摘要
油液光谱分析技术可以检测出机械设备润滑油中各种元素的成分和浓度,但光谱分析数据受噪声、补油和换油的影响严重,不能很好地反映出机械设备的磨损情况。对光谱数据进行提升小波分解,通过分析细节信号,检测出野点;用LS-SVM对趋势项建模,对各层细节信息建立AR模型,分别预测出补、换油后及野点处的值并进行叠加;为提高预测精度,用QPSO对LS-SVM的参数进行了优化。结果表明:该方法能有效修正原始数据,从而提高对机械设备磨损状态监测和故障诊断的精度。
Oil spectral analysis technology can be used to analyze the composition and concentration of elements in mechanism equipment oil.But spectrometric data are greatly influenced by noise,oil supply and change,thus the mechanism equipment wear estate can not be well reflected.Spectrometric data were decomposed by lifting wavelet transform and exception data were found through analyzing the detail signals.The model of approximation signal was set up by the method of least square support vector machine （LS-SVM）.AR model to the detail signals was set up.Values which should be forecasted were predicted and added by these two models.In order to improve the accuracy of prediction,LS-SVM was optimized with Quantum-behaved Particle Swarm Optimization （QPSO）.The results show that original signals are effectively amended through these methods,which improve the accuracy of predicition for monitoring wear estate and diagnosing fault of mechanism equipment.
出处
《润滑与密封》
CAS
CSCD
北大核心
2010年第2期59-63,共5页
Lubrication Engineering
基金
国家自然科学基金项目(50705097)
