Performance and Energy Consumption of Smart Sensors for Vibration-Based Anomaly Detection

Anomaly detection in vibration signals is vital for condition monitoring in industrial and structural health applications. This study evaluates cost-effective, energy-efficient smart sensors combining MEMS accelerometers with microcontrollers against a laboratory setup. Two platforms - Arduino Nano 33 BLE and Analog Devices MAX78000 - were assessed for performance, energy consumption, and real-time anomaly detection feasibility. Results indicate that all embedded systems achieve accuracy comparable to the reference setup. Notably, the MAX78000 offers superior inference speed and processing capabilities with a modest energy trade-off. Integrated micro-electromechanical systems (MEMS) and microcontroller solutions prove to be viable, scalable alternatives for on-site condition monitoring.