Abstract: In view of the large difference in fuel consumption and other performance between the truck regulation working condition and the actual working condition, the vehicle networking platform was used to obtain the real vehicle driving data of the target sample vehicle, and the construction method of the actual working condition of the truck based on Gaussian mixture model clustering and Markov chain was proposed. The kinematic segments were classified by principal component analysis and Gaussian mixture model cluster analysis. In the class, the driving state was divided by the speed interval, and the Laplace Smoothing method was used to estimate the state transition probability. The Markov chain method was used to synthesize the candidate working conditions in the class, and the best working conditions in the class were selected to construct the actual driving conditions of the target truck. The results show that compared with the original data, the joint probability density distribution of velocity-acceleration of the constructed working condition is in good agreement, and the average relative error of the overall characteristic parameters is 4.32%. Compared with the measured fuel consumption per 100 km, the relative errors of the simulated fuel consumption under regulation and constructed working conditions are 26.33% and 4.86%, respectively. The constructed working condition has higher accuracy and can effectively reflect the actual driving characteristics and fuel consumption performance of the vehicle, which can provide conditions and basis for the refined design of the vehicle and the vehicle performance evaluation on the actual road.

Key words: vehicle engineering; trucks; performance evaluation; Gaussian mixture model; Markov chain; actual working condition

摘要： 针对货车法规工况与实际使用工况的油耗等性能相差较大问题，利用车联网平台获取目标样车实车行驶数据，提出基于高斯混合模型聚类和马尔可夫链结合的货车实际工况构建方法。通过主成分分析、高斯混合模型聚类分析将运动学片段分类，类中以速度区间划分行驶状态，采用Laplace Smoothing方法来估计状态转移概率，利用马尔可夫链法合成类内候选工况，选择类内最佳工况，构建目标货车实际行驶工况。结果表明：与原始数据相比，所建工况速度-加速度联合概率密度分布吻合较好，整体特征参数平均相对误差为4.32%；与实测百公里油耗相比，法规工况和所建工况的仿真油耗相对误差分别为26.33%、4.86%。所建工况精度较高，能够有效反映车辆实际行驶特征及油耗性能，可以为车辆的精细化设计和实际道路整车性能评价提供条件和依据。

关键词: 车辆工程；货车；性能评价；高斯混合模型；马尔可夫链；实际工况