Implementation of Compressed Sensing Technique for Vibration Measurement with a MEMS Accelerometer

Abstract

This research demonstrates the feasibility of utilizing the compressed sensing technique with a low-cost measurement system consisting of a micro-electromechanical systems (MEMS) accelerometer and a microcontroller for vibration measurement. The vibration signals were sampled non-uniformly and processed using the compressed sensing technique. This approach enables the use of a maximum sampling frequency at approximately half the Nyquist rate, while reducing the amount of data by a factor of three to four compared to conventional sampling methods. The study comprised two main parts: simulation and experimental validation. The simulation phase examined the conceptual potential of compressed sensing, while the experimental phase confirmed its practical implementation. Vibration signals from single-piston and two-piston air compressors were measured using the low-cost system. The compressed sensing results were compared with signals simultaneously measured by a conventional system. The results confirmed that compressed sensing effectively reconstructed signals from low-sampling-rate data, closely matching those obtained from conventional measurements. Overall, compressed sensing proved particularly effective for signals with large amplitudes and prominent peaks in the frequency domain. In such cases, the percentage of matched peaks exceeded 80%, while the percentages of unrecovered and false peaks were less than 20%.