基于时间序列的动车组跟踪试验自动频谱分析

摘要

对实际运营列车进行实时跟踪试验是高速列车安全性研究不可缺少的重要环节.由于跟踪试验测点多、持续时间长、测试对象复杂、数据种类繁多,需要一种简单、有效、方便的分析方法,对加速度和温度等多种试验数据进行处理和挖掘,准确找出运营列车主要性能参数的变化规律.长期跟踪试验得到的数据具有随机性,可以用时间序列方法分析,时间序列分析的首要问题是建立合适的时序模型.通过对不同时序模型和判定函数进行研究,提出一种自动识别最优时序模型的方法,并根据该模型得到的功率谱密度函数,实现对试验数据的自动频谱分析.利用动车组跟踪试验测得的轴箱加速度数据进行仿真研究,其结果优于传统的周期图法,适合在线故障预报,验证了该方法的正确性.%Real-time tracking test of the trains in operation is vital to the safety of high-speed trains.Due to the large number of tracking test measuring positions, long test duration, complex test objects and a wide variety of data types in tracking test system, a simple, effective and convenient analysis method is needed to process and excavate the test data such as acceleration and temperature data and to find out the the change rules of the main performance parameters of the operating trains.The data obtained from long-term tracking test is stochastic and can be analyzed by time series method.The most important thing of time series analysis is to build an appropriate time-series model.Through the study of different time series models and criterion functions, a method of automatically identifying the optimal time-series model was proposed.The automatic spectrum analysis of the test data was realized based on the power spectrum density function obtained by the model.Finally, the axle box acceleration data measured by the tracking test of EMU was used in a simulation analysis.The result is better than that of the traditional periodogram method.The proposed method is suitable for online fault prediction.The correctness of the method is verified.