基于因果推理的智能控制系统故障溯源与自愈策略

Fault Traceability and Self-healing Strategies for Intelligent Control Systems Based on Causal Inference

摘要

智能控制系统故障难溯源、自愈欠精准，本文融合因果推理与生物免疫启发提出一种一体化自主健康管理框架。从溯源层面上构建了基于时序因果发现的可解释模型（ExCTM），能够精准定位故障根本原因。在自愈层面上设计了因果知识引导的生物免疫启发式强化学习方法（BIH-RL），建立了一套高效、特异性的恢复策略。在田纳西-伊士曼过程和机械臂系统上的实验表明，能能显著提升根因定位准确率和自愈效率，为构建具备认知自愈能力的下一代智能系统提供了新途径。

Abstract

To address the core challenges of difficult fault tracing and imprecise self-healing in intelligent control systems, this paper proposes an integrated autonomous health management framework that combines causal inference with bio-immune inspiration. At the tracing level, an interpretable model based on temporal causal discovery (ExCTM) is constructed to achieve precise localization of root causes. At the self-healing level, a causal-knowledge-guided, bio-immune-inspired reinforcement learning method (BIH-RL) is designed to generate efficient and specific recovery strategies. Experiments on the Tennessee-Eastman process and a robotic arm system demonstrate that this framework significantly improves root cause localization accuracy and self-healing efficiency, providing a new pathway for building next-generation intelligent systems with “cognitive self-healing” capabilities.

关键词

因果推理 / 故障溯源 / 自愈控制 / 强化学习 / 智能系统

Key words

Causal Inference / Fault Root Cause Tracing / Self-Healing Control / Reinforcement Learning / Intelligent Systems

参考文献

[1] PEARL J. Causality: Models, Reasoning, and Inference[M]. 2nd ed. Cambridge: Cambridge University Press, 2009.

[2] ZHENG X, DAN C, ARAGAM B, et al. Learning Sparse Nonparametric DAGs[C]// Proceedings of the 23rd International Conference on Artificial Intelligence and Statistics. 2020: 3414-3425.

[3] 张化光, 王迎春. 智能故障诊断与自愈控制[M]. 北京: 科学出版社, 2018.

[4] LEVINE S, FINN C, DARRELL T, et al. End-to-End Training of Deep Visuomotor Policies[J]. The Journal of Machine Learning Research, 2016, 17(1): 1334-1373.

[5] HAO J, CHEN H, ZHANG R, et al. Biologically Inspired Self-Healing Control Systems: A Review[J]. IEEE Transactions on Neural Networks and Learning Systems, 2022, 33(12): 6987-7001.

[6] 刘强, 周东华. 数据驱动的工业过程故障诊断研究综述[J]. 自动化学报, 2016, 42(9): 1285-1299.