Study On Structural Vibration Signal De-noising Based On ICA And Different Additional Channels

It is one of the hot spots to study the structure health monitoring and the condition assessment for structure workers. The vibration examination is an important method in structure health monitoring; many vibration signals have been gained from examinations. Because of the complex environment influence to sensors such as white noise and all kinds of driving waves, the signals contain some stochastic ingredients except the vibration information; the stochastic ingredients are called noises. In order to gain reliable signals to the next process of damage recognition, de-noising has become an urgent problem to us.The de-noising to structure vibration signal is one of the signal pretreatment problems; so many methods of signal processing can be used. Independent component analysis (ICA) is a creative method in this field. They have got many achievements in real. The ICA method is tried to reduce the noises of structure vibration signals in this paper. The main research work is followed:(1) The theory and algorithms of independent component analyzes is presented. The pre-processing of ICA and the fixed-point ICA algorithm based on negative entropy are introduced.(2) The noise origin and its characteristic are analyzed; the method of de-noising based on wavelet is presented. Wavelet method is used to reduce noises with the numerical example and the vibration experiment signals, but the limitation of it is expressed in real.(3) The noises reduction processing algorithm based on ICA is proposed in structure vibration signals. Because of the independence characteristic between the noises and structure vibration signals, the noise can be partly or completely separated from the sensor signals with ICA algorithm. Two arranging methods of additional channels are taken out in article. One is to increase the expanding channel in same position. The other is to increase the virtual or practical noise channel. Finally, These judge indexes such as: signal noise ratio, the FFT fair is used to value the de-noising effect to numerical example and the vibration experiment data, the good de-noising effect explains the ICA has the vast potential in the field of structure vibration signals.