PSI - Issue 64
Mohammad Shamim Miah et al. / Procedia Structural Integrity 64 (2024) 476–483 M.S. Miah and W. Lienhart / Structural Integrity Procedia 00 (2024) 000–000
479
4
Fig. 1. Sample measured data: (a) full-time series; (b) selected time-window [134 - 168 sec].
3. Results and discussion The results of this study consists of data post-processing, ARI model development, model validation, and, finally, forecasting. The experimentally measured displacement data has been utilized to achieve the goal of this study. Typically, the measured displacement data is considered to be less noisy in contrary to acceleration data. There is obviously an advantage of having displacement data which can directly be compared with reference position data without doing and integration and differentiation. However, it might not be so easy and costly to have such data due to the measuring range of displacement sensors e.g. laser type sensors, geodetic sensors. Furthermore, it might be very difficult to measured large deformation by using laser type displacement sensors. 3.1. Data post-processing It is usually unavoidable and also expected that the measured signals are going to be corrupted with noise for various inherent reasons e.g. sensors type, inputs. Hence, the data cleaning during the post-process (e.g. detrending) stage as well as filtering are essential. In this study, also data cleaning has been performed and a representative time-window is selected for the further use i.e. for model. A sample result from the measured data is depicted in Fig. 1. It needs to be mentioned that normally the displacement data are cleaner or less noisy than acceleration data and it can be be seen as well here in Fig. 1. In the next step, the data has been filtered though the measured signals seems to have very little or no noise. Fig. 2 shows the contrast of the original data versus filtered data. It can be seen in Fig. 2 that the filtered and unfiltered data shows good similarity. Further to ensure that do dynamics of the original and filtered signals are cross-checked by performing fast Fourier transformation (FFT) and power spectral density (PSD) estimation. As mentioned earlier that both the FFT and PSD estimations are conducted here to see the contrast among those two methods. To do this, in addition to conventional FFT, the Welch’s power spectral density (PSD) estimation technique has been utilized to identify the governing frequencies. The Welch’s PSD es-
Made with FlippingBook Digital Proposal Maker