Abstract:To improve the performance of online prediction of existing soft sensor models, we propose a dual updating strategy, i.e., integrating the methods of recursive partial least square (RPLS) model updating and the model output offset updating. In online applications, each update is activated rotationally. In this strategy, a new recursive PLS method is developed and implemented by updating the mean and variance of the training samples using the data acquired from the process, while the offset updating method takes into account both the old overall offset and the new bias between the actual measurement and the model prediction. Since the dual updating strategy takes the advantages of the two updating methods, it is more effective than any individual updating method in adapting process changes. The high performance of the strategy is demonstrated by the application of an industrial purified terephthalic acid (PTA) purification process in which prediction of average crystal particle size was within 2.5% with regard to the relative root mean square error (RMSE). In addition, the dynamic PLS method was found inferior to any of the three methods mentioned above, at least for this particular industrial application. The present dual updating method may also be extended to other industrial applications using process models outside PLS.

PRODUCT DESIGN VIA PLS MODELING: STEPPING OUT OF HISTORICAL DATA INTO UNKNOWN OPERATING SPACE

A Selective Moving Window Partial Least Squares Method and Its Application in Process Modeling

Online dual updating with recursive PLS model and its application in predicting crystal size of purified terephthalic acid (PTA) process

Development of Decay Based PLS Model and Its Economic Run-to-Run Control for Semiconductor Processes

Product Quality Prediction Method for Injection Molding Process Based on Operating Mode Recognition

The Construction and Application of Product-Form Semantic Space Based on PLSA

Controller Design Based on PLS Latent Variables Space

A Model Predictive Control Approach with Relevant Identification in Dynamic PLS Framework

Optimal Weighting Distance-Based Similarity for Locally Weighted PLS Modeling

KPLS model based product quality control for batch processes

Enhanced Batch Process Monitoring and Quality Prediction Using Multi-Phase Dynamic PLS

Stage-Based Process Analysis and Quality Prediction for Batch Processes

A new PLS approach with hybrid internal models

PLS-based Optimal Quality Control Model for TE Process

Fast Locally Weighted PLS Modeling for Large-Scale Industrial Processes

Multivariable Modeling Based on Dynamic Fuzzy PLS and Its Application

Robust Pls Approach For Kpi-Related Prediction And Diagnosis Against Outliers And Missing Data

Nonlinear Multivariate Quality Prediction Based on OSC-SVM-PLS

Statistical Prediction of Product Quality in Batch Processes with Limited Batch-Cycle Data

Nonlinear Pls Improved by Numeric Genetic Algorithm for Qsar Modelling

Quality relevant data-driven modeling and monitoring of multivariate dynamic processes: the dynamic T-PLS approach.