【作者】 刘明光；

【导师】 杨江天；

【作者基本信息】 北京交通大学 ， 机械（专业学位）， 2023， 硕士

【摘要】 电机故障预测对降低维修费用,提高机器性能,保证安全可靠满负荷运行有重要作用。本文研究从电机电流信号中提取混合熵作为诊断指标,用组合预测模型描述故障随时间的发展,取得成果如下:电机故障引起电机电流信号中产生新的成分,但是对电流波形的影响很小。因此跟踪电机故障,必须寻找对频谱变化敏感的参数。常规频域参数如:重心频率、频率标准差等,与电机型号、电机转速、信号采样频率等诸多因素有关,难以归一化使用。熵从系统的角度描述时间序列的复杂度,熵值可用来量化特定过程的动态变化,适合作为诊断指标。从电机电流信号中可以提取多种熵,反映电机的运行状况,但目前找不到一种熵函数能完整和充分地描述所有类型的电机故障。为了准确而可靠地预测电机故障,本文提出用主频能量熵和Renyi谱熵构造新的混合熵,分别从频域能量集中程度和频谱分布的均匀程度描述电流信号的变化。充分发挥两种熵函数的优势,能较好的反映不同故障下电机的工况。电机故障发展过程中包含多种趋势,其中既有确定性趋势,又有随机性趋势。为了准确预测故障,本文提出了基于小波变换的组合预测模型。先通过小波变换将时间序列分解成不同的趋势项,分别用背景值优化GM(1,1)模型和ARIMA模型描述其中的增长趋势和随机趋势,再将结果综合。本文提出的组合预测模型结合了两种预测方法的优势,具有预测精度高的优点,有望实现对机车电机故障的中长期预测。工业现场采集不同类型的机车电机电流信号,基于本文提出的方法建立组合模型,预测故障随时间的发展。实验证明本文提出的预测方法具有较高的模型拟合精度和预测精度,高度适用于机车电机剩余使用寿命预测。更多还原

【Abstract】 Motor failure prediction plays a key role in reducing costly unplanned maintenance and improving machine reliability,availability and safety.In this thesis,a novel hybrid entropy is extracted from the motor current signal as a diagnostic index,and the combined prediction model is employed to describe the changes in diagnostic index.The results are as follows:Motor failure causes new components in the motor current signal,but it has little effect on the current waveform.Therefore,the parameters sensitive to the spectrum change should be used to track the motor faults.Conventional frequency-domain parameters such as center of gravity of frequency,standard deviation of frequency,connect with the motor type,rotating speed and signal sampling frequency among many other factors,and not convenient to use.Entropy describes the complexity of time series from the views of systems.Since entropy measures are suitable to quantify such dynamic changes in the underlying process,they can be used as diagnostic index.Many kinds of entropy measurements can be extracted from the motor current signal.and are useful for condition monitoring.However,there is no single measurement can completely and adequately represent the development of any faults or malfunction.In order to predict motor faults accurately and reliably,a novel hybrid entropy,which includes the dominant frequency energy entropy and Renyi spectrum entropy is proposed.The hybrid entropy can describe the change in current signals from the views of concentration degree of frequency-domain energy and the uniformity degree of spectrum distribution systematically.Since the hybrid entropy takes advantage of two entropy measurements,it is highly suitable to quantify the faults development and to determine the machine condition.There are many trends in the development of motor fault,including deterministic trend and stochastic trend.In order to predict faults accurately,a combined prediction model based on wavelet transform is proposed.Firstly,the time series was decomposed into different trend items by wavelet transform,and the growth trend and random trend are described by the background value optimization GM(1,1)model and ARIMA model,and then the results were integrated.Since the combined prediction model takes the advantages of two prediction methods,has higher prediction accuracy and is the most promising tool for mid and long term prediction of motor faults.The motor current signals of different types of locomotive motors are sampled in industrial sites and the proposed approach is evaluated.Experimental results show that the proposed prediction method has high accuracy of model fitting and prediction,and is highly suitable for predicting the remaining service life of locomotive motor.更多还原