Diagnostic Feature Development based on Dynamic Time Warping of Dynamic Signals under Variable Machine Operating Condition

In condition monitoring (CM), measurements are often taken when a machine runs under different loads and speeds. The signals in these conditions show similar profiles but do not line up in time-axis for adequate comparison. Detection features derived in the time domain are not accurate enough to discriminate small changes in machine health conditions. In this paper, a Dynamic Time Warping Algorithm (DTW) is explored in aligning the signals under different machine operating conditions for both normal and faulty status. The feature and performance of DTW algorithms are reviewed and improved based on a set of simulated data consisting of common features in CM practice. Experimental data sets of electrical motor current signals have been studied using DTW. The performance of the event alignment is evaluated by the capability in detecting the fault with different severities. Preliminary results show the aligned data produces accurate results and hence can lead to better detection and diagnosis results.