Abstract

The fuser belt, a critical component in a laser printer's fuser assembly, frequently fails prematurely due to constant exposure to high heat and pressure. To address this, our study proposes a robust fault detection framework utilizing the motor's current signal. This method first employs ensemble empirical mode decomposition (EEMD) to isolate significant low-frequency features by filtering out irrelevant high-frequency noise. Subsequently, a convolutional autoencoder (CAE) reconstructs this filtered signal to identify deviations from normal operation. The framework's superior performance was validated through experiments designed to reflect real-world printing conditions.