Leilei Yin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Leilei Yin has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Dunbing Tang
College of Mechanical and Electrical Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China
1 shared publication - FuPeng Xie
School of Automation Nanjing Institute of Technology Nanjing China
1 shared publication - Fupeng Xie
Nanjing Institute of Technology, Nanjing, China
1 shared publication - JingWei Liu
School of Automation Nanjing Institute of Technology Nanjing China
1 shared publication - Li Shao
Nanjing Institute of Technology, Nanjing 211167, China
1 shared publication - Long Zhang
Phoenix Contact China Corporation, Nanjing, China
1 shared publication
Latest Publications
- 17 February 2026
- Published by Institute of Central Computation and Knowledge Inc. in ICCK Transactions on Intelligent Cyber-Physical Systems
- Vol. 1 (1) , 38-50
A Multi-Dimensional Data Learning-Based Production Quality Management Method for Intelligent Manufacturing
In the field of high-end precision manufacturing, quality control in production processes has long been challenged by both spatiotemporal data sparsity and error lag. Traditional offline sampling methods struggle to capture the dynamic fluctuations in production, while single-dimensional feedback controls fall short in addressing the nonlinear coupling between multi-dimensional process parameters and final product quality. To address these challenges, this paper proposes a production quality management system based on multi-dimensional data learning and an active error elimination method. First, to tackle the issue of sparse sampling, an Adaptive Gaussian Process Regression (AGPR) algorithm with mixed kernel functions is introduced to reconstruct continuous production quality time-series states, effectively resolving the "blind spot" problem caused by discrete monitoring. Second, a Dynamic Gated LSTM network with a "Correction Gate" is designed to explicitly model the dynamic intervention mechanism of process control variables on quality evolution, advancing from passive prediction to active deduction. Most importantly, this paper develops an active error elimination strategy using gradient inversion. By minimizing the quality deviation objective function, the optimal combination of process parameters is inversely determined. In practical terms, this approach enables the "dynamic re-matching of machine tool state and process requirements"—intelligently adjusting process parameters (such as injection pressure and holding time) according to the equipment's real-time state (e.g., thermal drift, wear). This method compensates for physical equipment performance degradation through dynamic scheduling. Experimental results show that the system significantly reduces the rejection rate in precision injection molding scenarios, marking a paradigm shift in production from "post-event rejection" to "pre-event self-healing".
- 17 February 2026
- Published by Institute of Central Computation and Knowledge Inc. in ICCK Transactions on Intelligent Cyber-Physical Systems
- Vol. 1 (1) , 38-50
A Multi-Dimensional Data Learning-Based Production Quality Management Method for Intelligent Manufacturing
In the field of high-end precision manufacturing, quality control in production processes has long been challenged by both spatiotemporal data sparsity and error lag. Traditional offline sampling methods struggle to capture the dynamic fluctuations in production, while single-dimensional feedback controls fall short in addressing the nonlinear coupling between multi-dimensional process parameters and final product quality. To address these challenges, this paper proposes a production quality management system based on multi-dimensional data learning and an active error elimination method. First, to tackle the issue of sparse sampling, an Adaptive Gaussian Process Regression (AGPR) algorithm with mixed kernel functions is introduced to reconstruct continuous production quality time-series states, effectively resolving the "blind spot" problem caused by discrete monitoring. Second, a Dynamic Gated LSTM network with a "Correction Gate" is designed to explicitly model the dynamic intervention mechanism of process control variables on quality evolution, advancing from passive prediction to active deduction. Most importantly, this paper develops an active error elimination strategy using gradient inversion. By minimizing the quality deviation objective function, the optimal combination of process parameters is inversely determined. In practical terms, this approach enables the "dynamic re-matching of machine tool state and process requirements"—intelligently adjusting process parameters (such as injection pressure and holding time) according to the equipment's real-time state (e.g., thermal drift, wear). This method compensates for physical equipment performance degradation through dynamic scheduling. Experimental results show that the system significantly reduces the rejection rate in precision injection molding scenarios, marking a paradigm shift in production from "post-event rejection" to "pre-event self-healing".
- 26 August 2025
- Published by Springer Nature in The International Journal of Advanced Manufacturing Technology
- Vol. 140 (3-4) , 1581-1600
Digital twin-driven decision support approach of production process for smart manufacturing
- 26 August 2025
- Published by Springer Nature in The International Journal of Advanced Manufacturing Technology
- Vol. 140 (3-4) , 1581-1600
Digital twin-driven decision support approach of production process for smart manufacturing
- 9 February 2021
- Published by Wiley in Journal of Computing and Information Science in Engineering
- Vol. 21 (4)
Risk Analysis of Engineering Change for Distributed Product Design
Nowadays, customer demand for satisfactory product developed in limited time is rapidly posing a major challenge to product design, and more distributed products are developed to address these concerns. In the distributed product design, engineering change (EC) is an inevitable phenomenon and consumes much production time. It is necessary to assess the design change effectively in advance. Some methods and tools to predict and analyze the change propagation influence have been provided. From the perspective of design change duration, our work extends the method of assessing design change by incorporating risk factors from different working groups in multiple design sites, and functional maintenance during the change propagation. The primary result of this work is the provision of a design support to acquire the optimal design change scheme by estimating the duration. In this paper, risk factor for distributed design is applied to the influence evaluation of change propagation, which implies an increase of change propagation influence due to the varying levels of expertise and possible lack of communication. Besides, a deterministic simulation model is proposed to assess the change propagation schemes. The model combines the effects of design change parallelism, iteration, and change propagation for the distributed product design. Based on the simulation results, a more focused discussion and identification of suitable design change schemes can be made. A case study of an assembly tooling for the reinforced frame is implemented to demonstrate how the developed method can be applied. Finally, the method is initially discussed and evaluated.
- 9 February 2021
- Published by Wiley in Journal of Computing and Information Science in Engineering
- Vol. 21 (4)
Risk Analysis of Engineering Change for Distributed Product Design
Nowadays, customer demand for satisfactory product developed in limited time is rapidly posing a major challenge to product design, and more distributed products are developed to address these concerns. In the distributed product design, engineering change (EC) is an inevitable phenomenon and consumes much production time. It is necessary to assess the design change effectively in advance. Some methods and tools to predict and analyze the change propagation influence have been provided. From the perspective of design change duration, our work extends the method of assessing design change by incorporating risk factors from different working groups in multiple design sites, and functional maintenance during the change propagation. The primary result of this work is the provision of a design support to acquire the optimal design change scheme by estimating the duration. In this paper, risk factor for distributed design is applied to the influence evaluation of change propagation, which implies an increase of change propagation influence due to the varying levels of expertise and possible lack of communication. Besides, a deterministic simulation model is proposed to assess the change propagation schemes. The model combines the effects of design change parallelism, iteration, and change propagation for the distributed product design. Based on the simulation results, a more focused discussion and identification of suitable design change schemes can be made. A case study of an assembly tooling for the reinforced frame is implemented to demonstrate how the developed method can be applied. Finally, the method is initially discussed and evaluated.