汽车运行工况的构建方法

A Method for Constructing Automobile Operation Condition

为快速准确地构建汽车运行工况,使用行车记录仪采集设备记录了上海市某辆轻型汽车的行驶数据,连续3次采集了包括3个不同时间段的信息,采样频率为1 Hz。为研究汽车行驶工况,首先使用平滑处理、剔除、划归滑窗法对庞大的汽车运行工况数据进行了预处理,将724处时间不连续的情况进行了调查,分为3类进行不同处理。采用移动平均滤波器Smooth函数进行了平滑处理,以消除速度异常值。将长期停车进行怠速处理后,使用滑窗法处理了怠速数据。然后对运动学片段进行切取并提取特征值,将13个特征降维得到4个主成分。接着在轮廓系数法等3种方法选择类别的基础上,应用K-means聚类分析法,不断验证确定将数据集划分成了3类最优情况,分别为走走停停、高速行驶和低速行驶,与现实情况基本吻合。最后根据每类的时间占比,从中选取最具代表性的15个片段,构建出了时长为1254 s的汽车运行工况,得到的结果误差很低,误差率不超过10%。此汽车运行工况的构建模型对汽车运行工况数据处理过程十分细致,且能在初始聚类中心的基础上不断寻优。

In order to quickly and accurately construct the operation condition of vehicle,we recorded the driving data of a light vehicle in Shanghai by driving recorder collection equipment,and collected the information of 3 time periods for 3 consecutive times,with the sampling frequency of 1 Hz.In order to study the driving condition of the vehicle,first,we preprocessed the huge data of vehicle operating conditions by using smooth processing,elimination,attribution and sliding window method.We investigated 724 time discontinuities and divided them into 3 categories for different treatments.We conducted smooth processing by using smoothing function of the moving average filter to eliminate the abnormal speed values.After idle processing the long-term parking,we processed the idling data by using the sliding window method.Then,we cut the kinematic segments and extracted the feature values,and reduced the dimensionalities of 13 features to obtain 4 principal components.Afterwards,on the basis of selecting categories by 3 methods including contour coefficient method,by applying K-means cluster analysis method,we continuously verified and determined to divide the data set into 3 categories,namely,stop and go,high-speed driving,and lowspeed driving,which are basically consistent with the reality.Finally,according to the time proportion of each category,we selected the most representative 15 segments and constructed the vehicle operating condition with a duration of 1254 s,and the error obtained is very low,with an error rate not exceeding 10%.The constructed model of the vehicle operation condition is very detailed for the data processing process of the vehicle operation condition,and can continuously seek optimization based on the initial clustering center.