基于多尺度可扩张卷积和DMWT-Mamba的小样本机械故障诊断OA

Research of intelligent diagnosis technology for mechanical failure can ensure the safe and stable operation of equipment.However,in industrial production,it is difficult to obtain large number of high-quality labeled data samples and avoid the influence of noise when collecting the vibration signals.Therefore,this paper proposes a small-sample mechanical fault diagnosis model based on multiscale dilatable convolution and DMWT-Mamba.Firstly,a multiscale dilatable convolution block is designed to extract multiple local receptive field features from vibration signals,which can significantly reduce the number of learning parameters and computational volume.Secondly,the discrete multi-wavelet transform is combined with Mamba to dynamically select important time-step information,ignore irrelevant noise interference,and extract key information while fully fusing features in each frequency band component,thereby enhancing the model's anti-jamming performance and feature extraction ability under small sample conditions.Finally,experiments are conducted using two sets of mechanical failure datasets.The results show that the model can effectively improve the accuracy of fault diagnosis with small samples and has strong anti-interference capability.