Special issue on data-driven modeling and analytics for optimization of complex manufacturing systems

As the core issue for improving manufacturing system performance (such as product quality, production efficiency and cost, etc.), modeling, analysis and optimization of manufacturing systems have always been studied towards smart manufacturing. However, due to the large-scale and dynamic characteristics of complex manufacturing systems, the causal relationship between system performance and manufacturing process parameters is difficult to determine (Sun, Qin, and Zhuang 2022). Therefore, traditional models and algorithms are now facing some challenges such as the ‘Curse of Dimensionality’, high computational complexity and so on.

With the rapid development of information technology, massive production data are captured. Practitioners and academics are paying more and more attention to the huge value hidden behind the data (Qin, Zha, and Zhang 2018). Data-driven technology can effectively help for revealing the inherent laws of complex manufacturing processes, transforming data into production and operation knowledge, optimizing production processes, improving product quality and production efficiency, as well as enhancing product lifecycle management level.

Data-driven modeling and analysis has become one of the most promising methods for optimization of complex systems, and has made important breakthroughs in many research areas (Runge et al. 2019; Severson et al. 2019). In biology science, data-driven modeling and analysis has been used to quantitatively identify direct dependencies between genes, reconstruct gene regulatory network and causal relations (Zhao et al. 2016), and identify cell behaviours affecting the observed aggregation dynamics without full knowledge of the underlying biological mechanisms (Cotter et al. 2017). In medical science, data-driven modeling and analysis can not only systematically explore the molecular complexity of specific diseases, but also identify disease modules and pathways, as well as the molecular relationships between distinct phenotypes (Cheng, Kovács, and Barabási 2019). These successful cases provide a new way of thinking for the modeling, analysis and optimization of complex manufacturing systems in the industrial field. How to use the big data to establish an effective model describing the complex manufacturing system? How to analyze the data-driven model to reveal the law of system operation? How to dynamically regulate the data-driven model to optimize the system performance?

Researchers have been developing new models, methods, solutions and tools for these problems (Zhuang et al. 2022). This special issue provides an opportunity for academia and practitioners to share state-of-the-art research and cases in the field of modeling, analysis and optimization in manufacturing systems. There are 13 papers accepted, covering new approaches for enhancing complex manufacturing systems by making full use of emerging technologies such as machine learning, complex network and intelligence algorithms.

The first paper A hybrid machine learning method for procurement risk assessment of non-ferrous metals for manufacturing firms, by Jian Ni, Yan Hu and Ray Y. Zhong, proposes a data-driven approach using the state-of-art machine learning techniques to assess and forecast the procurement risks of non-ferrous metals associated with complex manufacturing systems. This research can serve as a useful reference for the effective assessment and control of procurement risk of nonferrous metals in industries, such as mechanical manufacturing, aerospace, electricity and household appliances.

The next paper A data-driven robust optimization method for the assembly job-shop scheduling problem under uncertainty, by Peng Zheng, Peng Zhang, Junliang Wang, Jie Zhang, Changqi Yang and Yongqiao Jin, studies the production scheduling problem in an assembly manufacturing system with uncertain processing time and random machine breakdown. This problem is modelled as a bi-objective optimization problem with the aim to simultaneously minimize makespan and performance deviation of the schedule. Utilizing data-driven idea, this paper uses a boosting radial basis function network as a surrogate model to efficiently and effectively estimate the performance deviation.

The paper New approaches for rebalancing an assembly line with disruptions, by Yuchen Li, Zixiang Li and Francisco Saldanha-da-Gama, proposes two policies, a periodic rebalancing policy and a data-driven rebalancing policy, to solve the assembly line balancing problem under disruptions. The goal is to minimize the total cost, which include production and rebalancing costs. This study shows that considering specific moments in time for rebalancing an assembly line is a possibility worth considering when dealing with uncertainty in the processing times.

The fourth paper Deep learning for industrial image: challenges, methods for enriching the sample space and restricting the hypothesis space, and possible issue, by Tianyuan Liu, Jinsong Bao, Junliang Wang and Jiacheng Wang, provides an overview of approaches commonly used in industry by enriching the sample space and limiting the hypothesis space. This paper can guide more researchers to address the current and future directions of the deep learning research and application in industrial image.

The next paper A framework for credit-driven smart manufacturing service configuration based on complex networks, by Shijie Wang, Yingfeng Zhang, Cheng Qian and Dang Zhang, proposes a credit-driven service configuration method, which provides readers with a clear view of the inter-layer relationship between the manufacturing physical device layer and virtual manufacturing service layer. Under the framework of complex manufacturing network with a four-layer structure, the theoretical basis, application scenarios, advantages and feasibility of the proposed items are expounded in combination with the complex network theory.

In the sixth paper A sub-assembly division method based on community detection algorithm, by Yu Zheng, Liang Chen, Peng Jiang and Huanchong Cheng, assemblies comprising parts with different connection relationships are modelled as the weighted graph in which the weighted adjacent list is used to accomplish calculation of the subassembly division. The proposed method is applied in the subassembly division process of a practical pumping unit, with the influence of the weight assignment and deviation parameters being analysed in detail.

The paper KGAssembly: Knowledge graph-driven assembly process generation and evaluation for complex components, by Bin Zhou, Jinsong Bao, Zhiyu Chen and Yahui Liu, proposes a knowledge graph-driven assembly process generation and evaluation method for complex components. It provides assembly expert knowledge support for the evaluation method of interference detection of assembly sequence based on point cloud assembly feature recognition. This method is further evaluated by assembling an aero-engine compressor rotor.

The next paper knowledge representation in Industry 4.0 scheduling problems, by Daniel A. Rossit and Fernando Tohmé, proposes a knowledge architecture to improve the ability to solve scheduling problems by incorporating artificial intelligence-based complements to manufacturing scheduling systems. In this way, it summarizes the implicit criteria used by human schedulers. The architecture presented here records this knowledge in data structures compatible with the structure of scheduling problems. In further iterations, this knowledge crystallizes into a sound and smart structure.

The ninth paper Development of dynamic scheduling in semiconductor manufacturing using a Q-learning approach, by Yeou-Ren Shiue, Ken-Chuan Lee and Chao-Ton Su, develops a dynamic scheduling scheme that uses a Q-learning-based multiple-dynamic-scheduling-rule (MDSR) selection mechanism to support shop floor control in semiconductor wafer fabrication. Their results demonstrated that the proposed scheme provided superior production performance, according to various production performance measures, when compared with an approach that uses fixed scheduling (dispatching) decision rules and the classic MDSR approach.

The paper A multi-agent and internet of things framework of digital twin for optimized manufacturing control, by Qingwei Nie, Dunbing Tang, Haihua Zhu and Hongwei Sun, proposes a framework for the intelligent digital twin shopfloor, to reasonably allocate resources according to the production requirements and effectively avoid the influence of external interference on the manufacturing system. A design case is studied to illustrate that the intelligent digital twin shopfloor can configure resources and deal with disturbances effectively.

The eleventh paper Competitive strategy and production strategy of the original equipment manufacturer and the third-party remanufacturer in remanufacturing, by Minzhen Xu, Tong Shu, Shou Chen, Shouyang Wang and Shulin Lan, proposes two different competitive strategies of the original equipment manufacturer and the third-party remanufacturer in remanufacturing: investment strategy and non-investment strategy. This paper also provides some management insight for the industry and government administration.

The twelfth paper A memetic algorithm for energy-efficient scheduling of integrated production and shipping, by Jian Chen, Tong Ning, Gangyan Xu and Yang Liu, proposes a memetic algorithm to incorporate a knowledge-driven local search strategy considering the balance between exploration and exploitation. This research aims to address an energy-efficient scheduling problem of production and shipping for minimizing both makespan and energy consumption. It practically contributes to improving productivity and energy efficiency for the production-shipping supply chain of make-to-order products.

The final paper Reinforcement learning based optimal decision making towards product lifecycle sustainability, by Yang Liu, Miying Yang and Zhengang Guo, develops a method that can adopt the most suitable AI techniques to support decision-making for sustainable operations based on the available lifecycle data. It aims to address the challenge relates to the scarcity and uncertainties of data across the product lifecycle. This paper contributes to knowledge by demonstrating an empirically grounded industrial case using reinforcement learning for improved product lifecycle sustainability by effectively prolonging the product lifetime and reducing environmental impact.

We would like to thank all the reviewers who gave their significant comments and suggestions for improving the published papers in this special issue. Thank you all the contributors to make the publication of this special issue. Special thanks should be given to Professor Stephen T Newman, Editor-in-Chief of the International Journal of Computer Integrated Manufacturing and Professor Aydin Nassehi, Senior Editor, who gave their great support. We hope that this special issue will bridge the academic and practitioners to enhance the complex manufacturing systems in the future.