Dynamic Analysis of a Dam using Strong Motion Accelerograph Data

As part of the operation of the Victorian Water Industry seismic network, the Seismology Research Centre recently installed triaxial strong motion accelerographs (SMA) on Glenmaggie Dam.

One was installed near the centre of the crest of Glenmaggie Dam, and another at an existing weak motion seismic monitoring site located near the left abutment of the dam.

The SMA located on the crest of the dam (GLMMD) monitors the response of the structure to significant nearby earthquakes, while the SMA located near the abutment (GLMMA) acts as a bedrock reference site to determine the ground motion input that influences the response of the dam.

Measuring the response of the dam to strong earthquake motion will help confirm that the dam behaves as expected, confirm design parameters, and may detect hidden damage to the structure following a large earthquake.

In a low earthquake hazard region such as Australia, because the primary aim of these instruments is to record infrequent, moderate to large, nearby earthquakes, most of the time these instruments are simply recording ambient seismic noise.

In this context ambient noise is a generic name for persistent vibration of the ground (or structure) generated by a range of sources including human-related activities, wind and other atmospheric phenomena, river currents, and ocean waves.

We can use the ambient noise recorded by SMA instruments, as well as records of small to moderate earthquakes, to determine some of the dynamic characteristics of the dam. The SRC has undertaken an analysis of this data for Glenmaggie.

Two different procedures for examining the ambient noise at Glenmaggie Dam were used. The first technique involves transforming the time domain recordings of the ambient noise into the frequency domain using a Fast Fourier Transform (FFT) and examining each channel individually. If we assume that the input ambient seismic noise is primarily white noise (i.e. equal intensity across all recorded frequencies) then any peaks in the data recorded on the dam should provide us with information about the dam. The Fourier spectra data recorded on the abutment is uniform across all frequencies the instrument is capable of recording (see figure 1 below), which confirms the input noise to the dam is indeed white noise.

Figure 1. Time domain (red) and frequency domain (black) data from Abutment.

The Fourier spectra data recorded on the upstream/downstream channel of the crest SMA show two distinct peaks at 8.6 and 9.3 Hz, indicating that the white noise input is exciting a response in the dam at these frequencies. See Figure 2.

Figure 2. Time domain (red) and frequency domain (black) data from Crest

The second technique for analysing the ambient noise vibrations again involves looking at the data in the frequency domain, but using the ratio of the horizontal to vertical motion (H/V) to tell us something about the dynamic properties of the dam. Figure 3 shows the individual channels from the crest SMA in the frequency domain after some filtering and windowing has been applied and the blue trace (upstream/downstream) again shows a peak response about 9.3 Hz.

Figure 3. H/V data from the crest

Earthquake Response

On June 23, 2023 the SMA instruments at Glenmaggie recorded the largest of the Woods Point earthquake aftershocks (to date), a magnitude MLv 4.7 event at a distance of 60 kilometres. Although this event is not large enough or close enough cause any damage to the dam, if we assume that the dam will respond in a similar fashion to both large and small earthquakes, the data from this small event may tell us something about how the dam may respond to larger earthquakes.

Figure 4 shows the earthquake as recorded by the abutment SMA (the top three channels) and the crest SMA (the lower 3 channels).

Figure 4. SMA recording of MLv 4.7 event at 60 km distance.

The data indicates that the motion was amplified by a factor of almost x1.6 in the direction along the dam wall, between the abutments, by a factor of x11 in the upstream/downstream axis and by around x1.6 for the vertical motion.

The maximum amplitude recorded in the upstream/downstream axis was 331.6 mm/s² (0.0338 g) however as Figure 5 illustrates, the peak vector sum acceleration recorded on the dam crest was slightly higher at 350.7 mm/s². Peak displacement of the crest was less than 1 mm.

Figure 5. SMA Vector Sum Acceleration, Displacement and Velocity plot.

The relatively small size of the earthquake means that there is insufficient duration of significant motion for a full H/V type analysis of the motion of the dam. There are not enough “windows” of earthquake motion to accurately and reliably capture the frequency, however there is some useful data within the recording.

The Fourier transform of the north channel of data recorded before, during, and after the event shows a significant peak of energy around 9.55 Hz, see figure 6.

Figure 6. Frequency domain data from the upstream/downstream crest SMA.

Based on both the ambient noise data and the recording of the earthquake on the crest it appears that the natural frequency of the dam is around 9.5 Hz, slightly higher than previous estimates for the structure of around 7.4Hz for the spillway area.

Assuming that the dam will behave linearly with respect to amplification for both large and small earthquakes, the data from the H/V analysis recorded on the crest and the ratio of motion recorded from the MLv 4.7 earthquake at both bedrock and crest sites indicate that an amplification ratio of around 10 may be expected on the dam crest.

In addition to these useful insights, a H/V analysis of ambient vibrations recorded at the bedrock reference SMA site (GLMMA) demonstrate a very flat response, indicating that there are no soft sediments on site that are likely to generate any significant site response. This is useful information for any future seismic hazard analyses carried out for the site.