Search

To address the issues in reliability bench tests for drive motor system of electric vehicle, particularly low correlation between reliability test conditions and real-world user driving scenarios, as well as non-equivalence in damage effects, a method for constructing reliability test cycles based on user driving models is proposed. Firstly, driving scenarios are analyzed and actual driving characteristics of users are identified by utilizing big data from real-world driving. Secondly, K-means clustering analysis is conducted to establish typical operating conditions, from which segments with unit damage intensity ranging from the 95th percentile to the 99th percentile are selected into an alternative library. Finally, reliability test cycles are constructed through a genetic algorithm and Markov process. The results of damage analysis show that the operating conditions constructed for reliability test maintain the typical driving characteristics of users and demonstrate a good acceleration test effect. When the motor product is continuously loaded on the test bench for 1625.7h and 1302.1h respectively, the shafting and gear train are considered achieving the users' damage expectation, equivalent to a whole life cycle of 300,000 km.