Modeling of GA-CNN degradation process based on phase space-SVM parameter combination

The high-voltage relay is a critical component in electric vehicles, with its reliability having a direct impact on overall vehicle safety and performance. To address modeling inaccuracies resulting from traditional degradation modeling approaches that overlook the nonlinear and high-dimensional characteristics of degradation data, a novel methodology is proposed that integrates phase space reconstruction with support vector machine (SVM) parameter optimization. Wavelet transform is employed to denoise the raw degradation signals and extract key features. The C-C method is employed to estimate the optimal delay time and embedding dimension for phase space reconstruction, thereby recreating the degradation trajectory. Three modeling strategies are employed: a regression-based model, a phase space-based genetic algorithm-convolutional neural network (GA-CNN), and a GA-CNN model incorporating SVM parameter optimization within the phase space framework. The Root Mean Square Error (RMSE) was adopted as the criterion for model evaluation. A specific high-voltage relay was investigated to verify our approach. The results demonstrate that the GA-CNN model, incorporating SVM parameter optimization within the phase space framework, more accurately captures the degradation trajectory of the high-voltage relay.