CS-SHAP: Extending SHAP to Cyclic-Spectral Domain for Better Interpretability of Intelligent Fault Diagnosis

2025-02-10

AI Summary

To address the limited interpretability of neural network decisions in intelligent mechanical fault diagnosis, this paper proposes CS-SHAP—the first interpretable method operating directly in the cyclic spectral domain. CS-SHAP extends SHAP value theory to the cyclic spectrum for the first time, enabling direct quantification of feature contributions across both carrier and modulation frequency dimensions without model modification or signal preprocessing, thereby preserving end-to-end learning while strictly adhering to the underlying physical mechanisms of mechanical faults. By integrating frequency-domain feature attribution with multi-source vibration signal modeling, CS-SHAP achieves significant performance gains over existing SHAP variants on three public benchmark datasets, simultaneously improving both explanation clarity and attribution accuracy. The implementation is publicly available and demonstrates strong cross-task transferability.

Technology Category

• Machine Learning: Feature Construction/Reformulation
• Computer Vision: Interpretability, Explainability, and Transparency
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Neural networks (NNs), with their powerful nonlinear mapping and end-to-end capabilities, are widely applied in mechanical intelligent fault diagnosis (IFD). However, as typical black-box models, they pose challenges in understanding their decision basis and logic, limiting their deployment in high-reliability scenarios. Hence, various methods have been proposed to enhance the interpretability of IFD. Among these, post-hoc approaches can provide explanations without changing model architecture, preserving its flexibility and scalability. However, existing post-hoc methods often suffer from limitations in explanation forms. They either require preprocessing that disrupts the end-to-end nature or overlook fault mechanisms, leading to suboptimal explanations. To address these issues, we derived the cyclic-spectral (CS) transform and proposed the CS-SHAP by extending Shapley additive explanations (SHAP) to the CS domain. CS-SHAP can evaluate contributions from both carrier and modulation frequencies, aligning more closely with fault mechanisms and delivering clearer and more accurate explanations. Three datasets are utilized to validate the superior interpretability of CS-SHAP, ensuring its correctness, reproducibility, and practical performance. With open-source code and outstanding interpretability, CS-SHAP has the potential to be widely adopted and become the post-hoc interpretability benchmark in IFD, even in other classification tasks. The code is available on https://github.com/ChenQian0618/CS-SHAP.

Problem

• Enhance interpretability of fault diagnosis models
• Extend SHAP to cyclic-spectral domain for better insights
• Align explanations with fault mechanisms for accuracy

Innovation

• Extends SHAP to cyclic-spectral domain
• Evaluates carrier and modulation frequencies
• Enhances interpretability in fault diagnosis