Accelerometer prototype with combined filtering for noise attenuation using an embedded system and low-cost MEMS sensors for building monitoring

Obtaining sufficient seismic information is very important for decision making to warn disasters or damage, especially in public buildings. In this work, we propose criteria to design, implement and evaluate the performance of a low cost triaxial accelerometer with MEMs and an embedded system, for buildings monitoring. The prototype is implemented with a Raspberry Pi V3 B+ embedded system, two integrated circuits known as Motion Processing Unit (MPU), based on MEMS, and Wi-Fi communication system to send data to a web server. The whole system is inside of a box and fixed on the ceiling of a building. The noise is filtered using two MPU signals fixed orthogonally and 4th order Band Pass between 0.01 Hz and 25 Hz. The processing performance was tested in a vibrational simulator known as Hexapod, which has programmable vibrational frequency in Hertz and amplitude in millimeter. The evaluation is considering the noise variance, the fundamental frequency of the recorded signal, the magnitude of that frequency, the total harmonic distortion (THD) and the signal-to-noise ratio (SNR) in several experimental tests. The results show that the combination of two MPUs (MEMS) and a fourth-order bandpass filter reduces noise variance by up to approximately 25% and the frequencies of the acquired signals with errors less than 1.0%; In summary, MEMS have efficient performance, according THD and SNR, except for signal frequencies close to 1.0 Hz and displacements close to 1.0 mm.