基于灰色二次回归的轴温预测模型  

Prediction Model of Axle Temperature Based on Grey Quadratic Regression

作　　者：王超[1] 邹益胜[1] 邓佳林 罗怡澜 WANG Chao;ZOU Yi-sheng;DENG Jia-lin;LUO Yi-lan(Institute of Advanced Design and Manufacturing,School of Mechanical Engineering,Southwest Jiaotong University,Sichuan Chengdu 610031,China)

机构地区：[1]西南交通大学机械工程学院先进设计与制造研究所,四川成都610031

出　　处：《机械设计与制造》2020年第10期37-41,共5页Machinery Design & Manufacture

基　　金：国家重点研发计划—复杂环境下轨道车辆全生命周期能力保持与优化研究(2017YFB1201201)。

摘　　要：轴承温度实时监控和预测是保障高速列车安全运行的重要手段。GM(1,1)模型具有建模样本量小、计算效率和精度高等优点,适用于轴温的实时预测。但在基于GM(1,1)模型的轴温预测中存在两个问题:1.用于建模的轴温监测数据是离散整型,平滑性欠佳,导致预测精度不高;2.由于GM(1,1)模型在本质上是指数函数,具有单调性,导致在轴温升降趋势变化的拐点处预测误差较大。为此,提出一种基于灰色二次回归的轴温实时预测模型:首先将采集到的轴温数据进行迭代三次的滑动平均处理,再将GM(1,1)模型和二次多项式进行融合重构,并采用最小二乘法求取重构后模型的参数值。应用该模型对某高速列车的后序5分钟轴温进行实时预测,结果表明:在轴温先升后降、先降后升和连续波动的样本中该模型比GM(1,1)模型的预测误差分布更集中且数值更小;在不同通道类型的连续波动样本中,这里模型的预测结果均好于GM(1,1)模型,验证了模型的通用性。Real-time monitoring and prediction of bearing temperature is an important way to ensure the safe operation of high-speed trains.The GM(1,1)model has the advantages of small sample size,high computational efficiency and high precision,so it is suitable for real-time prediction of axle temperature.However,there are two problems in the prediction model of axle temperature based on the GM(1,1)model:1.the axis temperature monitoring data is discrete integer,hence the smoothness is poor and the prediction accuracy is not high;2.as a result of The GM(1,1)model is an exponential function,so that there is a large prediction error at the inflection point of axle temperature.For this reason,a real-time prediction model of axle temperature based on grey quadratic regression is proposed.First,the data of axle temperature is processed by three times sliding average;second,the GM(1,1)model and the second polynomial are fused to build a new model;finally,the least squares method was used to obtain the parameter values of the reconstructed model.The prediction results that the next 5 minutes to axle temperature of high speed trains show that:in the sample where the axle temperature first rises and then drops,first drops and then rises and continuous fluctuation,the prediction error distribution of this model is more concentrated and smaller than the GM(1,1)model;among the continuous fluctuation samples with different channel types,the prediction results of this model are better than the GM(1,1)model,which verifies the universality of this model.

关 键 词：轴温 预测 灰色二次回归 灰色理论 高速列车 

分 类 号：TH16[机械工程—机械制造及自动化] U271.91[机械工程—车辆工程]