Srovnání akcelerometrů pro frekvenční analýzu vibrací v prediktivní údržbě

Abstract

This bachelor thesis compares two digital MEMS accelerometers, the ADXL372 and KX134-1211, in terms of their suitability for vibration frequency analysis in predictive maintenance. Drivers, measurement applications communicating via SPI with an NXP MCXN947 microcontroller, and an application for computing a 2048-point FFT using the PowerQuad hardware accelerator were imple- mented for both sensors. Static measurements revealed that the ADXL372 exhibits spurious tones caused by silicon anomaly er003 and a quantization step of 100 mg/LSB. The KX134-1211 does not exhibit such anomalies and its quantization step is approximately fifty times finer (1.953 mg/LSB). A discrete artifact linked to the batch FIFO readout period was identified in this sensor and sup- pressed below the noise level by selecting an appropriate SPI clock configuration. The KX134-1211 is more suitable for vibration frequency analysis in predictive maintenance.