ABIGX: A Unified Framework for eXplainable Fault Detection and Classification
This addresses the need for interpretable fault diagnosis in industrial systems, offering a novel explanation framework that bridges traditional methods, but it is incremental as it builds on prior work like contribution plots and reconstruction-based contributions.
The paper tackles the problem of explainable fault detection and classification by proposing ABIGX, a unified framework that provides variable contributions for general models, and it outperforms existing gradient-based methods in mitigating fault class smearing and shows general superiority in experiments.
For explainable fault detection and classification (FDC), this paper proposes a unified framework, ABIGX (Adversarial fault reconstruction-Based Integrated Gradient eXplanation). ABIGX is derived from the essentials of previous successful fault diagnosis methods, contribution plots (CP) and reconstruction-based contribution (RBC). It is the first explanation framework that provides variable contributions for the general FDC models. The core part of ABIGX is the adversarial fault reconstruction (AFR) method, which rethinks the FR from the perspective of adversarial attack and generalizes to fault classification models with a new fault index. For fault classification, we put forward a new problem of fault class smearing, which intrinsically hinders the correct explanation. We prove that ABIGX effectively mitigates this problem and outperforms the existing gradient-based explanation methods. For fault detection, we theoretically bridge ABIGX with conventional fault diagnosis methods by proving that CP and RBC are the linear specifications of ABIGX. The experiments evaluate the explanations of FDC by quantitative metrics and intuitive illustrations, the results of which show the general superiority of ABIGX to other advanced explanation methods.