Threshold Selection for Circuit Board Nodes Based on LSTM Method

With the advancement of modern electronic integration technology, the complexity of circuit board composition has significantly increased, making fault detection more challenging. This study proposes a fault diagnosis method based on the current and voltage test values of circuit nodes. By setting up test points on the circuit board, the method determines whether the board is functioning correctly by evaluating if the test values fall within the specified thresholds. Leveraging existing circuit board test data, the study first performs data cleansing, followed by sliding window processing. Long Short-Term Memory (LSTM) networks are then employed to predict the test values for each window, and the 3σ method is used for conditional judgment to derive appropriate thresholds. This approach enhances detection accuracy and the ability to identify subtle anomalies, thereby significantly improving the reliability and precision of circuit board fault detection without damaging the circuit board.