Field Experiments of Distributed Acoustic Sensing Measurements

Abstract

Modern, large bridges and tunnels represent important nodes in transportation arteries and have a significant impact on the development of transportation. The health and safety monitoring of these structures has always been a significant concern and is reliant on various types of sensors. Distributed acoustic sensing (DAS) with telecommunication fibers is an emerging technology in the research areas of sensing and communication. DAS provides an effective and low-cost approach for the detection of various resources and seismic activities. In this study, field experiments are elucidated, using DAS for the Hong Kong–Zhuhai–Macao Bridge, and for studying vehicle trajectories, earthquakes, and other activities. The basic signal-processing methods of filtering and normalization are adopted for analyzing the data obtained with DAS. With the proposed DAS technology, the activities on shore, vehicle trajectories on bridges and in tunnels during both day and night, and microseisms within 200 km were successfully detected. Enabled by DAS technology and mass fiber networks, more studies on sensing and communication systems for the monitoring of bridge and tunnel engineering are expected to provide future insights.

1. Introduction

Bridges provide important channels for transportation and frequently represent the economic lifelines of a country. Their construction and maintenance are both important components of the transportation infrastructure. One study topic that has received a great deal of interest throughout the years is the detection of structural problems [1] in bridges and tunnels. Health monitoring technology can be a powerful tool and a reliable way of maintaining structural and driving safety during long-term operation, and in particular for large-span bridges and tunnels. The traditional safety monitoring mainly relies on manual inspections, supplemented with the corresponding detection equipment, but it is difficult to achieve the real-time and continuous monitoring of bridges and tunnels. With the continuous development of the Internet of Things, significant advances in technology, cooperation between several disciplines and smart cities [2], health and safety monitoring, and the maintenance of bridges and tunnels mainly includes the two major components of data collection and data transmission processing. In particular, onsite data collection is crucial and various types of sensors are required for real-time data collection.

Distributed acoustic sensing (DAS) is an emerging and promising technology in the fields of oil and gas resources [3,4,5], transportation [6,7,8,9], marine geophysics [10,11,12,13], natural earthquakes [14,15,16,17,18,19,20,21], and perimeter security [22] due to the outstanding distributed sensing ability of vibration signals. DAS uses optical fiber as the sensing element, which has the advantages of being light, low-loss, immune to electromagnetic interference, passive, and robust for long-term monitoring. DAS can also benefit from the deployment of dense fiber-optic telecommunication networks [23,24,25] which provide a more realistic and cost-effective platform. There are a number of studies on DAS, including studies covering the optimization of algorithms [26] to improve detection performance, underwater applications [27], and geological detection [28]. However, these studies only demonstrate the application of DAS in a single scenario. The simultaneous monitoring of complex scenarios, such as including bridges, islands and tunnels, still needs to be explored. For example, the monitoring of the Hong Kong–Zhuhai–Macao Bridge, which is of an exceptional length and extremely high construction difficulty, is particularly significant and meaningful.

In this study, field experiments using DAS for monitoring different events based on fiber-optic cables were demonstrated in the Pearl River estuary area of the South China Sea. For the sensing part of the system, phase-sensitive optical time-domain reflectometry (OTDR) technology was used to detect the acoustic signals caused by various activities. In this way, the simultaneous monitoring of complex scenarios, including those of the bridge, land and ocean, was achieved. On land, earthquakes and onshore events were recorded. On the bridge, vehicle-driving information was observed. In the ocean and tunnel, data relating to ocean waves were obtained. Here, the focus is on the sensing capability of DAS based on earthquake detection and vehicle monitoring in the bridge and tunnel. With the acquisition and analysis of these data, these events were able to be recorded and studied, providing a helpful tool for oceanography [29], geology [30], bridge monitoring [26], and other research fields [31,32,33].

2. Methods

The basic principle of practical DAS adopts OTDR architecture, utilizing backward Rayleigh scattering phase change or spectral shift to achieve the linear recovery of the amplitude, frequency, and phase information of vibrations. In order to obtain effective information, it is crucial to accurately analyze the data recorded from DAS (Silixa IDAS), which is usually installed in the equipment control room. Then, the existing communication facilities can be fully utilized, such as the optical cables and other onshore facilities. The phase information can be extracted from the Rayleigh backscattering signals and used to detect the activities around the sensing fibers. Due to the impact of the surrounding environment and other noise, the measured data from DAS are complicated and very large, exceeding normal data sizes.

In the field test, DAS was carried out using a telecommunication optical cable of ~35.2 km, as shown in Figure 1. The DAS recording instrument was placed in the telecoms control room on the shore for recording. Optical cables were arranged along the bridge, as shown in the inset to Figure 1. The arrangement of optical fibers was not parallel to the direction of the bridge, and the fibers were very entangled. In the diagram, the fiber shown represents an approximate path due to the lack of exact deployment. On the map of China, the path along the bridge can be divided into several sections, i.e., the first section of 2.8 km near the shore, the second section of 22.7 km on the bridge and the third section of 9.7 km, buried in an undersea tunnel. Various activities were recorded in several minutes using the optical cables. The real-time monitoring was achieved through the software interface [19] which could display the map, waterfall plot, and information of the detected various activities, including the time, location, and magnitude. The time interval and channel interval during data collection were applied as 1 ms and 4.0838 m, respectively. In the test, the data were collected every minute with 8704 data channels in space, which can be seen as 8704 sensing elements in total. The raw data were measured directly using the DAS, which was monitored in real time through the software interface. The strain-rate data were obtained from the raw data. Then, the strain-rate data were processed and analyzed using signal processing methods to obtain the experimental results. In our results, the original data referred to the strain-rate data. For the signal processing, frequency-wavenumber (f-k) domain analysis, filtering, and normalization were adopted. The f-k domain analysis was implemented via the application of two-dimensional (2D) fast Fourier transform. The filtering was implemented using the bandpass filter on the data of each channel to remove the disturbing frequency components. The normalization was implemented by dividing the data by the difference between the maximum and minimum values.

3. Results and Discussion

During the field test, the measured results were obtained within 21 min after UTC 2021-08-06 08:10:06, in the sea area near the Hong Kong–Zhuhai–Macao Bridge Port. The original wave swarm obtained from the DAS was first preprocessed and the results by the time-distance plots within 21 min were demonstrated, as shown in Figure 2. There is no signal processing, and the plot is obtained from the absolute value of the strain-rate data. Due to the large amount of data exceeding the scope of computer processing, it should be noted that the sampling rates of the data points both in time and space were reduced to four times the amount of the actual data for the convenience of drawing. In addition, the channel spacing for the data collection remains unchanged and the data interval for plotting increases. In this section, we will discuss in detail the results of our field test based on the signal processing method.

In Figure 2, the horizontal axis in the figure represents the measurement time with unit seconds (s). The vertical axis is along the fiber direction from Zhuhai Port (bottom) to Hong Kong Port (top). The unit of fiber length is meters (m). The color scale represents the strain rate variation in the data measured at the fiber optic vibration source. The associated unit is expressed in units of nm/m/s. The gray dashed lines mark the sections of coastal region, bridge, and ocean illustrated in the figure. The swarm plot can be divided into three sections, as mentioned above. At the bottom of the waterfall chart, the first section of the coastal region involves optical fibers ranging from 0 to ~2.8 km, where earthquakes, ocean waves and construction activities on shore can be expected to be detected. The second section of the bridge, including many clear and bright trajectories, covers the optical fibers from ~2.8 km to ~26.5 km in the middle of the figure. Here, many tilted trajectories can be observed, indicating the driving situation of the vehicles on the bridge during the observation period. The last section of the ocean corresponds to the fiber lengths ranging from ~26.5 km to ~35.2 km, where earthquakes, ocean waves, or other underwater activities may be detected in the tunnel. These three parts of the data evidently have different distribution characteristics from each other. For example, the tilted trajectory can only be observed in the middle section, and the ocean waves can be found in the other two sections. Therefore, the following analysis will be divided into three parts to discuss the characteristics of each part of the data. In addition, the data processing was separately performed on the three parts for analysis. The f-k domain analysis was mostly used to study the ocean dynamics. In order to ascertain the f-k spectrum, a 2D fast Fourier transform was applied on the original data from the sensing module output, which records the acoustic and vibration information with the submarine fiber. The original data refer to the strain-rate data, and there are no pre-processing steps performed before the Fourier analysis.

3.1. Coastal Region

Figure 3 shows the results of the first section of the original wave swarm in Figure 2, where the fiber length involved is 0–2.8 km. As shown in Figure 3a, the original wave swarm is provided again for the convenience of analysis and comparison. No signal processing is performed, and the plot is obtained from the absolute value of the strain-rate data. The red dashed box marks the section within 500 m closest to the land. Figure 3b shows the calculated f-k spectrum based on our test with the dataset, where the 2D fast Fourier transform is used. The horizontal axis represents the frequency with units of 1/s, and the longitudinal axis represents the wavenumber with units of 1/m. Then, the slope of the line is equal to the reciprocal of the wave velocity. The trend of the spectra shows the ocean-surface gravity dispersion relationship, which also describes the ocean surface gravity waves and indicates the energy propagating axially along the cable. With this f-k spectrum, the phase velocity can be calculated as the ratio between the frequency and wavenumber for the different oceanic acoustic signals with different frequencies, which describe the ocean wave propagation. There are two main areas observed in this f-k spectrum. One area is around the 11 m/s, whose frequency is below 0.4 Hz. This area indicates the natural ocean waves with a small wavenumber (<0.04 1/m) and a slow phase velocity (around 11 m/s). The other area is between 200 m/s and 1000 m/s, which shows the waves with a small wavenumber (<0.015 1/m) and a higher frequency (>15 Hz). These waves with large phase velocities of several hundred m/s are caused by the propagation and conversion of some seismic waves in the fluid ocean [34]. This implies that the telecommunication fiber cables with DAS technology can also be utilized for oceanic seismic monitoring and observation. Figure 3c shows the zoomed-in view of the 0–500 m range corresponding to the red dashed box in Figure 3a where no signal processing is performed. There are some activities between 200 and 300 m in the swarm. Then, the further analysis was conducted.

3.1.1. The Section of 0–480 m

The first section of 0–2.8 km in Figure 3a shows some activities have been detected within 480 m of the optical cable distance, where the transition connection area occurs from the control room to the corridor. The zoomed-in view of this section further indicates these activities are in the range of 200–300 m, as shown in Figure 3c. Thus, the spectrum analysis within this range is presented, where fast Fourier transform (FFT) is used. The examples of the results during the entire observation period are also provided in Figure 4. The position of the optical fiber is calculated as approximately 210, 270, and 280 m. Two main frequencies of ~10.4 Hz and 115 Hz are observed. There have been significant changes in the frequency and intensity of the frequency spectrum at 280 m. Multiple frequencies appear in the spectra, such as 100, 185, and 258 Hz. The intensity at a frequency of around 10.4 Hz exceeds that of 115 Hz, and becomes the maximum value. The generation of these frequencies [1] may principally be caused by the equipment located at the ventilation openings, such as the air conditioning. These frequencies were only observed in the first coastal region, and were not observed in the bridge and further ocean areas. This indicates that the frequencies are caused by events on land near the coastal region.

3.1.2. The Remainder of 0–2.8 km Section

The remainder of the section of 0–2.8 km in Figure 3a shows no activities. To verify this result, further analysis was adopted. Figure 5 shows the preprocessing results after filtering and normalization using the time-distance plots with fiber cable from 480 m to 2800 m. The filtering is implemented using a 2–15 Hz bandpass filter to remove the disruptive frequency components. Then, the normalization is adopted. The data plot within 1 km shows that it is still located near the coast. There is no significant activity in the transition zone located from the coast to the bridge.

To further demonstrate the feasibility of the DAS process, the results obtained during earthquakes that occurred at other times are also provided. We have successfully observed inland, coastal, and offshore earthquakes based on our DAS. In order to ensure the consistency, the same filtering and normalizing processing was adopted. As shown in Figure 6, as an example, the results of three continuous microseism swarms with a Richter magnitude (ML) of 1.1, 2.6, and 1.2 were recorded at UTC 2021-08-11 17:16:36 in the sea area near Yangxi city in Guangdong Province, China (21.74°N, 111.79°E). The three continuous microseisms occurred at UTC 2021-08-11 17:18:11 (white line), UTC 2021-08-11 17:18:17 (red line), and UTC 2021-08-11 17:19:33 (orange line). The epicenter is located approximately 200 km away from our detection fiber.

Using the frequency spectrum analysis, we identified that the detected nearby microseisms can be observed with more high-frequency components (>5 Hz); however, for remote earthquakes, mainly low-frequency components (<2 Hz) are observed with our testbed. Since the frequencies of the natural ocean waves are almost below 0.3 Hz, for the case shown in Figure 6, we used a 2–15 Hz bandpass filter on the data of each channel to remove the disruptive frequency components. This helped us to distinguish the seismic signal from the other ocean waves. For a better display of the detected earthquake, max–min normalization could be used for each channel to reduce the difference; however, ideally, we could use every single channel to monitor the seismic events. For example, the P wave and S wave of ML = 2.6 arrived at our testbed about 35.8 s and 56 s, respectively, after the earthquake occurred. With these time intervals, we can calculate the average propagating velocities of the P wave and S wave, which are 5.39 km/s and 3.4 km/s, respectively. The three lines in Figure 6 should be noted as indicating that the earthquake time occurred at 95 s, 101 s, and 177 s. The P and S waves are not straight with a certain tilt and radian, which is due to the actual laying location of the submarine fiber cable. The distances between the multiple fiber-sensing elements and the epicenter are slightly different; thus, the P and S waves arrived earlier at the closer fiber sections. This proves that the submarine DAS has good sensitivity for microseismic event monitoring. In fact, more unknown acoustic signals were observed with our testbed, which are not recorded here. Since the seismic stations are quite limited, especially for the offshore scenario, it is possible that we observed some offshore microseisms that were not recorded by the conventional seismic stations. In addition to earthquakes, it is worth noting that other artificial transient acoustic signals can also be detected by our system. Spectrum and time domain analysis with advanced machine learning algorithms will help distinguish the seismic signals from other disturbing vibrations.

3.2. Bridge Section

In the middle of the waterfall plots in Figure 2, the vehicles’ driving trajectories are plotted with the optical fiber section from 2.8 km to 26.5 km, where the optical fiber is arranged under the Hong Kong–Zhuhai–Macao Bridge. No signal processing is performed and the plot is obtained from the absolute value of the strain-rate data, amongst which, several features are noticeable. The direction of vehicle travel can be determined from the trajectory. For example, there is a driving trajectory illustrated from Hong Kong Port to Zhuhai Port, which is both very clear and the longest. According to the driving trajectory, it can also be estimated that the average vehicle speed during the observation time is about 84 km/h.

In addition, in the waterfall plot, it can be seen that there are obvious protrusions in every driving trajectory, as shown in Figure 7. This is due to the vibration signal generated by the vehicles passing through the gaps between the bridge piers. Based on the two adjacent strong vibration signals, the spacing between the bridge piers can be estimated. For example, there are two strong vibration signals in the trajectory marked by a rectangular box, as shown in Figure 7a, where the direction of travel is from Zhuhai Port to Hong Kong Port with a speed of 82 km/h. As a result, the distance between the two vibrations of the optical fiber is ~700 m. The same results can also be obtained on other driving trajectories. However, the arrangement of the optical fibers (such as any entanglement) on the bridge is unknown due to insufficient permissions; therefore, it is impossible to accurately estimate the spacing between the bridge piers. The vibration signals can be detected in our DAS, indicating the potential applications of DAS in bridge conditions. At the same time, the details of the vibration signal are also demonstrated, as shown in Figure 7b. The vibration signal propagates along the optical fiber at a speed of ~3500 m/s.

3.3. Ocean Section

The results of the rest section of the optical fiber cable are shown in Figure 8. The processing method of filtering and normalizing is the same as that adopted above. The optical fiber cable is arranged in the submarine tunnel, with a length of 9.7 km. Figure 8a shows the time–distance plot of the original wave detected without earthquakes, where the absolute values of the strain-rate data are used for plotting. The same plotting with microseisms is shown in Figure 8b. The outputs of the filtering and normalization in both cases are shown in Figure 8c and Figure 8d, respectively. In Figure 8a, the values are measured during the day and in Figure 8b, the values are measured at night. When there is no earthquake, the vehicle trajectories can still be observed in the optical cable section near the bridge, as shown in Figure 8a,c. This should be due to many vehicles traveling during the day, resulting in strong vibration signals. When there are microseisms, the driving trajectory is displayed after the data processing, as shown in Figure 8d. This is not obvious in the original data waterfall diagram, as shown in Figure 8b. This indicates that DAS based on submarine tunnel optical cables exhibits strong detection capabilities. At the same time, microseisms are also observed in the tunnel, with the intensities decreased significantly compared to those shown in Figure 6. The results present the simultaneous monitoring of complex scenes near the Hong Kong–Zhuhai–Macao Bridge, which integrates the bridges, tunnels, and artificial islands. In addition, the background noise measured in the tunnel section is high both during the day and at night, so the requirement for further data analysis is outstanding.

4. Conclusions

In this study, we present the field experiments of using DAS for the Hong Kong–Zhuhai–Macao Bridge based on telecommunication fiber optic cables. The experimental results show that DAS can be used for traffic analysis, earthquake and microseism detection, and bridge monitoring. The raw data from the DAS need to be preprocessed with basic signal processing including the operations of filtering and normalization. Different events have been extracted from the data waterfall swarm. The vehicles’ driving trajectories were successfully obtained during the day and night in the bridges and tunnels. The driving speed and the direction of the vehicles can be estimated. As an example, the longest driving trajectory from Hong Kong Port to Zhuhai Port was demonstrated with an average speed of about 84 km/h. At the same time, vibration signals were also observed, which are caused by collisions between high-speed vehicles and bridge piers. These observations will provide important information for analyzing the health status and subsequent maintenance of the bridge. In addition, an earthquake swarm was detected and analyzed with the DAS method, which utilized the telecommunication fiber optical cable between two ports in the Pearl River estuary area of the South China Sea. This work is a demonstration of a standard signal processing procedure for using DAS to detect and measure earthquakes. More studies in this field are expected in the future, focusing on the engineering of bridges and tunnels. The implementation of the DAS process can help us obtain more information from the wider environment and provide promising prospects in intelligent integrated monitoring platforms for the oceanography, geology, and bridge structure fields.