Independent Component Analysis And Its Applied Research In Deformation Monitoring Data Processing And Analysis

Deformation monitoring relates to engineering geology, structure mechanics, computer science and so on, it is a comprehensive technology and has become one of the most active research areas. It contains two mainly content:one is understand the stability of building, including finding question timely and taking action so as to guarantee its safety operation; another is science significance, including grasping deformation mechanism, carrying ticking to design and building deformation foresee model.With the development of deformation monitoring methods and technology, the monitoring data become more and more abundant, as a result it can provide more information for monitoring the safety state, however it can also made the deformation analysis be more complex. In order to improve the accuracy of deformation analysis and forecast, on the one hand the accuracy of survey data should be improved, on the other hand information fusion technology is needed so as to improve the accuracy of structure healthy diagnosis and disaster forecast.The method of ICA is introduced into deformation monitoring data processing and analysis, on the basis of signal de-noising, regression model was built.The main research work of this article is followed:1. The research of signal processing based on independent component analysis(1).On the basis of statistics independence principle, by mean of analyzing the higher order statistical characteristic among many dimensions data, independent components can be gained. By mean of the result of simulation test, we know that the independent components obtained by ICA are similar with source signal except for the uncertainty of order and amplitude of signals.(2).The method of ICA is used in dam survey data and the effect in reality application of it can be illustrated. Dam deformation signal is affected by water level, temperature, time component and noise. Before using the method, the quantity of sensor must be equal to or more than the quantity of source signals.(3). ICA can not distinguish useful signal and noise, so the components separated by ICA will be distinguished according to other prior experience. In this paper, signals and noise were effective distinguished.(4). The comparison of de-noising between ICA and wavelet was made. Evaluation index includes signal to noise ratio and correlation coefficient. After simulation test and example were made, conclusion was obtained that ICA de-noising is superior to wavelet de-noising, moreover the accuracy of components abstracted by ICA is much better.2. The research of multiple regression analysis based on independent component analysis(1). Principal component regression has been applied in many field, now, we use Partial Least squares Regression, Principal component regression and independent component regression to solve the assumption regression model, on the basis of analyzing of result, we know that independent component regression can be used in multiple regression analysis, moreover, factors gained through this method can explain the deformation better.(2). After the dam monitoring signal was handled by ICA and the noise signal was identified and removed, the rest factors which affect the deformation of dam were used to calculate regression model, on the basis of comparison of measured value and predicted value, we can estimate the prediction accuracy of the model.