hot hydrogen peroxide (H 2 O 2 ) solution is commonly practised in cotton fabric manufacture. The purpose of the bleaching process is to remove color from the cotton, achieving a permanent white before proceeding to dyeing or shape matching. Normally, the visual ratings of whiteness on the cotton are measured based on whiteness index (WI). However, it is found that there is little research on chemical predictive modelling of the cotton fabric's WI compared to experimental study. Analytics using predictive modelling can forecast the outcomes, leading to better-informed cotton quality assurance and control decisions. Up to date, there is limited study applying least square support vector regression (LSSVR) model in the textile domain. Hence, the present study aims to develop a multi-output LSSVR (MLSSVR) model using bleaching process variables and results obtained from two different case studies to predict the WI of cotton. The predictive accuracy of the MLSSVR model was measured by root mean square error (RMSE), mean absolute error (MAE), and the coefficient of determination (R 2 ). The obtained results were compared with other regression models including partial least square regression, predictive fuzzy model, locally weighted partial least square regression, and locally weighted kernel partial least square regression. Our findings indicate that the proposed MLSSVR model performed better than other models in predicting the WI as it showed significantly lower values of RMSE and MAE. Furthermore, it provided the highest R 2 values which are up to 0.9999.

No abstract