Abstract: Single or double parameters are adopted in traditional discriminant methods of running condition to distinguish the working condition. In order to improve the identification accuracy, a multi-source information fusion of the motor running condition recognition method is proposed, which aims at the double power sources characteristic of the hybrid cars. The data characteristic information of the decomposed signal in different frequency bands of each sensor is obtained by using single wavelet reconstruction method, so that feature information of different frequencies for each sensor data are obtained. Then different frequencies data feature information from different sensors are conducted by a clustering recognition based on variable weight fuzzy C-average clustering method. Finally, the membership values under different frequencies, belonging to a particular category, are obtained. The final running condition recognition is conducted by the SOM self-organizing map network secondary clustering fusion. In order to improve the vehicle's economy target,the proposed method is applied to torque distribution control of hybrid electric vehicle, and a three-layer feedforward neural network model with different driving conditions with different torque distribution neural network model was assigned. The experimental results verified the effectiveness of the proposed method.

Key words: clustering, Daubechies wavelet, driving cycle, fuzzy c-means, information fusion, SOM network