基于PIPs样本优化的机车车辆轴温实时预测方法  

A Real-Time Temperature Prediction Method Based on PIPs Sample Optimization for Train Bearings

作　　者：杨则云 YANG Zeyun(CRRC QINGDAO SIFANG CO., LTD., Qingdao 266111, China)

机构地区：[1]中车青岛四方机车车辆股份有限公司,山东青岛266111

出　　处：《机械》2018年第4期1-5,10,共6页Machinery

基　　金：国家高新技术研究发展计划(863)计划(2015AA043701)

摘　　要：建立了一种基于最小二乘曲线拟合的轴温预测模型,并采用PIPs对样本进行优化改进,提高了预测精度。首先基于PIPs对建模样本点进行优化选择,选取历史温度数据中能表征温度变化趋势的重要点作为建模样本点,再利用最小二乘曲线拟合法建立实时温度预测模型对机车车辆轴承温度进行实时预测。以某型机车车辆履历轴承温度数据为例,采用本文模型对任意时刻温度的后5分钟温度进行预测,将预测结果与实际监测结果进行连续对比,验证了所建立的轴温预测模型及其改进模型的有效性:基于最小二乘曲线拟合预测模型的综合平均相对误差为3.47%,综合最大相对误差为20.27%,而进行PIPs改进后的综合平均相对误差为2.67%,同比降低了23.05%,综合最大相对误差为16.67%,同比降低了17.76%。In this paper, a bearing temperature prediction model based on the least square curve fitting is established, and the sample is optimized by using the perceptually important points to improve the prediction accuracy. First, the modeling sample points were selected from historical temperature data which can represent the important points of the temperature change trend based on the PIPs, then least squares curve fitting was used to establish a real-time temperature prediction model of train bearing temperature real-time prediction. According to a train bearing temperature data, use this model to predict temperature in the next five minutes. Compared with the actual measurement results, and it shows that the proposed temperature prediction model is valid： the total average error of the prediction model based on the least squares curve fitting is 3.47%, the total maximum relative error is 20.27%. After the improvement of PIPs, the total average prediction error is 2.67%, reduces 23.05% compared with before. The total maximum relative error is 16.67%, reduces 17.76% compared with before.

关 键 词：轴承温度 时间序列 曲线拟合 PIPs 

分 类 号：TK730.322[交通运输工程—轮机工程] TF068.21[动力工程及工程热物理—流体机械及工程]