Learning-Based Predictions and Soft Sensing for Process Industries
Theory, Methodology and Applications
- 1st Edition - July 1, 2026
- Latest edition
- Authors: Hamid Reza Karimi, Yongxiang Lei
- Language: English
Learning-Based Predictions and Soft Sensing for Process Industries: Theory, Methodology and Applications covers prediction and soft sensing in industrial processes that are subjec… Read more
Learning-Based Predictions and Soft Sensing for Process Industries: Theory, Methodology and Applications covers prediction and soft sensing in industrial processes that are subject to specific challenges with AI-empowered learning algorithms. With the aid of a data-driven modeling strategy, the book explores the problems of industrial prediction and soft sensing and formulates a series of learning-based theory, methodologies, and applications. The book introduces the basics of prediction and soft sensing backgrounds, including different categories of prediction theory. Secondly, covers the foundations of machine learning methodologies, including supervised learning prediction, semi-supervised, and self-supervised prediction. Finally, the book examines novel learning-based models/architectures.
- Covers the benefits and an explanation of recent developments in prediction and soft sensing systems
- Unifies existing and emerging concepts surrounding advanced prediction models/architectures
- Provides a series of the latest results in, including, but not limited to, supervised learning, semi-supervised learning, self-supervised learning, probabilistic learning
Researchers and practitioners working on complex systems, mining industry, process industrial systems, applied mathematics, artificial intelligence and mechatronics
Section 1: Theory
Introduction of Soft Sensing
Overview of Single-Variable Soft Sensing
Overview of Multiple-Variable Soft Sensing
Preview of This Book
Abbreviations and Notations
1. Introduction of Prediction
1.1 Introduction of Data-Driven Prediction
1.2 Overview of Single-Variable Prediction
1.3 Overview of Multiple-Variable Prediction
1.3.1 Review of Time Series Prediction
1.3.2 Review of Recursive Prediction
1.3.3 Review of Multiple-Horizon Prediction
1.3.4 Review of Linear-Model-Based Prediction
2. Theoretical Foundations of Paste-Filling System
2.1 Theory of Paste-Filling System
2.1.1 Typical Functions
2.1.2 Typical Definitions
2.1.3 Basic Properties
2.2 Prediction Requirements and Challenges
2.2.1 Description of Prediction Tasks
2.2.2 Description of Soft Sensing Tasks
3. Foundation of Aluminium Electrolysis System
3.1 Theory of Aluminium Electrolysis System
3.1.1 Typical Functions
3.1.2 Typical Definitions
3.1.3 Basic Properties
3.2 Prediction Requirements and Challenges
3.2.1 Description of Prediction Tasks
3.2.2 Description of Soft Sensing Tasks
Section 2: Methodology
4. Machine Learning Basics for Prediction & Soft Sensing
4.1 ARIMA
4.2 LSTM
4.3 Conv
4.3.1 SVM
4.3.2 Transformer
4.3.3 ELM
4.4 Probabilistic Bayesian Model
4.5 Conclusion
Section 3: Application
5. A Novel Supervised Soft Sensor Framework Based on Convolutional Laplacian Extreme Learning Machine: CNN-LapsELM
5.1 Introduction
5.2 Problem Statement
5.3 Main Framework
5.4 Numerical Examples
5.5 Conclusion
6. A Novel Semi-Supervised Soft Sensor Framework Based on Stacked Auto-Encoder Wavelet Extreme Learning Machine: SAE-WELM
6.1 Introduction
6.2 Problem Statement
6.3 Main Framework
6.4 Numerical Example
6.5 Conclusion
7. A Novel Soft Sensor Based on Laplacian Hessian Semi-Supervised Hierarchical Extreme Learning Machine: LHSS-HELM
7.1 Introduction
7.2 Problem Statement
7.3 Main Results
7.4 Numerical Examples
7.5 Conclusion
8. A Self-Supervised Prediction Framework Based on Deep Long Short-Time Memory for Aluminum Electrolysis: SSDLSTM
8.1 Introduction
8.2 Problem Statement
8.3 Main Results
8.4 Simulation Results
8.5 Conclusion
9. A Self-Supervised Prediction Framework Based on Convolutional Deep Long Short-Time Memory for Aluminum Temperature Application: CNN-SSDLSTM
9.1 Introduction
9.2 Problem Statement
9.3 Main Results
9.4 Numerical Examples
9.5 Conclusion
10. A Novel Probabilistic Prediction Framework Based on Bayesian Machine Learning: BLSTM
10.1 Introduction
10.2 Problem Statement
10.3 Main Results
10.4 Numerical Example
10.5 Conclusion
11. Direct Data-Driven Quantile Regressor Forecaster for Underflow Concentration Soft Sensing: DDQRF
11.1 Introduction
11.2 Problem Statement
11.3 Main Results
11.4 Numerical Example
11.5 Conclusion
12. A Novel Key-Quality Prediction Framework for Industrial Deep Cone Thickener: DualLSTM
12.1 Introduction
12.2 Problem Statement
12.3 Main Results
12.4 Numerical Example
12.5 Conclusion
13. A Deeply-Efficient Long Short-Time Memory Framework for Underflow Concentration Prediction: DE-LSTM
13.1 Introduction
13.2 Problem Statement
13.3 Main Results
13.4 Numerical Example
13.5 Conclusion
14. An Ensemble Prediction Method for Probabilistic Forecasting of Aluminium Electrolysis Process
14.1 Introduction
14.2 Problem Statement
14.3 Main Results
14.4 Numerical Example
14.5 Conclusion
Introduction of Soft Sensing
Overview of Single-Variable Soft Sensing
Overview of Multiple-Variable Soft Sensing
Preview of This Book
Abbreviations and Notations
1. Introduction of Prediction
1.1 Introduction of Data-Driven Prediction
1.2 Overview of Single-Variable Prediction
1.3 Overview of Multiple-Variable Prediction
1.3.1 Review of Time Series Prediction
1.3.2 Review of Recursive Prediction
1.3.3 Review of Multiple-Horizon Prediction
1.3.4 Review of Linear-Model-Based Prediction
2. Theoretical Foundations of Paste-Filling System
2.1 Theory of Paste-Filling System
2.1.1 Typical Functions
2.1.2 Typical Definitions
2.1.3 Basic Properties
2.2 Prediction Requirements and Challenges
2.2.1 Description of Prediction Tasks
2.2.2 Description of Soft Sensing Tasks
3. Foundation of Aluminium Electrolysis System
3.1 Theory of Aluminium Electrolysis System
3.1.1 Typical Functions
3.1.2 Typical Definitions
3.1.3 Basic Properties
3.2 Prediction Requirements and Challenges
3.2.1 Description of Prediction Tasks
3.2.2 Description of Soft Sensing Tasks
Section 2: Methodology
4. Machine Learning Basics for Prediction & Soft Sensing
4.1 ARIMA
4.2 LSTM
4.3 Conv
4.3.1 SVM
4.3.2 Transformer
4.3.3 ELM
4.4 Probabilistic Bayesian Model
4.5 Conclusion
Section 3: Application
5. A Novel Supervised Soft Sensor Framework Based on Convolutional Laplacian Extreme Learning Machine: CNN-LapsELM
5.1 Introduction
5.2 Problem Statement
5.3 Main Framework
5.4 Numerical Examples
5.5 Conclusion
6. A Novel Semi-Supervised Soft Sensor Framework Based on Stacked Auto-Encoder Wavelet Extreme Learning Machine: SAE-WELM
6.1 Introduction
6.2 Problem Statement
6.3 Main Framework
6.4 Numerical Example
6.5 Conclusion
7. A Novel Soft Sensor Based on Laplacian Hessian Semi-Supervised Hierarchical Extreme Learning Machine: LHSS-HELM
7.1 Introduction
7.2 Problem Statement
7.3 Main Results
7.4 Numerical Examples
7.5 Conclusion
8. A Self-Supervised Prediction Framework Based on Deep Long Short-Time Memory for Aluminum Electrolysis: SSDLSTM
8.1 Introduction
8.2 Problem Statement
8.3 Main Results
8.4 Simulation Results
8.5 Conclusion
9. A Self-Supervised Prediction Framework Based on Convolutional Deep Long Short-Time Memory for Aluminum Temperature Application: CNN-SSDLSTM
9.1 Introduction
9.2 Problem Statement
9.3 Main Results
9.4 Numerical Examples
9.5 Conclusion
10. A Novel Probabilistic Prediction Framework Based on Bayesian Machine Learning: BLSTM
10.1 Introduction
10.2 Problem Statement
10.3 Main Results
10.4 Numerical Example
10.5 Conclusion
11. Direct Data-Driven Quantile Regressor Forecaster for Underflow Concentration Soft Sensing: DDQRF
11.1 Introduction
11.2 Problem Statement
11.3 Main Results
11.4 Numerical Example
11.5 Conclusion
12. A Novel Key-Quality Prediction Framework for Industrial Deep Cone Thickener: DualLSTM
12.1 Introduction
12.2 Problem Statement
12.3 Main Results
12.4 Numerical Example
12.5 Conclusion
13. A Deeply-Efficient Long Short-Time Memory Framework for Underflow Concentration Prediction: DE-LSTM
13.1 Introduction
13.2 Problem Statement
13.3 Main Results
13.4 Numerical Example
13.5 Conclusion
14. An Ensemble Prediction Method for Probabilistic Forecasting of Aluminium Electrolysis Process
14.1 Introduction
14.2 Problem Statement
14.3 Main Results
14.4 Numerical Example
14.5 Conclusion
- Edition: 1
- Latest edition
- Published: July 1, 2026
- Language: English
HK
Hamid Reza Karimi
Dr. Karimi received the B.Sc. (First Hons.) degree in power systems from the Sharif University of Technology, Tehran, Iran, in 1998, and the M.Sc. and Ph.D. (First Hons.) degrees in control systems engineering from the University of Tehran, Tehran, in 2001 and 2005, respectively. His research interests are in the areas of control systems/theory, mechatronics, networked control systems, intelligent control systems, signal processing, vibration control, ground vehicles, structural control, wind turbine control and cutting processes. He is an Editorial Board Member for some international journals and several Technical Committee. Prof. Karimi has been presented a number of national and international awards, including Alexander-von-Humboldt Research Fellowship Award (in Germany), JSPS Research Award (in Japan), DAAD Research Award (in Germany), August-Wilhelm-Scheer Award (in Germany) and been invited as visiting professor at a number of universities in Germany, France, Italy, Poland, Spain, China, Korea, Japan, India.
Affiliations and expertise
Professor of Applied Mechanics, Department of Mechanical Engineering, Politecnico di Milano, Milan, ItalyYL
Yongxiang Lei
Dr Yongxiang Lei received the B.Sc. Degree in Automation from the University of South China in 2017 and an M.Sc. in control engineering from Central South University, Changsha, China in 2020. In 2024, received his Ph.D. degree in Mechanical Engineering from Politecnico di Milano. Dr Lei’s research interests are in the areas of machine learning, prediction & control, industrial process modeling, simulation and application, soft sensing.
Affiliations and expertise
Politecnico di Milano, Italy