Recent Progress in Data-Driven Intelligent Modeling and Optimization Algorithms for Industrial Processes

Abstract

This editorial discusses recent progress in data-driven intelligent modeling and optimization algorithms for industrial processes. With the advent of Industry 4.0, the amalgamation of sophisticated data analytics, machine learning, and artificial intelligence has become pivotal, unlocking new horizons in production efficiency, sustainability, and quality assurance. Contributions to this Special Issue highlight innovative research in advancements in work-sampling data analysis, data-driven process choreography discovery, intelligent ship scheduling for maritime rescue, process variability monitoring, hybrid optimization algorithms for economic emission dispatches, and intelligent controlled oscillations in smart structures. These studies collectively contribute to the body of knowledge on data-driven intelligent modeling and optimization, offering practical solutions and theoretical frameworks to address complex industrial challenges.

1. Introduction

As industrial systems become increasingly complex, traditional mechanistic modeling methods are struggling to address the significant challenges posed by modern production processes. This is especially true for systems characterized by strong multivariable coupling, nonlinearity, and time-variance, where the limitations of expert-based and traditional modeling methods have become increasingly evident. As a result, data-driven modeling approaches have gradually become the focus of research. By mining the vast amounts of data embedded in industrial production processes and combining data-driven models with optimization algorithms, the real-time monitoring, state prediction, control, optimization scheduling, and fault diagnosis of industrial processes can be achieved. This provides solid technical support for the intelligent transformation of modern manufacturing, while ensuring the stability of production processes and product quality [1,2,3].

Traditional data-driven modeling and optimization algorithms primarily include multivariate statistical methods and machine learning models. Multivariate statistical methods, such as the use of one-way ANOVA for the detection of abnormalities in the sintering process [4] and Gaussian process regression for performance prediction in sintering [5], establish statistically based models to provide reliable decision support for industrial operations. With the rapid development of technologies such as big data, cloud computing, and the Internet of Things, researchers have become increasingly interested in the application of machine learning models in industrial production. Examples include resource allocation decision support systems that combine multi-objective optimization and meta-learning [6], artificial neural network optimization for solar energy multi-supply systems [7], and dynamic compressive strength prediction models for freeze–thaw rock based on a swarm intelligence optimization of hybrid support vector regression [8]. Machine learning has significant advantages in handling complex, dynamically changing data by automatically learning and adapting to different data patterns.

As modern industry continues to advance toward automation and intelligence, advanced artificial intelligence techniques, such as deep learning, reinforcement learning, and ensemble learning, have become mainstream in data-driven intelligent modeling. Examples include the development of optimal control strategies for nonlinear industrial production systems using spatiotemporal graph convolutional networks [9] and reinforcement learning [10], as well as bearing fault detection methods using convolutional neural networks [11] and sintering endpoint prediction models [12]. Artificial intelligence methods are highly effective in processing high-dimensional, nonlinear, temporal, and dynamically changing data. They possess strong feature extraction and learning capabilities, enabling them to adapt to real-time and complex optimization problems in big data environments.

Moreover, hybrid methods combining physical models with data-driven models can effectively integrate the advantages of physical models in causal analysis with the efficiency of data-driven methods in correlation analysis. For example, seismic damage prediction methods using finite element calculations in conjunction with multi-particle swarm optimization algorithms [13], integrated machine learning methods combined with physics-based empirical models for ship operation status recognition [14], and energy control strategies that combine mechanistic modeling with machine learning [15] all demonstrate the unique advantages of hybrid physical and data-driven models in addressing practical problems.

However, numerous challenges have yet to be addressed, particularly when tackling high-dimensional, nonlinear, and multi-scale industrial problems. Therefore, further research efforts are essential to advance both the theoretical and practical developments in this critical field.

2. An Overview of Published Articles

In the ever-evolving landscape of industrial process modeling and optimization, data-driven intelligent algorithms have emerged as a transformative force. This Special Issue aims to explore the intersection of data-driven approaches, intelligent modeling, and optimization algorithms in the context of industrial processes. With the relentless growth of Industry 4.0, the integration of advanced data analytics, machine learning, and artificial intelligence has become imperative to opening up new possibilities to improve production efficiency, sustainability, and quality assurance in industrial processes.

This Special Issue aims to explore the multifaceted aspects of data-driven intelligent modeling and optimization algorithms for industrial processes. The main objectives are to harness the power of data to improve control, decision making, and parameter optimization, and to drive industrial systems to unprecedented levels of efficiency, reliability, and adaptability. The research areas of this Special Issue include data-driven modeling, intelligent data representation, integration/hybrid modeling, machine learning and optimization, advanced machine learning algorithms, hybrid models with optimization algorithms, adaptive learning algorithms, intelligent process monitoring, real-time data monitoring and analysis, soft sensing technologies, operating mode perception and recognition, decision support systems, intelligent decision support systems, the integration of optimization algorithms, and human–machine collaboration for enhanced decision making.

The analysis of the papers published in this Special Issue is shown in

Table 1. Analysis of the published contributions in the Special Issue.

No.	DOI	Research Area	Focus	Type of Research	Industry	Country
1	10.3390/a17030102	Optimization algorithms	hot strip rolling; looper; production stability; root cause traceability	Mathematical modeling	Manufacturing	China
2	10.3390/a17050204	Process optimization	modeling finite elements; cutting tool; surface quality; stress and strain distribution; chip size	Mathematical modeling	Aerospace	Morocco
3	10.3390/a17070313	Optimization algorithms	combined economic emission dispatch; load shifting; demand side management; crow search algorithm; arithmetic optimization algorithm	Mathematical modeling	Power systems	South Africa
4	10.3390/a17070319	Optimization algorithms	hybrid energy storage system; adaptive sliding-mode controller; battery degradation; supercapacitor; Zeta converter	Mathematical modeling	Power systems	Colombia
5	10.3390/a17050183	Decision support strategies	work sampling; observations; proportions; interdependence between activities	Empirical research	Production management	Slovenia
6	10.3390/a17050188	Decision support strategies	process choreography; data-driven; process mining	Mathematical modeling	Production management	Mexico
7	10.3390/a17110530	Decision support strategies	ideal solution; sequential three-way decisions; muti-attribute decision making	Mathematical modeling	Production management	China
8	10.3390/a17060233	Control strategies	load frequency control; switching system; event-triggered; model-free adaptive control	Mathematical modeling	Power systems	China
9	10.3390/a17110505	Control strategies	vibration; intelligent control; piezoelectric structures; H2 criterion; H-infinity criterion	Mathematical modeling	Manufacturing	Greece
10	10.3390/a17120543	Control strategies	Sliding mode control; data-driven techniques; intelligent algorithms	Review	Manufacturing	Colombia
11	10.3390/a17050216	Process monitoring	control chart; variability; variance; interface; Shiny package	Mathematical modeling	Manufacturing	Brazil
12	10.3390/a17050204	Optimization algorithms	maritime emergency rescue; intelligent navigation; path planning; A* algorithm; B-spline interpolation; regional search	Mathematical modeling	Maritime rescue	China

. A considerable amount of research has been conducted on data-driven modeling and optimization algorithms, covering a wide range of research areas related to automatic control. References [16,17,18,19] focus on optimization algorithms in data-driven models, while References [20,21,22] examine the decision support strategies in industrial production. References [23,24,25] explore the control strategies, Reference [26] investigates process optimization, and Reference [26] addresses process monitoring.

It is worth noting that References [16,23,27] all involve the application of fault detection methods. Reference [16] investigates an optimization algorithm for fault tracking in hot rolling processes, using an equipment process accuracy evaluation to trace the root causes of failures. Reference [23] applies data-driven methods to detect faults in power systems and proposes an effective load frequency control strategy to address potential fault scenarios. Reference [27] employs real-time process monitoring techniques to assess fluctuations in the production process and combines these with statistical analysis to provide early warnings for fault detection, ensuring production stability.

References [18,24,25] all involve the application of system control strategies. Reference [18] proposes an effective battery management strategy that uses controllers to mitigate battery aging and manage high-frequency transients. Reference [24] explores the application of intelligent control strategies in piezoelectric structures, employing smart piezoelectric patches and advanced control methods to effectively suppress vibrations in engineering structures. Reference [25] introduces the application of data-driven, model-free sliding mode control techniques.

References [20,21,22] focus on the application of decision support strategies. Reference [20] presents a data analysis method based on the work sampling technique, studying work activities and their interrelationships in production management and providing decision-makers with more accurate production optimization tools. Reference [21] helps production managers optimize workflows by discovering the process choreography. Reference [22] proposes a sequential three-way classification ranking method based on the VIKOR method, aiding decision-makers in making optimal choices.

In addition, Reference [19] introduces real-time monitoring in maritime rescue missions through path planning and intelligent scheduling methods. Reference [26] utilizes finite element modeling to simulate surface quality and stress distribution in the ultrasonic vibration-assisted milling process, optimizing material processing. Reference [17] presents a hybrid crow search arithmetic optimization algorithm for solving the weighted combined economic emission dispatch problem in power systems, optimizing emissions and cost issues in the power scheduling process.

Finally, the industries covered in these studies include manufacturing, power systems, maritime rescue, and aerospace. The authors are mainly from China, but scholars from Slovenia, Mexico, Morocco, Brazil, South Africa, Colombia, and Greece also contributed to our Special Issue.

3. Conclusions

This Editorial presents the latest advancements in data-driven intelligent modeling and optimization algorithms, with a particular focus on the application of cutting-edge methods and technologies in recent years. These powerful intelligent algorithms leverage data for control, decision making, and parameter optimization, driving industrial systems to unprecedented levels of efficiency, reliability, and adaptability. The research indicates that with the rapid development of Industry 4.0, data-driven intelligent modeling and optimization algorithms have become a key driving force in enhancing production efficiency, promoting sustainability, and ensuring product quality.