A multi-model fusion soft sensor modelling method and its application in rotary kiln calcination zone temperature prediction

Tian, Zhongda; Li, Shujiang; Wang, Yanhong; Wang, Xiangdong

doi:10.1177/0142331215573099

Cited by 32 publications

(4 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In such situations, it is possible to approximate the nonlinear behavior of process by designing a so-called multimodel inferential sensor (MIS) [9]. The MISs found their use mostly in the same fields as standard inferential sensors, e.g., in the petrochemical industry [9], in manufacturing [10], and in the process industry [11].…”

Section: Introductionmentioning

confidence: 99%

Design of Multi-model Linear Inferential Sensors with SVM-based Switching Logic

Mojto¹,

Fikar²,

Paulen³

2022

Preprint

View full text Add to dashboard Cite

We study the problem of data-based design of multi-model linear inferential (soft) sensors. The multi-model linear inferential sensors promise increased prediction accuracy yet simplicity of the model structure and training. The standard approach to the multi-model inferential sensor design consists in three separate steps: 1) data labeling (establishing training subsets for individual models), 2) data classification (creating a switching logic for the models), and 3) training of individual models. There are two main issues with this concept: a) as steps 2) & 3) are separate, discontinuities can occur when switching between the models; b) as steps 1) & 3) are separate, data labelling disregards the quality of the resulting model. Our contribution aims at both the mentioned problems, where, for the problem a), we introduce a novel SVM-based model training coupled with switching logic identification and, for the problem b), we propose a direct optimization of data labelling. We illustrate the proposed methodology and its benefits on an example from the chemical engineering domain.

show abstract

Section: Introductionmentioning

confidence: 99%

Design of Multi-model Linear Inferential Sensors with SVM-based Switching Logic

Mojto¹,

Fikar²,

Paulen³

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Therefore, there is still a need to explore new soft measurement methods to improve the predictive performance of the model under complex operating conditions. The modeling method of multi-model fusion provides an effective idea to solve the above problems and has attracted extensive attention in soft measurement research (Zhongda et al, 2016). Zhang et al (2020) predicted air quality (PM2.5) through a weighted fusion approach.…”

Section: Introductionmentioning

confidence: 99%

A multi-model fusion soft measurement method for cement clinker f-CaO content based on K-means ++ and EMD-MKRVM

Zhang

Wang

et al. 2022

Transactions of the Institute of Measurement and Control

View full text Add to dashboard Cite

The content of free calcium oxide (f-CaO) in cement clinker is a key indicator for testing the quality of cement clinker. To address the problem that the content of f-CaO cannot be detected online, a multi-model fusion soft measurement method based on K-means++ clustering, empirical modal decomposition combined with multi-kernel relevance vector machines (EMD-MKRVM) is proposed to predict f-CaO content under different operating conditions. First, time-series analysis and matching of input variables with f-CaO content were performed, based on which a combination of empirical modal decomposition (EMD) and sample entropy (SE) denoising method was used to filter out high- frequency noise from the original data and extract effective signal information for reconstruction. Second, the K-means++ algorithm was used to cluster the processed training sample data, and multi-kernel relevance vector machine (MKRVM) sub-models were established by training the sample data of each sub-class and then the affiliation between the test samples and each sub-class was calculated as the weights of the sub-model output values, and the final model prediction output was obtained by multi-model fusion. Finally, the real data from cement plants were used for validation. The results show that compared with the single MKRVM model, multi–relevance vector machine (RVM) model, multi–support vector machine (SVM) model, and multi-MKRVM model using only EMD denoising method, the mean absolute error (MAE) of the multi-MKRVM model proposed in this paper was reduced by 42%, 7%, 14%, and 35%; root mean square error (RMSE) is reduced by 28%, 10%, 12%, and 21%; squares due to error (SSE) is reduced by 51%, 24%, 27%, and 41%; Theils inequality coefficient (TIC) is reduced by 27%, 17%, 21%, and 19%; [Formula: see text] is improved by 64%, 18%, 39%, and 91%; and Index of agreement (IA) is improved by 28%, 9%, 22%, and 13%. The multi-MKRVM model proposed in this paper has higher accuracy, better generalization ability and stability, and provides an effective method for f-CaO content prediction under complex multiple working conditions.

show abstract

“…On the other hand, SPE contributes to the lack-of-fit of the approximate model to the one it was originally brought from. As a matter of fact, diverse extensions have been introduced to PCA in order to overcome some of its limitations such as nonlinear PCA (NLPCA) (Jia et al, 1998), considering the system’s dynamics using dynamic PCA (DPCA) (Ku et al, 1995), Kernel PCA (KPCA) (Lee et al, 2004; Zhongda et al, 2016), and multiscale PCA (MSPCA) (Mirin, 2013) that combines the PCA with wavelet analysis. However, the control charts still need more thoughtful attention from researchers in order to find new alternatives that further enhance the process monitoring efficiency.…”

Section: Introductionmentioning

confidence: 99%

Online monitoring scheme using principal component analysis through Kullback-Leibler divergence analysis technique for fault detection

Bounoua

Benkara

Kouadri

et al. 2019

Transactions of the Institute of Measurement and Control

View full text Add to dashboard Cite

Principal component analysis (PCA) is a common tool in the literature and widely used for process monitoring and fault detection. Traditional PCA is associated with the two well-known control charts, the Hotelling’s T2 and the squared prediction error (SPE), as monitoring statistics. This paper develops the use of new measures based on a distribution dissimilarity technique named Kullback-Leibler divergence (KLD) through PCA by measuring the difference between online estimated and offline reference density functions. For processes with PCA scores following a multivariate Gaussian distribution, KLD is computed on both principal and residual subspaces defined by PCA in a moving window to extract the local disparity information. The potentials of the proposed algorithm are afterwards demonstrated through an application on two well-known processes in chemical industries; the Tennessee Eastman process as a reference benchmark and three tank system as an experimental validation. The monitoring performance was compared to recent results from other multivariate statistical process monitoring (MSPM) techniques. The proposed method showed superior robustness and effectiveness recording the lowest average missed detection rate and false alarm rates in process fault detection.

show abstract

A multi-model fusion soft sensor modelling method and its application in rotary kiln calcination zone temperature prediction

Cited by 32 publications

References 22 publications

Design of Multi-model Linear Inferential Sensors with SVM-based Switching Logic

Design of Multi-model Linear Inferential Sensors with SVM-based Switching Logic

A multi-model fusion soft measurement method for cement clinker f-CaO content based on K-means ++ and EMD-MKRVM

Online monitoring scheme using principal component analysis through Kullback-Leibler divergence analysis technique for fault detection

Contact Info

Product

Resources

About