Feasibility of using low‐sampled accelerometer measurements for bolt joint looseness detection

In this work, the feasibility of using low-sampled vibration signals for bolt joint tightness detection was investigated. Testing was carried out on multiple bolt joint configurations using a bench top electrodynamic shaker rig. Two data-processing methods were successfully used to deduce bolt joint loosening from the accelerometer measurements, namely the resonant frequency and regression methods (ARX and AR-ARX). Both methods were able to detect loosening of bolt joints, however, the latter possesses higher sensitivity in detecting the position of the loosened bolt among an array of bolts. As the resonant frequency of wind turbines is low (0.35–2 Hz), the minimum sampling rate for bolt joint tightness detection is consequently also low (twice the resonant frequency). This facilitates potential use of existing accelerometer instrumentation on wind turbines, typically sampled at low rates.