To address the issues of slow convergence, susceptibility to local optima, and limited evaluation metrics in identifying and localizing acceleration faults in hybrid electric vehicles (HEVs), this study proposes a fault diagnosis method based on an improved genetic algorithm (IGA) optimized least squares support vector machine (LSSVM). The proposed IGA introduces adaptive crossover and mutation mechanisms to enhance search efficiency and employs the F1-score as the fitness evaluation metric, thereby improving robustness in recognizing imbalanced multi-class fault data. The experimental results demonstrate that the proposed IGA-LSSVM model achieves an overall diagnostic accuracy of 96.28% on the test set. Compared with the GA-LSSVM model, the accuracy is improved by approximately 6%, and the F1-score is increased to 0.95. These results verify the effectiveness and practicality of the proposed method for multi-class fault diagnosis tasks.
Ji et al. (Tue,) studied this question.