In efforts to improve precision engineering, a new approach surfaces to break the limitations of current modeling techniques. Researchers Chen Luo, Ao-Jin Li, Jiang Xiao and Ming Li are led by Professor Yun Li of the Shenzhen Advanced Research Institute, and China University of Electronic Science and Technology introduced practical solutions. Their study, published in a scientific report, explains an approach called the Grey-box State Space Model (SSM), which combines the simplicity, accuracy, and transparency of dynamic modeling.
This gray box hybrid model combines basic scientific principles with advanced data analysis, fusing the laws of the physical in the white box, which are symbolic rules that describe things like movement and energy in the real world, as well as general functional approximators such as connected artificial neural networks, involving data-driven training or prediction. This combination creates a model that not only explains but also adjusts various complexities in the real world. Professor Li said: “By integrating expert knowledge into a strong AI framework, we ensure that these models are understandable and effective under different conditions.”
Testing this method on a highly sensitive temperature control system used in a cleaning room proves its effectiveness. These systems are free from dust and contaminants and require temperature regulation of air and water. The Grey-Box model goes beyond the performance of traditional methods and manages unpredictable system changes and unique features better than independent methods.
The Grey-Box model developed using SSM structure uses two transformations. One converts the set irregular nonlinear differential equation into a conventional linear global SSM white box, and the other converts its state-dependent parameters into a conventional local function approximator. Therefore, the laws of physics form the basis of the model, and machine learning is used to dynamically adjust parameter settings. For example, in air temperature control in clean rooms, the model relies on two energy transfer principles that explain how heat moves between objects, while real-time data (information collected as an event occurs) achieves optimal performance. Professor Li explained: “Our model can predict the behavior of new scenarios with significant accuracy, which is crucial for industries where operating conditions change frequently.”
Addressing shared challenges such as incomplete information (involving gaps or missing data and inefficient calculations), the gray-frame framework demonstrates higher adaptability while still providing insights on how it works. This fusion of adaptability and clarity is crucial for actual industrial use.
The future possibilities of the Grey-Box SSM cover a variety of areas, including aerospace, involving the design and production of aircraft and spacecraft, and energy management, which focuses on the efficient use of resources. Professor Li sees this approach as part of a broader action towards smarter, more transparent engineering technology. This transformation represents a future in which the machine not only performs, but also explains its functions, thereby increasing trust and efficiency. “Our goal is to develop efficient and explainable engineering tool AI,” Professor Li said.
Journal Reference
Luo, C., Li, A., Xiao, J., Li, M. &Li,Y. Scientific Reports, 2024. https://doi.org/10.1038/s41598-024-67259-4
About the Author
Yun Li (Classmate, IEEE) received a Ph.D. In 1990, he received his degree from Strathclyde University in Glasgow, England. From 1991 to 2018, he was a lecturer, senior lecturer and professor at the University of Glasgow, and a founding director of the University of Singapore, University of Glasgow, Singapore. He is currently the chairman of the School of Electronic Science and Technology, Shenzhen, China. He has written or co-authored more than 300 papers, one of which is the most popular paper in the IEEE Control System technology transaction almost every month since its publication in 2005. Professor Li is interested in the next generation, explainable artificial intelligence and its engineering applications.
Dr. Chen Luo Her PhD is from the University of Earth Sciences, Wuhan, China. She is currently a postdoctoral researcher in engineering artificial intelligence. Her work addresses key scientific challenges for smart cities and large-scale engineering projects to ensure they are both robust and understandable. Dr. Luo’s contribution aims to support smarter, safer and more sustainable urban development, making her a key figure in the integration of AI and engineering science.
ao-jin li He received his bachelor’s degree from Henan University of Technology in 2021. His research interests include intelligent control, robotics and specific intelligence.
Jiang Xiao In 2022, he received his bachelor’s degree from China University of Electronic Science and Technology in Chengdu, China. He is currently studying for an MS degree at the Shenzhen Advanced Institute of China University of Electronic Science and Technology, Shenzhen, China. His recent research interests include computing intelligence, large language models and their applications in communication systems.
Ming Li He received a bachelor’s degree from China Normal University in Guangzhou, southern China. He is currently a graduate student at the Institute of Advanced Research, Shenzhen University of Electronic Science and Technology, Shenzhen, China. His work focuses on neural network compression. Ming Li is committed to advancing machine learning technology, especially in neural network optimization.
