Integrating Physics-Based and Data-Driven Models Using Reinforcement Learning and Adversarial Learning for Intelligent Manufacturing

Guangda Xu; Jihong Chen; Huicheng Zhou; Jianzhong Yang; Dehai Huang

doi:10.1016/j.eng.2025.12.039

Engineering ›› :202512039 DOI: 10.1016/j.eng.2025.12.039

Research

research-article

Integrating Physics-Based and Data-Driven Models Using Reinforcement Learning and Adversarial Learning for Intelligent Manufacturing

Author information +

History +

PDF

Abstract

Hybrid models that integrate physics-based (PB) and data-driven (DD) approaches have gained increasing attention in intelligent manufacturing. To address the dual challenges of limited dynamic adaptability in static hybrid models and the difficulty of traditional residual learning in handling unmodeled dynamics, we propose a PB-DD hybrid model using reinforcement learning (RL) and adversarial learning (AL), namely, RA-PBDD, which is different from a DD-centric hybrid approach that often suffers from limited training data representativeness, whereas PB-centric hybrid methods face challenges because of model incompleteness. RL is employed for one-time parameter identification in PB models, overcoming the limitations of conventional stepwise techniques. A residual value model using AL with neural networks is proposed, which achieves improved accuracy through iterative refinement compared with standard learning approaches. Additionally, we analyze the operational principles of the PB, DD, and hybrid PB-DD models from a state-space perspective which is a particular angle to systematically examine and contrast the characteristics of these model types. The effectiveness of RA-PBDD is validated through a case involving a machine tool feed system. The results confirm the superiority of RA-PBDD in both prediction and generalizability accuracy over the stand-alone PB, DD models and other two hybrid models. Specifically, it improves the prediction accuracy by up to 73% from 15.4 to 4.2 μm and machining precision by up to 46.7% from 15.2 to 8.1 μm.

Keywords

Hybrid model / Reinforcement learning / Adversarial learning

Cite this article

Download citation ▾

Guangda Xu, Jihong Chen, Huicheng Zhou, Jianzhong Yang, Dehai Huang. Integrating Physics-Based and Data-Driven Models Using Reinforcement Learning and Adversarial Learning for Intelligent Manufacturing. Engineering 202512039 DOI:10.1016/j.eng.2025.12.039

登录浏览全文

4963

注册一个新账户忘记密码

References

Publishing order | Descend order by publishing year | Descend order by cited within

[1]	Zhou J, Li P, Zhou Y, Wang B, Zang J, Meng L. Toward new-generation intelligent manufacturing. Engineering 2018; 4(1):11-20.

[2]	Banerjee S, Jesubalan NG, Kulkarni A, Agarwal A, Rathore AS. Developing cyber-physical system and digital twin for smart manufacturing: methodology and case study of continuous clarification. J Ind Inf Integr 2024; 38:100577.

[3]	Lou S, Hu Z, Zhang Y, Feng Y, Zhou MC, Lv C. Human-cyber-physical system for Industry 5.0: a review from a human-centric perspective. IEEE Trans Autom Sci Eng 2025; 22:494-511.

[4]	Liu S, Zheng P, Bao J. Digital Twin-based manufacturing system: a survey based on a novel reference model. J Intell Manuf 2024; 35(6):2517-46.

[5]	Singh APA, Gameti N. Digital Twins in manufacturing: a survey of current practices and future trends. Int J Sci Res Archive 2024; 13(1):1240-50.

[6]	Liu Y, Zhou Y, Liu Y, Xu Z, He Y. Intelligent fault diagnosis for CNC through the integration of large language models and domain knowledge graphs. Engineering 2025; 53:311-22.

[7]	Theodoridis S, Koutroumbas K. Pattern recognition. 4th ed. New York City: Academic Press; 2006.

[8]	Hu Z, Shukla K, Karniadakis GE, Kawaguchi K. Tackling the curse of dimensionality with physics-informed neural networks. Neural Netw 2024; 176:106369.

[9]	Tolle KM, Tansley DSW, Hey AJG. The fourth paradigm: data-intensive scientific discovery [Point of View]. Proc IEEE 2011; 99(8):1334-37.

[10]	Wang B, Tao F, Fang X, Liu C, Liu Y, Freiheit T. Smart manufacturing and intelligent manufacturing: a comparative review. Engineering 2021; 7(6):738-57.

[11]	Mourtzis D, Varsamis A, Zygomalas S, Angelopoulos J. Adaptive control of 3D printer based on digital twin concept. In: Terzi S, Madani K, Gusikhin O, Panetto H,editors. Innovative intelligent industrial production and logistics. Cham: Springer; 2023. p. 3-17.

[12]	Mourtzis D, Tsoubou S, Angelopoulos J. Robotic cell reliability optimization based on digital twin and predictive maintenance. Electronics 2023; 12(9):1999.

[13]	Bampoula X, Nikolakis N, Alexopoulos K. Condition monitoring and predictive maintenance of assets in manufacturing using LSTM-autoencoders and transformer encoders. Sensors 2024; 24(10):3215.

[14]	Huang D, Yang J, Xu G, Chen J. LuGre-Net: a hybrid neural network for friction modeling of feed systems in machine tools. Nonlinear Dyn 2024; 112(14):11927-43.

[15]	Karabacak YE. Triboinformatic modeling of the friction force and friction coefficient in a cam-follower contact using machine learning algorithms. Tribol Int 2023; 181:108336.

[16]	Rai R, Sahu CK. Driven by data or derived through physics? A review of hybrid physics guided machine learning techniques with cyber-physical system (CPS) focus. IEEE Access 2020; 8:71050-73.

[17]	Aykol M, Gopal CB, Anapolsky A, Herring PK, van Vlijmen B, Berliner MD, et al. Perspective—combining physics and machine learning to predict battery lifetime. J Electrochem Soc 2021; 168(3):030525.

[18]	Arias Chao M, Kulkarni C, Goebel K, Fink O. Fusing physics-based and deep learning models for prognostics. Reliab Eng Syst Saf 2022; 217:107961.

[19]	Li G, Song X, Tian S, Zhu Z. Intelligent drilling and completion: a review. Engineering 2022; 18:33-48.

[20]	Yan J, Liu Z, Leng J, Zhao J, Chen C, Zhang D. Human-centric artificial intelligence towards Industry 5.0: retrospect and prospect. J Ind Inf Integr 2025; 47:100903.

[21]	Leng J, Li R, Xie J, Zhou X, Li X, Liu Q. Federated learning-empowered smart manufacturing and product lifecycle management: a review. Adv Eng Inform 2025;65(Pt A):103179.

[22]	Anitescu C, İsmail Ateş B, Rabczuk T. Physics-informed neural networks:theory and applications. In: Rabczuk T, Bathe KJ,editors. Machine learning in modeling and simulation. Cham: Springer; 2023. p. 179-218.

[23]	Jia X, Zwart J, Sadler J, Appling A, Oliver S, Markstrom S, et al.Physics-guided recurrent graph model for predicting flow and temperature in river networks. In:Proceedings of the 2021 SIAM International Conference on Data Mining (SDM); 2021 Apr 29-May 1. Philadelphia:SIAM; 2021. p. 612-20.

[24]	Wang J, Li Y, Gao RX, Zhang F. Hybrid physics-based and data-driven models for smart manufacturing: modelling, simulation, and explainability. J Manuf Syst 2022; 63:381-91.

[25]	Ma S, Zhang S, Tang T, Yu D, Wang X, Zhang H, et al. A lightweight high-impact acceleration state reconstruction method for multibody dynamic systems by an extended Kalman filter-aided time neural network. IEEE Sens J 2024; 24(19):31524-37.

[26]	Leng J, Zuo K, Xu C, Zhou X, Zheng S, Kang J, et al. Physics-informed machine learning in intelligent manufacturing: a review. J Intell Manuf 2025;in press.

[27]	Jiewu L, Xuyang S, Zean L, et al, Diffusion model-driven smart design and manufacturing: prospects and challenges. J Manuf Syst 2025; 82:561-77.

[28]	Lutter M, Listmann K, Peters J.Deep Lagrangian networks for end-to-end learning of energy-based control for under-actuated systems. In:Proceedings of the 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS); 2019 Nov 3-8; Macau, China. Piscataway: IEEE; 2019. p. 7718-25.

[29]	Zhou T, Gani R, Sundmacher K. Hybrid data-driven and mechanistic modeling approaches for multiscale material and process design. Engineering 2021; 7(9):1231-8.

[30]	Mohammadi A, Krysander M, Jung D. Analysis of grey-box neural network-based residuals for consistency-based fault diagnosis. IFAC-PapersOnLine 2022; 55(6):1-6.

[31]	Li W, He S, Mao X, Li B, Qiu C, Yu J, et al. Multi-agent evolution reinforcement learning method for machining parameters optimization based on bootstrap aggregating graph attention network simulated environment. J Manuf Syst 2023; 67:424-38.

[32]	Yuan Y, Tang X, Zhou W, Pan W, Li X, Zhang HT, et al. Data driven discovery of cyber physical systems. Nat Commun 2019; 10(1):4894.

[33]	Rudy SH, Brunton SL, Proctor JL, Kutz JN. Data-driven discovery of partial differential equations. Sci Adv 2017; 3(4):e1602614.

[34]	Pyle R, Jovanovic N, Subramanian D, Palem KV, Patel AB. Domain-driven models yield better predictions at lower cost than reservoir computers in Lorenz systems. Philos Trans A Math Phys Eng Sci 2021; 379(2194):20200246.

[35]	Liu Q, Liang J, Ma O. A physics-based and data-driven hybrid modeling method for accurately simulating complex contact phenomenon. Multibody Syst Dyn 2020; 50(1):97-117.

[36]

Ahmed HF, Malik FK, Akhtar I, Bangash Z.Application of data-driven and physics-embedded neural networks in wake dominated flows. In:Proceedings of the 2021 International Bhurban Conference on Applied Sciences and Technologies (IBCAST); 2021 Jan 12-16; Islamabad, Pakistan. Piscataway: IEEE; 2021. p. 810-6.

[37]	Jiang Y, Chen J, Zhou H, Yang J, Xu G. Residual learning of the dynamics model for feeding system modelling based on dynamic nonlinear correlate factor analysis. Appl Intell 2021; 51(7):5067-80.

[38]	Mei Z, Ding J, Chen L, Pi T, Mei Z. Hybrid multi-domain analytical and data-driven modeling for feed systems in machine tools. Symmetry 2019; 11(9):1156.

[39]	Hu J, Wang Q, Ye Y, Tang Y. Toward online power system model identification: a deep reinforcement learning approach. IEEE Trans Power Syst 2023; 38(3):2580-93.