The Improved Fast ICA Algorithm in the Application of Vibration Signal Analysis Research

In the process of mechanical equipment condition monitoring, the sensors do not often collect independent source signals of vibration, acquisition is the mixture vibration signals of tested mechanical equipment and the neighbourhood, it brings certain difficulties to the mechanical equipment condition and fault monitoring. Aiming at this problem, it put forward an improved algorithm of fast independent component analysis (Fast ICA), mixed signal can be extracted into separations of acquisition. The algorithm based on maximum negentropy iteration establish the objective function, with the non gaussian signal as the yardstick, then, solve the target function. Finally, the method is verified by mechanical vibration simulation signals that it can effectively separate the mixed signals, separated results show that the algorithm has stability and accuracy, and provide an important guarantee for the mechanical fault diagnosis, practical it has practical significance.