Mixture of partial least squares experts and application in prediction settings with multiple operating modes

Abstract

This paper addresses the problem of online quality prediction in processes with multiple operating modes. The paper proposes a new method called mixture of partial least squares regression (Mix-PLS), where the solution of the mixture of experts regression is performed using the partial least squares (PLS) algorithm. The PLS is used to tune the model experts and the gate parameters. The solution of Mix-PLS is achieved using the expectation-maximization (EM) algorithm, and at each iteration of EM algorithm the number of latent variables of the PLS for the gate and experts are determined using the Bayesian information criterion. The proposed method, shows to be less prone to overfitting with respect to the number of mixture models, when compared to the standard mixture of linear regression experts (MLRE). The Mix-PLS was successfully applied on three real prediction problems. The results were compared with five other regression algorithms. In all the experiments, the proposed method always exhibits the best prediction performance.

An on-line weighted ensemble of regressor models to handle concept drifts

“…RPLS is a widely used algorithm in on-line process modeling to adapt to the process changes. Its popularity is motivated by its reduced computational time and computer memory requirements (Qin, 1998); and robustness under collinearity, measurement error and high dimensionality of input space, which are common characteristics in most industrial data sets (Souza and Araújo, 2014). AddExp.C and Learn þ þ .…”

A Robust Infinite Gaussian Mixture Model and Its Application in Fault Detection of Nonlinear Multimode Processes

“…As a consequence, greater demands are being placed on the process monitoring, especially fault detection and diagnosis, which plays an indispensable role in production safety, process stability and quality products (Qin, 2012;Ding, 2014;Shang et al, 2014). In the last two decades, data-driven process monitoring has been widely applied in the process industry, including chemicals (Yu, 2012;Luo et al, 2015), polymerization (Souza and Araújo, 2014), oil re nery (Chen et al, 2004) and wind turbine (Yin et al, 2014). e characteristics of modern industrial plants, however, may change frequently from one operating condition to another due to the changes of product speci cations, operation grade, set-point, some external restrictions, etc.…”

Student’s-t Mixture Regression-Based Robust Soft Sensor Development for Multimode Industrial Processes

“…Due to multiple product-grade requirements, feedstock changes, load variations, seasonal operations, etc., most industrial processes work with multiple operation modes [ 20 ]. The multimode characteristics result in process variables that are no longer Gaussian, and the functional relationship between primary and secondary variables being strongly non-linear [ 2 ], which increases the difficulty in developing high-accuracy soft sensor models.…”