Data Mining and Analytics in the Process Industry: The Role of Machine Learning

Ge, Zhiqiang; Song, Zhihuan; Ding, Steven X.; Huang, Biao

doi:10.1109/access.2017.2756872

Cited by 747 publications

(377 citation statements)

References 292 publications

Supporting

Mentioning

366

Contrasting

Unclassified

Order By: Relevance

“…To illustrate how DL‐HPE performs, five datasets are selected for test from the UCI machine learning library (The details of the datasets are referred to http://archive.ics.uci.edu/ml/datasets.html), whose brief description is given in Table . Firstly, the input attributes of the five datasets are processed using PCA‐MI, then the DL‐HPE model based on the nine regression models in Ge et al, which almost cover the commonly used data‐driven modeling methods is tested. The termination criterion for ensemble pruning is a 30% improvement in the mean absolute error relative to the best base model.…”

Section: Testing Of Uci Benchmark Datasetsmentioning

confidence: 99%

“…In recent years, the data‐based models, which feature little dependence on a priori knowledge and capability of capturing the characteristics of sophisticated process, have gained more and more attention in the prediction of chemical process outputs . These data‐based prediction models are mainly derived from nine base models, that is, multiple linear regression, partial least squares, principal component regression, artificial neural network, TREE, K nearest neighbors, support vector regression (SVR), Gaussian process regression (GPR), and random forest (RF) . According to Kadlec et al and Harrington et al, each type of the base model has its own merits, whereas no one is definitely superior to another and able to exactly capture the global characteristics of practical processes.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Prediction of liquid ammonia yield using a novel deep learning‐based heterogeneous pruning ensemble model

Dai

Yang

Zhang

et al. 2020

Asia-Pacific J Chem Eng

View full text Add to dashboard Cite

Liquid ammonia yield is the main index characterizing the process output of ammonia synthesis, the prediction of which is crucial for process control and optimization. However, the industrial process of ammonia synthesis involves multiple variables and strong nonlinearity, making it difficult to be accurately predicted using conventional mechanism‐driven model, or a certain type of data‐driven model. Therefore, a deep learning‐based heterogeneous pruning ensemble (DL‐HPE) model is proposed to overcome the limitations of conventional models. In this model, a “VDiv” pruning strategy that trades off diversity and accuracy is proposed and successfully applied to select the optimal subset from nine representative base models. Finally, the deep learning is employed to integrate the outputs of the involving models to generate the final prediction. The DL‐HPE method was applied to modeling the UCI standard dataset and the historical dataset collected from an ammonia plant in China. The results show that the method is superior to single model or other ensemble models both in accuracy and in robustness.

show abstract

Section: Testing Of Uci Benchmark Datasetsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Prediction of liquid ammonia yield using a novel deep learning‐based heterogeneous pruning ensemble model

Dai

Yang

Zhang

et al. 2020

Asia-Pacific J Chem Eng

View full text Add to dashboard Cite

show abstract

“…The batch process is an important part of modern industry, and the safety monitoring of the batch process is very important and meaningful . Theoretical research based on data‐driven modeling methods, including soft sensor modeling and process monitoring, has made significant progress and has become increasingly intelligent with industrial processes . For the batch process, multiway principal component analysis (MPCA) is one of the most basic and the most extensively used monitoring methods .…”

Section: Introductionmentioning

confidence: 99%

Just‐in‐time learning–multiple subspace support vector data description used for non‐Gaussian dynamic batch process monitoring

Yan

Jiang

et al. 2019

Journal of Chemometrics

View full text Add to dashboard Cite

Batch process data are time‐varying dynamic and non‐Gaussian distributed. In addition, for multivariate statistical process monitoring, their variability can be overwhelmed when considering local variability behavior. To address the abovementioned issues, an improved batch process monitoring approach is presented that integrates just‐in‐time learning and multiple subspace support vector data description (JITL‐MSSVDD). A new multiple subspace segmentation method is proposed that classifies a contribution array that is calculated on the mixing matrix of independent component analysis (ICA). Offline, the variable subspace segmentation rule can be obtained from the proposed method. The subspace monitoring models can reduce the risk of the variability being overwhelmed. Online, local modeling samples are collected through JITL, which can reduce the impact of the time‐varying dynamic on modeling accuracy. Then, in accordance with the variable subspace segmentation rule, which is obtained offline, MSSVDD models are constructed to solve non‐Gaussian problems. The advantage of the proposed JITL‐MSSVDD is demonstrated through the standard test model fed‐batch penicillin fermentation process.

show abstract

“…Traditional model‐based process monitoring methods have become difficult to adapt given these characteristics. In this circumstance, data‐driven multivariate statistical process monitoring (MSPM) methods have also become popular considering its convenience and relaxing of prior knowledge . Principal component analysis (PCA), independent component analysis (ICA), and partial least squares (PLS) are the basic MSPM methods.…”

Section: Introductionmentioning

confidence: 99%

Quality‐relevant dynamic process monitoring based on mutual information multiblock slow feature analysis

Zheng

Jiang

Yan

2019

Journal of Chemometrics

View full text Add to dashboard Cite

Slow feature analysis (SFA) is an efficient technique in exploring process dynamic information and is suitable for quality‐relevant process monitoring. However, involving quality‐irrelevant variables or features may introduce redundant information and degrade the monitoring performance. A novel multiblock monitoring scheme based on mutual information (MI) and SFA is proposed to detect an efficient quality‐relevant fault for dynamic processes. First, all process variables are divided into two blocks in accordance with their MI values with quality variables. Second, slow features (SFs) of two blocks are extracted via the SFA. The SFs, which are extracted from the quality‐relevant variables, are not all related to the quality variable. Thus, these SFs are further partitioned into two blocks in accordance with their MI values with quality variables. The SFs from three blocks are obtained, and monitoring statistics are constructed. Two simulation studies, including a numerical example and Tennessee Eastman process, demonstrate that the proposed method outperforms conventional methods.

show abstract

Data Mining and Analytics in the Process Industry: The Role of Machine Learning

Cited by 747 publications

References 292 publications

Prediction of liquid ammonia yield using a novel deep learning‐based heterogeneous pruning ensemble model

Prediction of liquid ammonia yield using a novel deep learning‐based heterogeneous pruning ensemble model

Just‐in‐time learning–multiple subspace support vector data description used for non‐Gaussian dynamic batch process monitoring

Quality‐relevant dynamic process monitoring based on mutual information multiblock slow feature analysis

Contact Info

Product

Resources

About