基于数据驱动的分段供电直线电机故障诊断研究

Data driven fault diagnosis for segment-powered linear induction motor

根据电磁发射系统分段供电直线电机的结构特点指出了这类直线电机的典型故障类型。分析了由于系统结构复杂,采用数据驱动模型而不是物理机理驱动模型进行故障检测和诊断的必要性。针对现有数据集故障样本少,数据集偏倚的问题,研究了采用仿真形成足量故障数据的方法。给出了一种提取直线电机分段通电过程电流包络均值的特征提取方法和数据标准化方法。在试验及仿真数据集上采用这种特征提取方法,应用BP神经网络模型,在故障检测、故障诊断和故障定位三种目标上均取得了较好的效果。在多样化的测试数据集上应用上述方法,验证了该方法的有效性和稳定性。

According to the structrue of segment-powered linear induction motor for electromagnetic emission system,typical fault types of this kind of motor were pointed out.Due to the complexity of structure data-driven model,rather than model based on physical mechanisms,necessity for fault detection and diagnosis was illustrated.Failure data by Simulink was generized to fix data set bias problem caused by the less of fault data.A method of current feature extraction and data set nomalization was presented.By using the method on the data set composed of experimental data and simulation data,BP neural network was applied,and good results have been achieved on fault detection,fault diagnosis and fault location.Effectiveness and stability of the proposed method are verified by the applied method on the diversified test data set.