2025 International Conference on Sensing, Measurement & Data Analytics in the era of Artificial Intelligence (ICSMD 2025)

Traditional motor fault diagnosis based on digital signal processing relies on analog-to-digital conversion and digital computation, facing issues of heavy computational load and high energy consumption. Inspired by decision trees, this paper proposes a novel pure analog circuit classifier to achieve near-zero-delay fault diagnosis. The core design is that hardware circuits directly process analog signals from accelerometers. An analog filter decision tree is constructed, and the filters cutoff frequency parameters are adjusted through an optimization algorithm to match circuit characteristics, directly extracting features and conducting classification in the analog domain. Results are output as real-time waveforms. The significant advantages of this scheme are completely avoiding the analog-to-digital conversion process, featuring zero-delay response, an efficient circuit structure, and retaining the interpretability of analog circuits. Experiments have confirmed that this method can accurately output diagnostic results, providing a convenient and highly real-time solution for intelligent maintenance of motors and rotating machinery.