摘要
为实现滚动轴承故障智能诊断,提出了一种基于核主元分析法(KPCA)、粗糙集(RS)和遗传算法(GA)改进离散隐马尔科夫模型(DHMM)的智能诊断方法。通过使用混合核函数的KPCA和RS对时域、频域参数进行约简,构造敏感性高、稳定性强,并能准确表征轴承状态的特征参数矩阵。应用GA优化了DHMM,克服了DHMM训练算法容易陷入局部极小的缺点。最后应用GA优化的DHMM训练算法得到的滚动轴承各状态下的DHMM,并通过比较测试样本在各DHMM下的对数似然概率,实现了轴承故障类型的有效识别。实验结果表明,该方法可以有效地识别滚动轴承的状态,具有较强的适用性。
An intelligent diagnosis method for rolling bearing is proposed,which is constructed on the basis of kernel principal component analysis（KPCA）,rough set（RS） and a discrete hidden Markov model（DHMM） optimized by genetic algorithm（GA）.KPCA-RS is used to optimize the time domain and frequency domain characteristic parameters and construct a more sensitive and stable characteristic parameters matrix which can accurately characterize the bearing condition.Additionally,DHMM is optimized by GA to solve the problem that DHMM training algorithm is easy to fall into local minimum.The DHMM based on GA can be obtained under each rolling bearing state.The fault state can be identified by calculating and comparing the logarithmic likelihood probability of each DHMM of testing data.The experimental results show that the method can effectively identify the rolling bearing fault state with high recognition accuracy.
出处
《测控技术》
CSCD
北大核心
2014年第11期21-24,28,共5页
Measurement & Control Technology
基金
国家自然科学基金资助项目(51375037
51075023)
国家973项目(2012CB026000)
教育部新世纪优秀人才支持计划(NCET-12-0759)
