Low cost wireless sensor network for vibration detection

Abstract

In this work, a real-time method for characterization and analysis of vibration that can be applied in the transport of sensitive loads was proposed, developed and tested. For that, sensors nodes from a wireless sensor network were employed, performing data acquisition and calculating the power spectral density of vibrations. Transmitting only the result of the two most significant points of the PSD to a control node, there was a reduction of 93.4 % of the amount of data transmitted, allowing to increase data acquisition time without battery recharging. Another way to reduce power consumption was to use activity interrupts on the accelerometer to wake up the processor, keeping it in a low power state most of the time. To validate the proposed system, several experiments were performed comparing data with a commercial system, where the average percentage error obtained for PSD amplitude was 41 % for the highest energy frequency and 45 % for the second analyzed frequency. These were high values, but they do not invalidate the proposed system, since the average absolute error was smaller than the resolution in the identification of the frequencies where these peaks occurred. Some changes can be made with the possibility of improving the proposed system, such as performing a system calibration instead of using the values reported by the manufacturer and physically detaching the accelerometer from the sensor node board.