Sensor fusion for simultaneous measurement of micro-vibrations

Abstract

Low-amplitude micro-vibrations are common in nature and engineering, throughout structural applications, and biology. The ability to accurately measure and analyse these vibrations in the presence of noise (unwanted signal content) has far reaching consequences across many fields of acoustics. Methodologies for the enhancement and improvement of such signals are, therefore, sought. We explore the capability of sensor fusion in combination with Kalman filtering (KF), using pairs of accelerometers to improve measurement of low-amplitude micro-vibrations. Prior research of pairing sensor fusion with machine learning approaches like KF, support vector machines, or coherence analysis reported up to 90% reductions in “ghost” detections. This research attempts to extend this success to micro-vibrations where broadband noise, and external perturbations can have dramatic impacts on measurability. A pair of accelerometers has been placed both parallel and perpendicular to the axis of an excitation in pine timber planks of varying dimensions and cuts. Simultaneous measurement of ∼5 N excitations with an automated hammer at varying distances are recorded and patterns observed in the time domain through preliminary analysis in MATLAB. Features identifiable from this data become clearer as compared to conventional approaches and have potential applications in non-invasive early predictive analysis of structures for timber pest control.