A wavelet-based approach for the detection and modelling of non-stationary effects in structural deformation analysis

The modelling of continuously observed structural deformation processes is typically done using specific tools of time series analysis and linear system theory. The system parameters are calculated from the entire available data sets, presuming their stationarity at least up to the 2^nd order. For some observed phenomena this assumption needs, however, a thorough consideration. Due to irregular influences or to different operation states of the monitored object, the recorded time series might suddenly change their statistical equilibrium state. In such situations, the standard modelling tools yield poor estimators of the systems parameters. This paper introduces concepts concerning an extended deformation analysis, which includes the handling and modelling of such signals. It is argued that the Wavelet Transform is building the proper framework in order to accomplish this task. The paper concludes with results obtained by the implementation of the proposed approach in monitoring projects.