Process Feature Change Recognition Based on Model Performance Monitoring and Adaptive Model Correction for the Gold Cyanidation Leaching Process

Abstract

The gold cyanidation leaching process (GCLP) is the central unit operation in hydrometallurgy, and satisfactory gold recovery is highly significant in practice. However, GCLP faces the challenge of an irregular slow time-varying feature (STVF), which seriously affects gold recovery, and blind treatment for STVF also has drawbacks, which results in the need for the recognition of STVF for purposeful, rather than blind, treatment. Meanwhile, it also faces the problem of change of working condition (COWC) due to the variability of mineral resources. Both STVF and COWC may cause degradation of the soft-measuring model, which presents the need for model correction. Therefore, a coping strategy is proposed to solve these existing problems. First, an improved model-based principal component analysis monitoring is proposed to detect model mismatch and monitor the change of process feature. Next, a support vector machine-based process feature change recognition method is presented to recognize change type, which not only provides guidance in treating STVF but also makes it possible to implement targeted model correction for STVF and COWC. Finally, an adaptive model correction strategy that combines case-based correction and just-intime learning-based correction is proposed. The simulation studies have verified the validity of the proposed coping strategy. INDEX TERMS Gold cyanidation leaching process, model-based principal component analysis, process feature change recognition, adaptive model correction.

A Data-Driven Iterative Optimization Compensation Method Based on PJIT-PLS for Gold Cyanidation Leaching Process

“…To cope with this problem, data-driven modeling methods are extensively used in industrial plants [4]. They are many kinds of data-driven modeling approaches like multivariate statistical methods such as principle component regression (PCR) [5], independent component regression (ICR) [6], [7], partial least squares (PLS) [8], and nonlinear methods such as kernel PLS [9], kernel PCR [10] and so on. Meanwhile, the machine learning methods have been widely applied in process modeling such as Gaussian process regression (GPR) [11], [12], support vector regression (SVR) [13], artificial neural networks (ANN) [14], [15], their application study has been analyzed in the literature [8], [16].…”

“…They are many kinds of data-driven modeling approaches like multivariate statistical methods such as principle component regression (PCR) [5], independent component regression (ICR) [6], [7], partial least squares (PLS) [8], and nonlinear methods such as kernel PLS [9], kernel PCR [10] and so on. Meanwhile, the machine learning methods have been widely applied in process modeling such as Gaussian process regression (GPR) [11], [12], support vector regression (SVR) [13], artificial neural networks (ANN) [14], [15], their application study has been analyzed in the literature [8], [16]. To solve the kinetic reaction rate expressions are difficult to be obtained accurately in the actual GCLP, Zhang J et al proposed a serial hybrid modeling method that combined the first-principle model (the mass conservation equations in steady-state mechanistic model) and data-driven model (two BP ANN models) [17].…”