Early diagnosis of helicopter gearboxes based on independent component analysis

Early gearbox fault signal is often very weak and SNR is low, which greatly constrains the use of developed diagnosis methods. Thus how to determine the true vibration signals is the key to improve the early diagnostic performance. The advent of Independent Component Analysis (ICA) provides one way-out for it. This paper proposes the application of ICA to early diagnosis of helicopter gearboxes. At first the mixing model of gearbox vibration signal is built; then one experiment is done on one actual faulty bearing and then the bearing vibration signal is separated by FastICA algorithm. Different PSDs using FFT between separated signal and original signal are compared and the results demonstrate that ICA can he used for early fault diagnosis easily indeed. In the end advice for the future is given.