eSysId: Embedded System Identification for Vibration Monitoring at the Extreme Edge

Enabling extreme edge processing functionalities will lead a breakthrough in the development of the next generation of Structural Health Monitoring (SHM) systems, thanks to the adoption of sensor–near data analtycs which will make the structural inference process faster and more advantageous in terms of power consumption and data volume. In this work, we specifically endorse this paradigm in the context of vibration–based diagnostics by proposing a novel, intelligent accelerom- eter sensor combining, in an embedded device, advanced edge data ana- lytics implementing System Identification algorithms, and energy–aware custom hardware supporting it. The effect of the bit–depth quantization of the collected signal on the quality of the retrieved structural param- eters is assessed; moreover, a cost–benefit analysis is also encompassed, showing how the developed solution might be globally more advanta- geous from an energy point of view, reaching up to 10x power saving if compared with standard alternatives.