Probabilistic machine learning based soft-sensors for product quality prediction in batch processes

Mowbray, Max; Kay, Harry; Kay, Sam; Caetano, P. Castro; Hicks, Aaron; Mendoza, Ćesar; Lane, Amanda; Martin, Philip A.; Zhang, Dongda

doi:10.1016/j.chemolab.2022.104616

Cited by 21 publications

(11 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Finally, to highlight the performance of the current soft-sensor, a soft-sensor designed in our previous work was used as a benchmark . In our previous work, MPLS was used as a linear dimensionality reduction technique feeding into either an HNN, BNN, or GP.…”

Section: Resultsmentioning

confidence: 99%

“…In conclusion, autoencoders are a viable dimensionality reduction method and can be used in conjunction with machine learning regression models to build a robust soft sensor able to predict the outcomes of industrial processes such as batch quality. In our previous work, 33 a PLS-based dimensionality reduction model was investigated but the robustness of the model was limited, presenting flaws in the sensor's responsiveness and generalization capabilities. This is mainly because the investigated product is a highly viscous, non-Newtonian fluid; thus, using an autoencoder proved useful because the viscosity and the process variables are nonlinearly related.…”

Section: Discussionmentioning

confidence: 99%

“…Finally, to highlight the performance of the current softsensor, a soft-sensor designed in our previous work was used as a benchmark. 33 In our previous work, MPLS was used as a linear dimensionality reduction technique feeding into either an HNN, BNN, or GP. The MPLS-GP results are used to benchmark with the results obtained from this autoencoder study as this soft sensor model performed the best with regard to the quantitative metrics.…”

Section: Soft Sensor Comparison and Selectionmentioning

confidence: 99%

See 2 more Smart Citations

Integrating Autoencoder and Heteroscedastic Noise Neural Networks for the Batch Process Soft-Sensor Design

Kay

Mowbray

et al. 2022

Ind. Eng. Chem. Res.

Self Cite

View full text Add to dashboard Cite

Viscosity represents a key product quality indicator but has been difficult to measure in-process in real-time. This is particularly true if the process involves complex mixing phenomena operated at dynamic conditions. To address this challenge, in this study, we developed an innovative soft sensor by integrating advanced artificial neural networks. The soft sensor first employs a deep learning autoencoder to extract information-rich process features by compressing high-dimensional industrial data and then adopts a heteroscedastic noise neural network to simultaneously predict product viscosity and associated uncertainty. To evaluate its performance, predictions of product viscosity were made for a number of industrial batches operated over different seasons. Furthermore, probabilistic machine learning techniques, including the Gaussian process and the Bayesian neural network, were selected to benchmark against the heteroscedastic noise neural network. Through comparison, it is found that the proposed soft-sensor has both high accuracy and high reliability, indicating its potential for process monitoring and quality control.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Soft Sensor Comparison and Selectionmentioning

confidence: 99%

See 1 more Smart Citation

Integrating Autoencoder and Heteroscedastic Noise Neural Networks for the Batch Process Soft-Sensor Design

Kay

Mowbray

et al. 2022

Ind. Eng. Chem. Res.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Another consideration to account for is the performance of the symbolic regression models in comparison to more traditional, noninterpretable, black-box models. In previous work, 26 MPLS and deep learning models including the GP and an HNN were developed on the process described in case study 1. Compared to the MPLS method (testing MAPE greater than 15%), the deep learning methods boast testing MAPEs of 11.4 and 13.7%, respectively.…”

Section: Discussionmentioning

confidence: 99%

Constructing a Symbolic Regression-Based Interpretable Soft Sensor for Industrial Data Analytics and Product Quality Control

Kay,

Mowbray

et al. 2024

Ind. Eng. Chem. Res.

Self Cite

View full text Add to dashboard Cite

Deriving physical models for key performance indicators (KPIs) has been a challenge for industries developing accurate control and optimization schemes. As a result, data-driven models have seen a rise in application within recent literature; however, commonly used "black-box" data-driven models suffer from a lack of interpretability, limiting their uptake within industrial settings. To address this challenge, we developed an interpretable soft sensor by integrating symbolic regression among dimensionality reduction, statistical feature engineering, and latent variable modeling. Through this approach, symbolic regression can effectively extract important statistical relations between the KPI and key process variables. To investigate their performance and compare against existing machine learning models, two industrial case studies were presented focusing on formulation product quality control. It is concluded that symbolic regression-generated soft sensors are of high accuracy, and their expressions can provide meaningful physical insights into the underlying processes, highlighting their great potential for future applications.

show abstract

“…However, in industrial processes, substantial time delays and expensive inspection costs cause great difficulties in monitoring some key variables and performance elements that reflect the operating conditions of the equipment or the processing status of the product instantaneously due to the production process and extreme manufacturing environment. [2,3] There are often essential characteristics such as strong nonlinearity, dynamic characteristics with slow time variation, and uncertainty; multilevel and strong coupling relationships prevail in many covariates. [4] For traditional manual inspection methods, cost-effective online measurements are lacking to produce high quality products, and therefore traditional mechanistic models are difficult to accurately describe real industrial processes.…”

Section: Introductionmentioning

confidence: 99%

Weighted target feature regression neural networks based soft sensing for industrial process

Guo,

Wang,

2023

Can J Chem Eng

View full text Add to dashboard Cite

There are often nonlinear and time‐varying characteristics in industrial processes. These characteristics cause difficulty in measuring product quality online. To address these issues, a weighted target feature regression neural network (WTFAER) was proposed for soft sensor modelling in this paper. The Pearson correlation coefficient was calculated to assign corresponding weights to process variables and design a weighted objective function. A target feature regression network (TFAER) was constructed using target correlation autoencoder with fully connected layer. After that, the weighted reconstructed information was applied to the TFAER model to extract deep quality‐related features and realize feature reuse. A deep network was formed by layered stacking to fully exploit the deep features for quality prediction. To make the proposed method domain adaptive, a maximum mean squared deviation (MMD) based regularization term was introduced in the loss function. Through the simulation experiments of debutanizer column and industrial polyethylene process, and compared with stacked autoencoder (SAE), variable‐wise weighted stacked autoencoder (VW‐SAE) and stacked target‐related autoencoder (STAE) methods, the effectiveness and generalization performance of the proposed modelling method were verified.

show abstract

Probabilistic machine learning based soft-sensors for product quality prediction in batch processes

Cited by 21 publications

References 35 publications

Integrating Autoencoder and Heteroscedastic Noise Neural Networks for the Batch Process Soft-Sensor Design

Integrating Autoencoder and Heteroscedastic Noise Neural Networks for the Batch Process Soft-Sensor Design

Constructing a Symbolic Regression-Based Interpretable Soft Sensor for Industrial Data Analytics and Product Quality Control

Weighted target feature regression neural networks based soft sensing for industrial process

Contact Info

Product

Resources

About