An Open-Source Industrial Digital Twin Platform for Explainable Predictive Maintenance: Architecture and Single-Node Performance Characterisation

We have described an industrial digital twin and predictive-maintenance platform built entirely from open components, covering acquisition over standard industrial protocols, separated stores for current state and history, an inference path whose explanations are audited against their own numbers, risk-aware scheduling by constraint solving, and an actuation model that binds natural-language commands to explicit risk tiers. The system runs as fourteen services on one commodity machine from a single command. Measured on that machine, the platform holds 3.66 GiB resident, sustains 80.7 readings per second, delivers live updates in 74 ms at the median, stores readings at 11.91 bytes each under a compression ratio of 15.2:1, answers fleet-wide queries 90 times faster from continuous aggregates than from raw history, and passes 450 automated tests. The memory distribution identifies the clustered twin store, at 54.8 per cent of the total, as the component to reconsider for constrained deployments. Four directions follow. The prediction path should be validated against instrumented machines, using the simulator's labelled output only for pre-training. The twin store should be compared against a lighter current-state representation under the same measurement protocol. The explanation audit should be evaluated with technicians rather than by construction, measuring whether an audited narration changes what a maintenance decision looks like. Finally, moving inference to the edge would reduce the volume crossing the network and allow the architecture to be measured under intermittent connectivity.

[1]               W. Kritzinger, M. Karner, G. Traar, J. Henjes, and W. Sihn, “Digital twin in manufacturing: A categorical literature review and classification,” IFAC-PapersOnLine, vol. 51, no. 11, pp. 1016–1022,2018. [2]               F. Tao, H. Zhang, A. Liu, and A. Y. C. Nee, “Digital twin in industry: State-of-the-art,” IEEE Transactions on Industrial Informatics, vol. 15, no. 4, pp. 2405–2415, 2019. [3]               J. Lee, F. Wu, W. Zhao, M. Ghaffari, L. Liao, and D. Siegel, “Prognostics and health management design for rotary machinery systems— Reviews, methodology and applications,” Mechanical Systems and Signal Processing, vol. 42, no. 1–2, pp. 314–334, 2014. [4]               T. Zonta, C. A. da Costa, R. da Rosa Righi, M. J. de Lima, E.S. da Trindade, and G. P. Li, “Predictive maintenance in the Industry 4.0: A systematic literature review,” Computers and Industrial Engineering, vol. 150, art. 106889, 2020. [5]               A. Saxena, K. Goebel, D. Simon, and N. Eklund, “Damage propagation modeling for aircraft engine run-to-failure simulation,” in Proc. International Conference on Prognostics and Health Management, 2008, pp. 1–9. [6]               F. T. Liu, K. M. Ting, and Z.-H. Zhou, “Isolation forest,” in Proc. 8th IEEE International Conference on Data Mining, 2008, pp. 413–422. [7]               G. Ke, Q. Meng, T. Finley, T. Wang, W. Chen, W. Ma, Q. Ye, and T.-Y. Liu, “LightGBM: A highly efficient gradient boosting decision tree,” in Advances in Neural Information Processing Systems 30, 2017,pp. 3146–3154. [8]               S. M. Lundberg and S.-I. Lee, “A unified approach to interpreting model predictions,” in Advances in Neural Information Processing Systems 30, 2017, pp. 4765–4774. [9]               M. T. Ribeiro, S. Singh, and C. Guestrin, “Why should I trust you? Explaining the predictions of any classifier,” in Proc. 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016, pp. 1135–1144. [10]             H. Nori, S. Jenkins, P. Koch, and R. Caruana, “InterpretML: A unified framework for machine learning interpretability,” arXiv:1909.09223, 2019.