Underrepresentation of driving cycles is an important reason for the difference in actual fuel consumption and standards. In previous studies, The amount of data for developing driving cycle are too small and the clustering method used cannot satisfy the increasingly complex samples. Therefore, this paper aims at the development of driving cycles, combines with the data processing methods of wavelet transform and principal component analysis, to analyze the main characteristic factors of vehicle kinematic sequences. The kinematic sequences model, self-organizing feature Map (SOM) and fuzzy C means (FCM) clustering analysis were used to establish a mathematical model for dividing kinematic sequences with different motion characteristics into different sample sets, and developing driving cycles of 1300, 1259, 1299s respectively. Through the selection of 15 main features of the vehicle motion characteristics of the evaluation system, a variety of algorithms under the rationality of operating conditions are compared and verified. Experimental results show that the accuracy of reconstruction using wavelet transform compression is normalized to the RMS value of 13.25, the cumulative error of SOM algorithms and FCM clustering analysis is 134.1% and 124.7%. The results of rationality verification show that FCM clustering model is superior to others. It provides the support of model algorithm for the development of driving cycles.

CICTP