摘要
滚动轴承是旋转机械中最主要的部件之一。智能故障检测与诊断作为一种重要的手段,对保证滚动轴承的稳定、可靠和安全运行起着至关重要的作用。然而,由于内部和外部环境的变化,数据分布偏移在实际场景中是不可避免的,在变工况下会导致模型性能严重下降。因此建立一个能够消除相同分布假设的有效故障诊断模型仍然具有挑战性。针对这些问题提出一种基于多表示深度域自适应(MDAN)的滚动轴承故障诊断方法,首先对采集的滚动轴承振动数据进行连续小波变换(CWT),把振动信号由时域信号转换为时频信号,输入到多表示深度域适应网络(MDAN)进行特征提取,通过多表示网络来获得不同尺度信息,将源域特征与目标域特征映射到高维空间,在高维空间进行特征对齐,然后引入MKMMD进行特征距离度量和计算域损失,最后引入Adam进行网络优化,加快模型收敛。在著名的西储大学轴承数据集进行实验验证,通过多个迁移方案对模型进行评估,实验结果证明,所提方法在变工况下具有优异表现,泛化性能力更强。
Rolling bearing is one of the most important parts in rotating machinery.As an important means,intelligent fault detection and diagnosis plays a vital role in ensuring the stable,reliable and safe operation of rolling bearing.However,due to changes in internal and external environments,data distribution deviation is inevitable in actual scenarios,which will lead to serious degradation of model performance under variable working conditions.Therefore,it is still challenging to establish an effective fault diagnosis model that can eliminate the assumption of the same distribution.To solve these problems,a rolling bearing fault diagnosis method based on multi-representation depth domain adaptive(MDAN)was proposed.Firstly,continuous wavelet transform(CWT)was performed on the collected vibration data of rolling bearing.The vibration signal from time domain signal conversion is frequency signal,the input to the said more depth domain adaptive network(MDAN)for feature extraction,with through the said network information for different scales,from the source domain with target domain characteristics is mapped to high-dimensional space,aligned in high-dimensional space characteristics,then introduce the MK-MMD characteristic distance measurement with calculation domain,Finally,Adam was introduced to optimize the network to accelerate the convergence of the model with carry out adaptive verification in the famous bearing dataset of Western Reserve University.The model was evaluated through multiple migration schemes.Experimental results show that the proposed method has excellent performance under variable working conditions with stronger generalization ability.
作者
张文兴
张林林
刘文婧
王建国
ZHANG Wenxing;ZHANG Linlin;LIU Wenjing;WANG Jianguo(School of Mechanical Engineering,Inner Mongolia University of Science and Technology,Inner Mongolia Baotou 014000,China)
出处
《机械设计与制造》
北大核心
2025年第12期258-262,271,共6页
Machinery Design & Manufacture
基金
国家自然科学基金(51865045)
内蒙古自治区自然科学基金项目(2020LH05025)
内蒙古自治区自然科学基金项目(2018ZD06)。
关键词
故障诊断
领域自适应
迁移学习
变工况
Fault Diagnosis
Domain Adaptation
Transfer Learning
Variable Condition
