Analysis of discharge parameters affected by electrode motion speed based on deep learning and traditional machine learning

Electrostatic discharge (ESD) is a critical factor affecting the safe operation of electronic devices, and its waveform characteristics are significantly influenced by various parameters, such as the approach speed. To this end, this study utilizes a self-developed electrode movement speed effect tester to collect ESD waveform data and preprocesses the raw data using equal-width binning and various normalization methods. Time-domain, frequency-domain, and morphological features are then extracted. Based on this, multiple regression prediction models are constructed, including traditional machine learning methods, such as Support Vector Machine (SVM), Random Forest, Decision Tree, and K-Nearest Neighbors, and deep learning methods, such as long short-term memory neural network Short-Term Memory Neural Network (LSTM), one-dimensional Convolutional Neural Network, Transformer, and Transformer + LSTM), for multi-parameter prediction of ESD parameters such as peak value and rise rate. Experimental results indicate that, on the dataset normalized using Min-Max, the SVM model performs the best in ESD parameter prediction, with R² values of 0.991 3 and 0.982 9, and significantly lower mean squared error (MSE) and mean absolute error (MAE) compared to other methods. This demonstrates that SVM exhibits higher robustness and generalization ability when handling small sample sizes and high-dimensional data. This study systematically compares various traditional and deep learning models, providing a novel approach to ESD parameter prediction and offering theoretical foundation and practical guidance for the optimization and application of electrostatic protection technology.