Vibration Signal Filtering Algorithm Based on Singular Value Subspace Decomposition

In the area of fault detect for rotating machinery, the vibration signal should be filtered before parameter detection and quality evaluation. For the filtering of the vibration signal, it is needed to keep the linear phase characteristics while obtaining good filtering effect. Aiming at the filtering problem of vibration signal, a filtering algorithm based on singular value subspace decomposition is proposed in this paper. The Hankel matrix is constructed by the vibration signal to carry on the singular value decomposition. According to the distribution characteristic of the singular value, the singular value space is divided into the signal singular value subspace and the noise singular value subspace. By setting the proper threshold value, the signal singular value subspace is preserved, and the noise singular value subspace is removed, then the filtering effect is achieved. The algorithm is analyzed and simulated based on the measured steam turbine’s vibration signal. The experimental results show that the proposed algorithm not only has good effect on noise removal, but also maintains the linear phase characteristics of the signal. At the same time, the calculation amount of this algorithm is smaller than other filtering algorithms, which can be applied to the real-time analysis of the vibration signal.