A multi-faults separation method based on improved sparse component analysis

Multi-faults which usually exists in roller bearing makes fault diagnosis difficult. Thus, it is of great significance to carry out fault diagnosis of rotating machinery to ensure the complete machinery system to perform in a normal statement. To effectively separate the multi-faults and achieve the fault diagnosis, sparse component analysis based on the sparsity of objective signals was proposed. However, the vibration signals are scantily sparse and it usually cannot be represented in a sparse way. Therefore, a novel approach is suggested to overcome the above problem. First, tunable Q-factor wavelet transform is used to obtain the sparse signal. Then, the quantity of source signal and the mixed matrix are approximated grounded on the potential function which is calculated according to the sparse signal. Ultimately, the shortest path method is taken advantage of gaining separation of the source signal. The simulate signals and vibration signals are used to verify the effectiveness of the proposed method. The results show that it is telling to accomplish the separation of multi-faults.