DKDL-Net: A Lightweight Bearing Fault Detection Model via Decoupled Knowledge Distillation and Low-Rank Adaptation Fine-tuning

TL;DR

DKDL-Net addresses the need for accurate yet ultra-lightweight bearing fault detection by training a compact single-layer CNN via Decoupled Knowledge Distillation (DKD) from a six-layer teacher and further enhancing performance with Low-Rank Adaptation (LoRA) fine-tuning. The approach yields a DKDL-Net with 6,838 trainable parameters, achieving an F1-Score of $99.50\%$ on the CWRU dataset while reducing parameters by about $90.20\%$ relative to the teacher, and it outperforms state-of-the-art methods by roughly $0.58\%$ with competitive inference speed around $1{,}757\mu$s. Key contributions include the combination of DKD with LoRA for efficient knowledge transfer and targeted fine-tuning, enabling robust, real-time fault detection on resource-constrained hardware. The work demonstrates a practical path toward industrial deployment of accurate bearing fault detectors with minimal computational burden.

Abstract

Rolling bearing fault detection has developed rapidly in the field of fault diagnosis technology, and it occupies a very important position in this field. Deep learning-based bearing fault diagnosis models have achieved significant success. At the same time, with the continuous improvement of new signal processing technologies such as Fourier transform, wavelet transform and empirical mode decomposition, the fault diagnosis technology of rolling bearings has also been greatly developed, and it can be said that it has entered a new research stage. However, most of the existing methods are limited to varying degrees in the industrial field. The main ones are fast feature extraction and computational complexity. The key to this paper is to propose a lightweight bearing fault diagnosis model DKDL-Net to solve these challenges. The model is trained on the CWRU data set by decoupling knowledge distillation and low rank adaptive fine tuning. Specifically, we built and trained a teacher model based on a 6-layer neural network with 69,626 trainable parameters, and on this basis, using decoupling knowledge distillation (DKD) and Low-Rank adaptive (LoRA) fine-tuning, we trained the student sag model DKDL-Net, which has only 6838 parameters. Experiments show that DKDL-Net achieves 99.48% accuracy in computational complexity on the test set while maintaining model performance, which is 0.58% higher than the state-of-the-art (SOTA) model, and our model has lower parameters. Our code is available at Github link: https://github.com/SPBU-LiPengyi/DKDL-Net.git.

Figures (6)

• Figure 1: Structure of Rolling Bearings
• Figure 2: Architecture of the Model. (top) Teacher Model, (below) Student Model
• Figure 3: Architecture of the DKDL-Net Model
• Figure 4: Data collection machine tools
• Figure 5: Confusion Matrix for Student Model and DKDL-Net Model on CWRU Dataset