【作者】 贾健；

【导师】 张淑清；

【作者基本信息】 燕山大学 ， 检测技术与自动化装置， 2010， 硕士

【摘要】 当前,工矿企业大型设备的数目越来越多,关键设备的检测与诊断技术所带来的社会效益和经济效益,也不断为人们所认识。因此,预防维修到预知维修的转变也成为必然,顺应了机械设备故障诊断的趋势。对于时间序列,利用时间序列的相空间重构可以保存系统的某些性质,在此基础上可以建立动力系统的非线性模型并进行预测。对于机械系统状态的预测,就是根据已知的某时刻以前机械系统的状态运动轨迹,预测该时刻以后最大可预测尺度内轨迹的运动状态。论文首先分析了混沌的复杂特性以及其他混沌识别方法的利弊。针对所作分析,提出了一种混沌相空间重构的方法:即运用互信息函数方法确定有效延迟时间;CAO方法确定最佳嵌入维数。然后根据相空间重构所确定的混沌时间序列的延迟时间以及嵌入维数,运用Volterra级数算法推算出未来一定时间内的时间序列,以达到对设备未来故障状态预测分析的目的。为解决强噪声环境中的大型机械故障的早期信号微弱的问题,运用了Duffing振子的间歇混沌现象对强噪声背景中的微弱信号进行检测,并将其应用在轧机设备的齿轮和转子的故障检测中,同时对实验数据及结果进行分析。最后,将整套理论方法应用到减速机的故障状态预测中:将所预测的时间序列运用Duffing振子来提取特征信号,并通过RBF神经网络进行推理决策,给出故障诊断结果,从而实现了对机械故障状态的预测分析。更多还原

【Abstract】 At present, more and more industry and mine corporations are constructed, the social and economic benefit bright out by the detection and diagnosis of the key equipment are more considered gradually. The change from prevention maintenance to prescient maintenance will be inevitable, it conforms the trend of machine fault’s diagnosis.For time series, the phase reconstruction can recover some of the system’s characteristics. Basing on it, nonlinear model of the power system can be created in order to do some prediction. Relying on the state of machine system before the moment that have been slected, the movement state after the moment can be predicted in the bound of the max prediction.First, chaos’s complex characteristic and the method of identifying them are analysed. Depending on these analysis, a new method on the phase reconstruction of chaos is proposed: the appropriate delay time is determined using the mutual information function; the best embedding dimension is determined using CAO method.Second, the time series are calculated in an allowable bound using the Volterra progression, relying on the delay time and the embedding dimension through the method of phase reconstruction, so the prediction on the fault of equipment will be completed.Third, in order to solute the question on the weak fault signal of large machine, intermittent chaos of the Duffing oscillator is used to detect the weak fault signal buried in the strong noise surroundings. The method is applied to detect the fault of the rolling mill’s rotors and gears and analysing the experiment result.At last, all theories are applied to predict the fault of speed-down machine: the characteristic signal contained in the predicted time series is detected by Duffing oscillator and the dignosis result is given using RBF nerve network. The prediction and analysis to machine fault state are completed.更多还原