Self-Calibrating Multiresonance Piezoelectric MEMS Sensor With Adaptive Learning for Biofuel Viscosity and Density Monitoring
Year: 2026
Type: preprint
Abstract: This paper presents a self-calibrating multiresonance piezoelectric MEMS sensor with adaptive deep learning for robust biofuel viscosity and density monitoring. The sensor exploits broadband spectral responses, with dominant resonances near 260, 320, and 545 kHz, to capture composition-dependent variations beyond a single resonance mode. Experimental characterization achieves normalized sensitivity, resolution, recovery, \(\:m\left(1\right)\), and \(\:m\left(2\right)\)values of 0.19, 0.135, 0.16, 0.26, and 0.18, respectively. An adaptive spectral-learning framework integrates multiresonance features with operating-condition information to mitigate temperature-induced variation, sensor drift, and measurement disturbances. The proposed system achieves viscosity and density RMSEs below 0.035 mPa·s and \(\:4.5\times\:{10}^{-4}\)g/mL, respectively, with corresponding \(\:{R}^{2}>0.985\), while limiting prediction degradation to < 5% over 30 days and reducing recalibration requirements by ≥ 50%.
Source: Research Square
Author: Md Abdulla Hasan
Institution: Daffodil International University
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225992399
