摘要
在相同状态下各种故障刚出现时和大噪声环境下所收集的信号特征极其相似,导致诊断精准性下降。提出基于深度自编码法和频域相关峭度(FCKT)以实现对轴承运行状况的智能分类。在特征得到加强的前提下,使得数据长度大幅减小,并且在很大程度上增强了算法的识别效率及精准性。研究结果表明:在最大偏移点数取值范畴逐渐扩大的基础上,辨别精准性呈现出不断提高的态势,在该数值为340时,可实现100%的辨别,且有着较强的稳定性。与时域指标相比,频域指标有着更高的辨别精准性。在将FCKT指标当作样本的情况下,稳定性最强、精准性最高。在FCKT计算结束以后、再实施深度自编码智能区分,运行时间减少46.32%,能够在很大程度上减少运行时间。
In the same state,when various faults first appear,the characteristics of the signals collected in the loud noise environment are very similar,leading to the decline of diagnostic accuracy.Based on depth coding and frequency domain correlation kurtosis(FCKT),the intelligent classification of bearing operating conditions is proposed.Under the premise of enhanced features,the data length is significantly reduced,and the recognition efficiency and accuracy of the algorithm are enhanced.The results show that,based on the gradual expansion of the value range of the maximum offset points,the discrimination accuracy presents a trend of continuous improvement.When the value is 340,the discrimination can be 100%,and has a strong stability.Compared with time domain index,frequency domain index has higher discrimination accuracy.When FCKT index is taken as the sample,it has the strongest stability and the highest accuracy.After the end of FCKT calculation,the deep self-coding intelligent differentiation is implemented,and the running time is reduced by 46.32%,which can reduce the running time to a large extent.
作者
史琼艳
杨风波
SHI Qiongyan;YANG Fengho(School of Mechanical Engineering,Changzhou Institute of Mechanical and Electrical Technology,Changzhou Jiangsu 213164,China;College of Mechanical and Electronic Engineering,Nanjing Forestry University,Nanjing 210037,China)
出处
《机械设计与研究》
CSCD
北大核心
2024年第2期143-146,共4页
Machine Design And Research
基金
国家自然科学基金资助项目(51675279)
江苏省自然科学青年基金资助项目(BK20160296)
江苏省教育厅江苏高校“青蓝工程”项目。
关键词
轴承
频域相关峭度
深度自编码器
故障诊断
bearing
frequency domain correlation kurtosis
depth autoencoder
fault diagnosis
