What question did this study set out to answer?

To develop a fusion framework that enhances fault detection in home appliances through acoustic and vibration analysis.

April 1, 2026Open Access

Acoustic–Vibration Fusion with Sparse Representation for Intelligent Fault Detection in Home Appliances

Puntos clave

To develop a fusion framework that enhances fault detection in home appliances through acoustic and vibration analysis.
Developed an acoustic-vibration fusion framework using sparse representation.
Integrated contact-based and noncontact sensors for improved data acquisition.
Applied multiscale wavelet decomposition and adaptive feature fusion based on Fisher Discrimination Criterion.
Conducted experimental validation on various home appliances, including a washing machine and refrigerator.
Adaptive fusion improved identification accuracy by 15 to 25% over static methods.
Recognition accuracy increased from 65% to over 85% under low signal-to-noise conditions.
Diagnostic confidence scores ranged from 0.72 to 0.85.

Resumen

The increasing complexity of home appliances necessitates robust fault detection systems to ensure operational safety and reduce maintenance costs.Therefore, we developed an acousticvibration fusion framework based on sparse representation (AVF-SR) to enable intelligent fault monitoring in home appliances.By integrating contact-based vibration sensors (accelerometers) and noncontact acoustic sensors (micro-electro-mechanical system microphones), AVF-SR overcomes the limitations of single-modality systems in noisy household environments.In AVF-SR, we used multiscale wavelet decomposition and an adaptive feature fusion mechanism based on the Fisher Discrimination Criterion to adjust sensor weights according to appliance state.The experimental results on a drum-type washing machine, a split-type air conditioner, and a refrigerator demonstrated that adaptive fusion improves identification accuracy by 15 to 25% compared with those of static weighting methods.SR enhanced recognition accuracy from 65 to over 85% even under low signal-to-noise ratio (-5 dB) conditions.Diagnostic confidence scores ranged from 0.72 to 0.85.The results of this study provide a basis for the development of advanced, lightweight sensors for deployment on low-power microcontrollers in consumer electronics.The limitations of this study related to fault types and laboratory-controlled simulations necessitate further research to enhance model generalizability across diverse appliances and investigate self-supervised learning, which addresses the scarcity of labeled fault data in home appliance deployments.

Me gusta

Guardar

Ver artículo completo