Open AccessResearch on construction of vehicle driving cycle based on Markov chain and global K-means clustering algorithm
Author(s) -
Yuanxin Wu,
Guangzhong Liu
Publication year - 2020
Publication title -
vehicle dynamics
Language(s) - English
Resource type - Journals
ISSN - 2529-7767
DOI - 10.18063/vd.v4i1.1135
Subject(s) - markov chain , cluster analysis , algorithm , sequence (biology) , driving cycle , state (computer science) , dimension (graph theory) , automotive industry , computer science , engineering , mathematics , artificial intelligence , machine learning , power (physics) , genetics , physics , electric vehicle , quantum mechanics , pure mathematics , biology , aerospace engineering
Vehicle driving cycle is a time-speed curve used to describe vehicle driving rules. The research and development of vehicle driving cycle not only provide theoretical basis for the test of vehicle fuel’s economy and pollutant emission level, but also guide the design and development of new models in the future. This paper adopts the actual driving data of light vehicles in Fuzhou City, Fujian Province collected by China Automotive Technology Research Center (CATC) through the data collection system, and analyzes and verifies the new data after standardized dimension reduction by combining the global K-means clustering and Markov chain principle. The specific work is divided into the following parts: 1. The global K-means clustering algorithm adopted to cluster the kinematic segment database after standardized dimension reduction; 2. Markov chain is applied to construct the working condition diagram. The basic principle of this method is to regard the short-stroke speed-time sequence as a complete random process, divide the speed intervals by lines, each of which represents a different speed state, and convert the speed into a speed state, so that the speed-time sequence becomes a state-time sequence. Since the next state is only related to the current one, a group of random state sequences can be randomly generated by the program as long as the transition probability between two adjacent states is determined and the matrix of state transition probability is established. 3. The state sequence is converted into a speed sequence, and finally a set of driving cycle conforming to the spatial characteristics of the samples is obtained.