论文部分内容阅读
为了处理车辆轴温可能出现的跳变、缺失、噪声等异常数据,有效降低误报率,提出了基于动态时间规整算法的车辆轴温状态监测方法,将轴温历史监测数据和历史统计数据进行指数平滑预处理,在训练阶段将数据反复迭代得到不同轴温模式的参考样本,计算了实时轴温和参考样本各数据帧之间的欧氏距离,得到帧匹配距离矩阵,运用动态规划和回溯的思想,求出累积距离矩阵和动态规整路径,将动态规整距离作为2个时间序列相似度的量化指标,找出最小动态规整距离对应的轴温模式,从而得到状态监测结果。仿真结果表明:在MATLAB仿真中,输入1 000个时长为50~300min的轴温测试样本,其最大响应时间小于0.4s,共出现29次错误匹配,误报率低于3%。通过对测试样本和参考样本的各数据帧进行指数平滑处理,有效消除车辆轴温出现跳变的干扰,虽然跳变值和跳变点数量不同,但相对动态规整距离无变化,对状态监测结果无影响。可见,车辆轴温状态监测方法能够满足车辆轴温状态监测的实时性和准确度要求,减少了误报率。
In order to deal with abnormal data such as jumping, missing and noise that may appear in the axle temperature of the vehicle and effectively reduce the false alarm rate, a method for monitoring the axle temperature of the vehicle based on the dynamic time warping algorithm is proposed. The historical data of the axle temperature and the historical statistics Exponential smoothing preprocessing, the data are iteratively iterated during the training phase to get the reference samples of different axial temperature modes, and the Euclidean distance between each data frame of the real-time axis mild reference sample is calculated to obtain the frame matching distance matrix, and the dynamic programming and backtracking , The cumulative distance matrix and dynamic regularization path are obtained, and the dynamic regular distance is used as the quantitative index of two time series similarity to find the shaft temperature mode corresponding to the minimum dynamic regular distance, so as to obtain the state monitoring result. The simulation results show that in the MATLAB simulation, the input of 1000 shaft temperature test samples with the time length of 50 ~ 300min, the maximum response time is less than 0.4s, a total of 29 error matching occurs, the false alarm rate is less than 3%. Exponential smoothing of each data frame of the test sample and the reference sample can effectively eliminate the interference of the vehicle axis temperature jump, although the number of jump values and the number of trip points are different, but the relative dynamic regularization distance does not change, no effect. Visible, vehicle temperature monitoring method to meet the state of the vehicle temperature monitoring of the real-time and accuracy requirements, reducing the false alarm rate.