摘要
轴承健康状况直接影响着机械设备的稳定性和安全性,对轴承的运行状态进行故障诊断尤为重要。基于此,在健康、内圈故障和外圈故障3种不同状况下,选择莱文贝格-马夸特(Levenberg-Marquardt,LM)算法、贝叶斯正则化(Bayesian Regularization,BR)算法和量化共轭梯度(Quantum Conjugate Gradient,QCG)算法,对在随时间变化的加速条件下滚动轴承振动数据进行训练和测试。在MATLAB R2023b软件中构建不同类型的深度学习模型,对比分析深度学习模型的均方误差值、回归R值、训练时长和训练轮数等多种指标。经过分析得出,在追求精度和准确性、内存资源和时间充足的情况下,应选用贝叶斯正则化法算法来训练深度学习网络模型。
Bearing health condition directly affects the stability and safety of machinery,fault diagnos is of bearing running state is particularly important.Based on this,Levenberg-Marquardt(LM)algorithm,Bayesian Regularization(BR)algorithm and Quantum Conjugate Gradient(QCG)algorithm are selected to train and test the vibration data of rolling bearings under the condition of time varying acceleration under three different conditions:health,inner ring fault and outer ring fault.Different types of deep learning models are constructed in the MATLAB R2023b software,and the mean square error value,regression R value,training time and training rounds of the deep learning model are compared and analyzed.After analysis,it is concluded that the Bayesian regularization algorithm should be selected to train the deep learning network model when the precision and accuracy,memory resources and time are sufficient.
作者
王睿川
胡一飞
WANG Ruichuan;HU Yifei(School of Mechanical and Power Engineering,Nanjing Tech University,Nanjing 211816)
出处
《现代制造技术与装备》
2024年第3期108-111,共4页
Modern Manufacturing Technology and Equipment
