This entry was posted in Ground Motions and tagged baseline, butterworth, Chase, corner, FFT, filtering, frequencies, noise, Rob, seismogram, SeismoSignal on Januby roch9726.
SEISMOSIGNAL LAST LINE LIMIT SERIES
The final corrected and filtered time series is shown in Figure 2.įigure 5: Selection of Corner Frequencies Don’t worry about the area circled in red as it is due to SeismoSignal and should not be there. The filtering can be seen as the Fourier amplitude is zero outside of the corner frequencies. Figure 6 shows the FFT plot after filtering. After applying the filter, the noise will be removed resulting in a corrected time series like Figure 2. Figure 5 shows the two corner frequencies chosen on the FFT plot. Selecting corner frequencies is up to the user and can be dependent upon the desired usage of the data. Since an FFT plot is in the frequency domain, it is easy to identify high and low frequency noise and consequently select the corner frequencies. A FFT plot is a great tool for selecting these values. For example, if 0.3 and 30 Hz are selected as the corner frequencies, all frequencies between those values will pass and be included in the data and everything outside will be filtered out. The corner frequencies determine the bounds of the filter. This noise is most easily removed by the use of a bandpass filter like a Butterworth Filter. This is shown in the two areas in the Fast Fourier Transform (FFT) plot in Figure 4. High and low frequency noise can also contaminate the signal. The corner frequencies (“Freq 1” and “Freq 2” in Figure 3) are the low and high bounds of the bandpass filter everything between these bounds is retained in the signal. Generally, an order of 2 or 4 is selected for the order type in seismology practices. A bandpass filter combines a lowpass and highpass filter to remove both high and low frequency noise. For seismological purposes, a Butterworth filter type is frequently used. Filtering, meanwhile, is typical done in the frequency domain to remove unwanted frequencies. Baseline corrections are typically done in the time domain and are used to remove unwanted trends. In the case of Figure 1, a linear correction was selected. This can be changed depending on the trend being removed. The filter requires multiple inputs which are shown below.įigure 3: SeismoSignal Filtering Filter InputsĪs shown, the baseline correction can be applied using different techniques. This will bring you to the area where the filter can be applied. After loading a seismogram in SeismoSignal, select the “Baseline Correction and Filtering” tab at the top of the page. Please reference the fantastic blog post, SeismoSignal, to get acquainted with the program. SeismoSignal is an incredibly useful program that contains all the necessary tools to filter a seismogram. Notice the displacement time history does not deviate towards the end and the peak ground displacement is aligned more closely with the largest acceleration and velocity values.įigure 2: Filtered Seismogram Filtering Using SeismoSignal By removing the noise, a seismogram more indicative of the actual event can be attained. It can be recognized as noise because it is physically impossible for the ground to continue to displace after the shaking has stopped. This trend is not representative of a typical ground motion and inaccurately represents the actual displacements in the earthquake. This is because the errors in the acceleration time history are magnified through the double integration needed to get from the acceleration to the displacement. Here the data deviates from zero showing increasingly large displacements with the final displacement being the largest. The noise is most evident in the displacement time history. Examples of Noiseįigure 1 is an example of noisy raw data. However, this post will focus on using SeismoSignal. Other programs such as Matlab have built in filtering functions as well. A program such as SeismoSignal is very help as it already has built in filters and other correction tools. Removing trends, making baseline corrections, and applying a filter can help remove the noise and provide data that is more representative of the actual event. Noise is apparent in multiple ways, such as deviation from the baseline, high frequency contamination, low frequency contamination, and other error due to trends.
It is important to remove this noise from the signal to get as close to the real ground motion as possible and provide a “clean” input into structural analysis models. This noise is due to numerous sources which alter the actual earthquake signal. Earthquake data which has not been processed is inherently noisy.
0 kommentar(er)
