An interpretable deep learning method for bearing fault diagnosis

TL;DR

This paper presents an interpretable deep learning approach for bearing fault diagnosis using CNNs and Grad-CAM, enabling visualization of feature importance and building a health library to improve trustworthiness in safety-critical maintenance.

Contribution

It introduces a method combining CNN and Grad-CAM for interpretability, creating a health library for sample retrieval, applicable to any CNN without architecture changes.

Findings

The method effectively visualizes feature importance in fault diagnosis.

It improves trustworthiness by providing interpretable prediction basis samples.

Experimental results demonstrate meaningful and physically intuitive sample selection.

Abstract

Deep learning (DL) has gained popularity in recent years as an effective tool for classifying the current health and predicting the future of industrial equipment. However, most DL models have black-box components with an underlying structure that is too complex to be interpreted and explained to human users. This presents significant challenges when deploying these models for safety-critical maintenance tasks, where non-technical personnel often need to have complete trust in the recommendations these models give. To address these challenges, we utilize a convolutional neural network (CNN) with Gradient-weighted Class Activation Mapping (Grad-CAM) activation map visualizations to form an interpretable DL method for classifying bearing faults. After the model training process, we apply Grad-CAM to identify a training sample's feature importance and to form a library of diagnosis knowledge (or health library) containing training samples with annotated feature maps. During the model evaluation process, the proposed approach retrieves prediction basis samples from the health library according to the similarity of the feature importance. The proposed method can be easily applied to any CNN model without modifying the model architecture, and our experimental results show that this method can select prediction basis samples that are intuitively and physically meaningful, improving the model's trustworthiness for human users.