Analog Circuit Fault Classification Detection Based on Adam-CNN Models

Analog electronic circuits are integral to the functionality of complex electronic devices. The automatic diagnosis of circuit faults is critical to prevent potential shutdowns and significant property losses. This paper presents a novel Adam-CNN model designed for soft fault detection in analog circuits. The model combines the robust feature extraction capabilities of the CNN model with the adaptive learning of the Adam gradient descent algorithm. Training on peak values of the frequency domain response and the maximum and minimum values of the transient time domain response of the Circuit Under Test (CUT), the proposed model aims to classify faults effectively. Validation experiments using a Sallen-Key bandpass filter circuit, with variations in resistor or capacitor values around 50% of their nominal values, showcase a fault detection accuracy of 99.3%. This highlights the superior performance of the proposed method compared to existing research, emphasizing its potential in enhancing the reliability of analog electronic circuits.