Low-cost MEMS accelerometer network for rotating machine vibration diagnostics

In this paper a wireless sensor network (WSN) with low-cost nodes (less than $US 30.00) is presented. The nodes are based on microelectromechanical systems (MEMS) accelerometers and a highly-integrated microcontroller with built-in antenna for Wi-Fi and Bluetooth Low-Energy (BLE). The system was specially designed to analyze unbalance, load, and rotor obstructions in a rotating machine. Two sensor nodes were installed on an apparatus: one in the shaft and the other on the support table. The acceleration signals were used to analyze the machine vibration in the frequency domain by its Fast Fourier Transform (FFT). This spectrum was pre-analyzed on the sensor node, and the most significant features were sent to a cloud platform. These nodes were mounted to acquire vibration from an electric bicycle motor. Three situations were simulated: unbalance, using weights at the endings of rim spokes; mechanical load, using neodymium magnets, through the principle of eddy currents; and rotor obstruction, using an object made with a nylon cable. The solution presented was used to measure vibration and calculate its spectrum and send it to the cloud. The signals were analyzed using three strategies: an FFT amplitude level comparison, logistic regression, and neural network (NN). The analyses were carried out using the signals of each sensor independently and as a sensor network, this last showing better results. It was possible to diagnose each type of fault inserted in the tests, proving that the device developed can be used in industries as a low-cost alternative to monitor the health of rotating machines.