Particle swarm optimization-based empirical mode decomposition–kernel independent component analysis joint approach for diagnosing wind turbine gearbox with multiple faults

Abstract

Under-determined blind source separation (BSS) of nonlinear mixed signals in multiple-fault detection of wind turbine gearbox has been considered a challenging issue for years. The paper addresses this problem and presents an efficient solution through a combination of empirical mode decomposition (EMD) and kernel independent component analysis (KICA) methods. The nonlinear mixture signals are firstly decomposed into a set of intrinsic mode function (IMF) components using EMD, which can be combined with the original observed signals to construct a set of new signals. Thus, the original problem can be effectively transformed into a problem of over-determined BSS, which can be solved by the use of KICA. The adoption of particle swarm optimization (PSO) algorithm can further enhance the performance of the EMD–KICA solution. The proposed solution is assessed through a set of simulation experiments and the numerical results demonstrate its effectiveness.