Research And Implementation Of Small Sample Intelligent Fault Diagnosis System With Enhanced Attention Mechanism

With the development of modern industrial technology,the working process of rotating machinery has become more integrated and intelligent.Due to the complexity,harshness,and uncertainty of the working environment,mechanical components will inevitably fail.If faults are not detected early,serious damage will be caused to the equipment,greatly increasing maintenance costs.Therefore,providing effective fault monitoring and health management for mechanical systems plays a crucial role.The deep learning method with multi-level non-linear transformation has been used recently for autonomous mining of statistical and structural relationships between data to establish reliable diagnostic models.Therefore,deep learning methods that can achieve high-dimensional feature information representation have been widely developed.However,the lack of annotated data,noise interference,and changes in operating conditions limit the deep learning method from moving closer to practical applications.This article focuses on the design and implementation of an intelligent fault diagnosis system for complex industrial environments.A multi-branch parallel attention convolution capsule network(WDACN)is designed for the system.To address the challenges of limited labeled data,background noise interference,and varying operating conditions in intelligent fault diagnosis,a multi-branch parallel attention mechanism(MBPAM)is proposed.The multi-branch MBPAM can guide feature representation in multiple directions,pay more attention to segments with more fault information,and effectively suppress noise and redundant information in the input vibration signal.Furthermore,combining MBPAM with the capsule network enhances the model’s ability to extract detailed features,leading to a significant improvement in its classification performance.The diagnostic platform system comprises five modules: registration and login module,user management module,data acquisition module,intelligent diagnostic model construction module,and diagnostic testing module.The data acquisition module leverages sensor technology to capture one-dimensional vibration signals from rotating mechanical components in different health states,which serve as inputs to the end-to-end data-driven model.The intelligent fault diagnosis model construction module relies on the proposed WDACN network structure to achieve fault classification of rotating mechanical components in industrial equipment.Finally,the diagnostic testing module analyzes and diagnoses the signal data and implements visualized push notifications.Finally,the effectiveness of the system is analyzed and validated with the bearing dataset from Case Western Reserve University(CWRU)and the gear dataset from Southeast University(SEU).The experimental results show that the proposed WDACN model performs remarkably well in handling a limited number of signal samples under background noise interference,and has good generalization ability and robustness.The designed intelligent fault diagnosis system has promising practical application prospects.