Applications of Artificial Intelligence in Process Systems Engineering

Front Cover
Applications of Artificial Intelligence in Process Systems Engineering
Copyright
Contents
Contributors
Chapter 1: Artificial intelligence in process systems engineering
	1. What is process system engineering
		1.1. Background
		1.2. Challenges
	2. What is artificial intelligence?
		2.1. Background
		2.2. Deep learning
	3. The AI-based application in PSE
		3.1. Physical properties prediction and product design
		3.2. Process modeling
		3.3. Fault detection and diagnosis
		3.4. Process optimization and scheduling
	4. Summary
	Acknowledgment
	References
Chapter 2: Deep learning in QSPR modeling for the prediction of critical properties
	1. Introduction
	2. Methodology
		2.1. The signature molecular descriptor
		2.2. Data preparation: Molecules encoding and canonizing
		2.3. Data preparation: Atom embedding from chemical bonds
		2.4. Deep neural network
		2.5. Model training and evaluation
	3. Results and discussion
		3.1. Embedding of bond substrings
		3.2. The DNN performance
		3.3. Comparisons with two existing methods
		3.4. Distinction of isomers
	4. Conclusions
	Acknowledgments
	References
Chapter 3: Predictive deep learning models for environmental properties
	1. Introduction
	2. Methodology
		2.1. Data acquisition and processing
		2.2. Tree structures in information processing
		2.3. Signature molecular descriptor and encoding rules
		2.4. Structural features and parameters of DNN
	3. Results and discussion
		3.1. List of molecular features
		3.2. Training process
		3.3. Generalization ability
		3.4. Applicability domain
		3.5. External competitiveness
		3.6. Discriminative power in isomers
	4. Conclusions
	Acknowledgments
	References
Chapter 4: Automated extraction of molecular features in machine learning-based environmental property prediction
	1. Introduction
	2. Methodology
		2.1. Data collection
		2.2. Feature extraction
		2.3. Neural network design
		2.4. Model training
	3. Results and discussion
		3.1. Feature vector
		3.2. Training process
		3.3. Model performance
		3.4. Comparison with reported models
	4. Conclusions
	Acknowledgments
	References
Chapter 5: Intelligent approaches to forecast the chemical property: Case study in papermaking process
	1. Introduction
	2. Literature review
	3. Intelligent prediction method
		3.1. ANN
		3.2. SVM
		3.3. Ensemble learning
	4. Beating degree prediction model of tissue paper
		4.1. Data collection and preprocessing
			4.1.1. Data collection
			4.1.2. Data preprocessing
		4.2. Modeling the beating degree of tissue paper
			4.2.1. Correlation analysis
			4.2.2. Algorithm description
			4.2.3. Parameter adjustment
		4.3. Industrial verification and feature importance ranking
	5. Conclusion
	Acknowledgments
	References
Chapter 6: Machine learning-based energy consumption forecasting model for process industry-Hybrid PSO-LSSVM algorithm el ...
	1. Introduction
		1.1. Literature review
		1.2. Algorithm selection
	2. Methodology
		2.1. Data preprocessing
		2.2. Input variables selection methods
			2.2.1. Correlation function
			2.2.2. Lag autocorrelation function
		2.3. The algorithm descriptions
			2.3.1. LSSVM algorithm
			2.3.2. PSO algorithm
			2.3.3. The hybrid PSO-LSSVM algorithm
	3. Results and discussion
		3.1. Input variables selection
			3.1.1. Case 1
			3.1.2. Case 2
		3.2. Model verification
	4. Conclusions
	Appendix A
	Appendix B
	Appendix C
	Appendix D
	References
Chapter 7: Artificial intelligence algorithm application in wastewater treatment plants: Case study for COD load prediction
	1. Introduction
	2. Literature review
	3. Artificial intelligence algorithm model
	4. Case study: COD prediction model based on the GBDT algorithm
		4.1. Data source
		4.2. Model evaluation index
		4.3. Model parameter selection
			4.3.1. Learning_rate
			4.3.2. N_estimators
			4.3.3. Max_depth
			4.3.4. Min_samples_split
	5. Discussion of results
	6. Conclusion
	References
Chapter 8: Application of machine learning algorithms to predict the performance of coal gasification process
	1. Introduction
	2. Materials and methods
		2.1. Dataset description
		2.2. Data preparation
		2.3. Evaluation criteria
		2.4. Modeling
			2.4.1. Sequential minimal optimization-based support vector regression
			2.4.2. Gaussian process regression
			2.4.3. Lazy K-star
			2.4.4. Lazy IBk
			2.4.5. Alternating model tree
			2.4.6. Random forest
			2.4.7. M5Rules
	3. Results and discussion
		3.1. Comparison of ML algorithms
		3.2. Comparison with previous studies
	4. Conclusions and future perspectives
	References
Chapter 9: Artificial neural network and its applications: Unraveling the efficiency for hydrogen production
	1. Introduction
	2. Artificial neural network
	3. Principle of ANN
	4. Methodology of ANN model
		4.1. Collection of data and its preparation
		4.2. Selection of structural parameters
		4.3. ANN model training
		4.4. Model testing and result analysis
	5. Applications of ANN model
		5.1. Forecasting of river flow
		5.2. ANN model for prediction of depression in aged population
		5.3. Application in various sectors
	6. ANN and hydrogen production
	7. Conclusion
	Acknowledgments
	References
Chapter 10: Fault diagnosis in industrial processes based on predictive and descriptive machine learning methods
	1. Introduction
	2. FDD in large-scale processes
		2.1. Existing FDD methods: Limitation and needs
		2.2. Selection criteria for FDD methods in process industries
	3. Data-driven FDD methods in industrial systems: A review
		3.1. Machine learning-based FDD methods
			3.1.1. Predictive FDD methods
				3.1.1.1. Artificial neural network (ANN)
				3.1.1.2. Support vector machines (SVM)
				3.1.1.3. Bayesian networks (BN)
			3.1.2. Descriptive FDD methods
				3.1.2.1. Decision trees
				3.1.2.2. Logical analysis of data
				3.1.2.3. Rough set theory
		3.2. Multi-variate statistical process monitoring methods
			3.2.1. Principal component analysis (PCA)
			3.2.2. Projection to latent structures (PLS)
			3.2.3. Independent component analysis
			3.2.4. Fisher discriminant analysis
			3.2.5. Kernel methods-kernel PCA (KPCA), kernel PLS (KPLS), and kernel ICA (KICA)
		3.3. Integrated and hybrid FDD methods
	4. Fault diagnosis in pulp and paper mills: Case studies
		4.1. Case 1: Black liquor recovery boiler (BLRB) in kraft pulp and paper mills
		4.2. Case 2: Reboiler system in thermomechanical (TMP) pulp mills
		4.3. Case 3: Heat recovery system for hot water production in TMPs
	5. Concluding remarks and lesson learned
	Acknowledgment
	References
Chapter 11: Application of artificial intelligence in modeling, control, and fault diagnosis
	1. Artificial neural network
		1.1. Introduction
		1.2. The architecture of neural networks
			1.2.1. Feedforward neural networks
			1.2.2. Recurrent neural networks
			1.2.3. Stacked neural networks
			1.2.4. Auto Encoder
			1.2.5. Hybrid neural networks
		1.3. Application of neural network in chemical processes
			1.3.1. Application of artificial neural network in simulation, modeling, and optimization
				1.3.1.1. Polymerization
				1.3.1.2. Fuel cell
				1.3.1.3. Combustion and fuel
				1.3.1.4. Petrochemical
				1.3.1.5. Environment
				1.3.1.6. Biotechnology
				1.3.1.7. Nanotechnology
				1.3.1.8. Mineral
				1.3.1.9. Other application
			1.3.2. Application of artificial neural network in control
				1.3.2.1. Polymerization
				1.3.2.2. Fuel cell
				1.3.2.3. Fuel and combustion
				1.3.2.4. Petroleum
				1.3.2.5. Environment
				1.3.2.6. Biotechnology
				1.3.2.7. Other application
			1.3.3. Application of network in fault detection and diagnosis
	2. Fuzzy logic
		2.1. Introduction of fuzzy
		2.2. Application of fuzzy controller in process control
		2.3. Simulation: A sample of process control by fuzzy controller
	3. Support vector machine
		3.1. Introduction
		3.2. SVM classifiers
			3.2.1. Separable case
		3.3. SVM for the nonlinear and nonseparable cases
		3.4. Nonlinear SVMs
	References
Chapter 12: Integrated machine learning framework for computer-aided chemical product design
	1. Introduction
	2. An integrated ML framework for computer-aided molecular design
	3. Establishment of ML model for computer-aided molecular design
		3.1. Data collection
		3.2. Data preprocessing and feature engineering
		3.3. Model establishment
		3.4. Chemical product design
	4. Case studies
		4.1. A ML-based atom contribution methodology for the prediction of charge density profiles and crystallization solvent d ...
		4.2. A ML-based computer-aided molecular design/screening methodology for fragrance molecules
	5. Conclusions
	Acknowledgments
	References
Chapter 13: Machine learning methods in drug delivery
	1. Introduction
	2. Types of machine learning methods used in drug delivery
	3. Applications of machine learning methods in drug delivery
		3.1. Artificial neural networks (ANNs) in modeling the drug loading in lipophilic nanoparticles
		3.2. Gaussian processes (GPs) in modeling the drug loading in lipophilic nanoparticles
		3.3. Utilizing unsupervised machine learning methods in detecting the stability of soft nanoscale drug carriers
		3.4. Capturing the highly loaded drugs on gelatin nanoparticles using different machine learning methods
		3.5. The application of support vector machines as tools to identify and classify compounds with potential as transdermal ...
	4. Conclusion
	References
Chapter 14: On the robust and stable flowshop scheduling under stochastic and dynamic disruptions
	1. Introduction
	2. Literature review
	3. Problem formulation
		3.1. Deterministic flowshop problem
		3.2. The proactive-reactive approach
	4. Hybridized evolutionary multiobjective optimization (EMO) methods
		4.1. Evaluation of baseline schedule in the proactive stage
			4.1.1. Slack-based and neighborhood-based surrogates (R1,S1)
			4.1.2. Breakdown and repair estimation-based surrogates (R2,S2)
			4.1.3. Simulation-based surrogates (R3,S3)
		4.2. A new hybridization strategy
		4.3. Four hybridized EMO algorithms
			4.3.1. Hybrid quantum genetic algorithm
			4.3.2. Hybrid particle swarm optimization
			4.3.3. Hybrid quantum-inspired iteration greedy
			4.3.4. Hybrid simulated annealing
	5. Computational study
		5.1. Testing instances
		5.2. Parameters setting
		5.3. Performance metrics for the Pareto front
		5.4. Computational results
			5.4.1. Analysis of the hybridization strategy
			5.4.2. Analysis of robustness and stability measures against determined schedule
			5.4.3. Analysis of the reactive methods
	6. Conclusions and future research directions
	Acknowledgments
	References
Chapter 15: Bi-level model reductions for multiscale stochastic optimization of cooling water system
	1. Introduction
	2. Methodology
	3. Optimal design of experiments
	4. Multisample CFD simulation
	5. Reduced models construction
		5.1. Level 1: Physics-based ROM
			5.1.1. Step 1: Singular value decomposition method
			5.1.2. Step 2: Kriging interpolation
		5.2. Level 2: Data-driven ROM
			5.2.1. High-dimensional model representation
		5.3. Model evaluation
	6. Illustrative example
		6.1. Example I: Model reduction
		6.2. Example II: Multiscale optimization
	7. Conclusion
	References
Chapter 16: Artificial intelligence algorithm-based multi-objective optimization model of flexible flow shop smart scheduling
	1. Introduction
	2. Literature review
	3. Flexible flow shop production scheduling model
		3.1. Problem description
		3.2. Production scheduling model
		3.3. GA method introduction
	4. Case study
		4.1. Production process of ball mill shop
		4.2. Optimal object
		4.3. Production cost model
		4.4. Experimental data
		4.5. Parameter setting
		4.6. Result analysis
	5. Conclusions
	References
Chapter 17: Machine learning-based intermittent equipment scheduling model for flexible production process
	1. Introduction
	2. Problem description and solution
		2.1. Dispatching status of intermittent production equipment
		2.2. NSGA-II algorithm overview
		2.3. Objective function and constraints
	3. Case study
		3.1. The pulping process
		3.2. Technical route
		3.3. Data collection
		3.4. Establishment of a liquid level calculation model
			3.4.1. Slurry flow rate
			3.4.2. Slurry tank liquid level
		3.5. Schedule planning model of pulping equipment based on the NSGA-II algorithm
		3.6. Industrial verification
			3.6.1. Adjustment of scheduling plan model parameters
			3.6.2. Results of industrial verification
	4. Conclusion
	References
Chapter 18: Artificial intelligence algorithms for proactive dynamic vehicle routing problem
	1. Introduction
	2. Main approaches for PDVRP
	3. Problem description
		3.1. The vehicle speed of time-dependence
		3.2. Penalty cost of soft time windows
		3.3. The conception of proactive distribution
			3.3.1. Calculating prospect value
			3.3.2. Proactive clustering algorithm
	4. TDVRPSTW formulation
	5. The algorithm
		5.1. Coding structure
		5.2. Genetic operator design
			5.2.1. Crossover operator
			5.2.2. Mutation operator
			5.2.3. Pseudo code of the SA-GA
	6. Numerical experiment
		6.1. Evaluation of the algorithms performance
		6.2. A real case study
			6.2.1. Customer classification
			6.2.2. Evaluation of dynamic customers
			6.2.3. Clustering for the targeted customers
			6.2.4. Proactive tour plan
			6.2.5. Route fitting
	7. Conclusions and future research directions
	References
Back Cover