Abstract:
Most traditional methods can detect signals in noise by detecting the change of mean value of test statistics. But these methods assume that the signal parameter is known...Show MoreMetadata
Abstract:
Most traditional methods can detect signals in noise by detecting the change of mean value of test statistics. But these methods assume that the signal parameter is known without considering unpredictability of the actual industrial process. Generalized Likelihood Ratio Test (GLRT) detector, one of most widely used detector, can replace unknown parameters with the results of maximum likelihood estimation. However, the classical detector cannot change with the working conditions. Also it can't solve the troubles of dynamic variation using prior knowledge, which lead to a very high rate of missing detection. To overcome this shortcoming, an adapted GLRT detector is proposed based on subspace alignment to highlight weak signal characteristics, then a mechanism changes time series model is established by Detrended Fluctuation Analysis (DFA). Experimental date is used to illustrate the effectiveness of the proposed detector by using three-phase stator current of Marine Current Turbine (MCT) under complex working conditions. The test result shows the proposed detector can extract weak features, and has good feature portability.
Published in: 2019 CAA Symposium on Fault Detection, Supervision and Safety for Technical Processes (SAFEPROCESS)
Date of Conference: 05-07 July 2019
Date Added to IEEE Xplore: 06 October 2020
ISBN Information: