An Unified Nonconvex Sparse Decomposition Method for Fault Feature Enhancement of Rotating Machinery

Aiming at the problem of early weak fault feature extraction of rotating machinery under strong noise interferences, an unified nonconvex sparse decomposition (UNSD) method is proposed for the weak fault feature enhancement in this paper. By introducing a reweighted strategy and L2 norm regularization into the traditional generalized minimax-concave (GMC) model, the proposed UNSD method can overcome the amplitude attenuation and overfitting problems to strengthen the stability of the iterative process. The convexity condition of the cost function in our UNSD model is theoretically derived and the optimization algorithm is developed with the forward-backward splitting (FBS) algorithm. Meanwhile, two Fitness functions are customized to optimize the key model parameters via the genetic algorithm and evaluate the decomposition performance of our UNSD model. The effectiveness and superiority of the proposed algorithm have been successfully confirmed in comparison with state-of-the-art approaches through both the simulation study and experiment validations.