Blind Information Extraction of Machine Faults Based on Separating Matrix

Abstract

Blind signal processing is an effective feature extraction method for mechanical vibration signals. However due to noise corruption, independent source signals can't always be accurately recovered or separated from the acquired sensor observations. Then feature information extracted from source signals can't naturally represent machine states of the detected mechanical equipment. Generally in blind signal processing, the separating matrix may contain as much information contents as the separated source signals. The separating matrix can directly be processed to extract its singular values as useful feature information by the singular value decomposition (SVD) method. Thus a blind information extraction method was proposed to extract singular values of separating matrix as the desired feature information of the detected machine. The experimental results of gear pump indicate that this method can be applied to feature extraction of mechanical equipment.