Blind source separation based on JADE algorithm and application
Kangrong Zhang, Guanyu Tian, Tian Lan · Advances in computer science research · 2015
Fault monitoring or detection online is very important for the safe operation of power equipment.The traditional contact-type detection system based on vibration feature exist some troubles under the high voltage and strong electromagnetic field conditions.In this paper, a non-contact sound detection method is proposed, which is based on audio features blind source separation (BSS).The joint approximate diagonalization of eigenmatrices (JADE) algorithm is applied in the detection method, which makes use of the forth order statics to automatically suppress Guassian background noises and enhance the non-Gaussian source signals.Compared with the fast independent component analysis (Fast ICA) algorithm, a clearer source signal can be estimated and separated.The simulation experiment results show that by the JADE algorithm, the similar coefficients between the separated signals and the original source signals are all above 0.9, far greater than that by Fast ICA algorithm.So it concludes that the sound BSS method can effectively separate source vibration sound of power equipment and JADE is more efficient than Fast ICA in the fault diagnosis for power equipment.