A phase space reconstruction based single channel ICA algorithm and its application in dam deformation analysis

Separating noise, systematic errors and deformation components that are caused by different forces such as water pressure and temperature should be identified from dam deformation monitoring data to facilitate effective dam deformation analysis and consequently determine the potential dam structural damage. As source signals are assumed to be mutually independent, independent component analysis (ICA) can be used to separate the source signals from their mixtures. In addition to signal independence, another requirement of ICA theory is that the number of mixtures should not be less than that of source signals. Unfortunately, in most cases of dam deformation monitoring, only a single channel of observed displacement data is available for each monitoring point. Therefore, ICA with single channel data input (called single channel ICA) is necessary in the application of deformation analysis. In this paper, we apply a phase space reconstruction based single channel ICA (PSR-ICA) algorithm to de-noise deformation monitoring data and separate deformation components introduced by different forces from these de-noised data. A numerical simulation is conducted, and results indicate that PSR-ICA is an efficient tool not only for denoising data but also for separating deformation components caused by different forces. PSR-ICA is then further utilised to process the displacement monitoring data of Wuqiangxi Dam. Results indicate that the two extracted main displacement components are nearly consistent with the displacement components; temperature and water level are used as variables and are computed by using a regression model. Both numerical simulated and real life dams demonstrate that PSR-ICA is an effective tool for separating deformation components by different causative forces and is therefore beneficial to dam deformation analysis.