A Low-Cost Embedded Inertial Spectrum Analyzer for Machine Condition Monitoring

Abstract

Vibration analysis is a widely used technique for machine condition monitoring, since mechanical faults often introduce characteristic spectral components in inertial signals. This paper presents a low-cost embedded spectrum analyzer for vibration monitoring based on an ADXL345 digital accelerometer and the BitDogLab open-hardware platform, which integrates a Raspberry Pi Pico microcontroller and an onboard OLED display. The proposed system employs SPI communication combined with direct memory access to improve data acquisition efficiency, while spectral analysis is performed locally using Hamming windowing and the fast Fourier transform. The computed spectra can be displayed in real time on the onboard OLED or transmitted to a host computer through USB. As a proof of concept, a vibroacoustic test setup based on an audio processor, power amplifier, and loudspeaker was implemented to generate controlled mechanical excitations. Experimental results show that the embedded system can identify relevant frequency components in sinusoidal, square-wave, and machine-sound-based excitation signals. The results indicate the feasibility of using low-cost open hardware for embedded spectral analysis and motivate further validation in real machine condition-monitoring scenarios.