High-speed 3D subway tunnel inspection; A San Francisco BART case study

The trend in tunnel condition assessment is toward comprehensive 3D measurement combined with high-quality images that document and quantify damages to a tunnel’s surface. This work can create a cohesive set of data of the structural conditions of a tunnel and can also be gathered throughout the tunnel’s lifespan. In addition, the faster such on-site measurements can be made, the less of an impact they have on the traffic conditions of these sophisticated subway systems.

One such device used to conduct these measurements, the Dibit high-speed 3D scanning system, is based on photogrammetry and light detection and ranging (Lidar) remote sensing technology. Performing an assessment with this system enables users to gather data in tunnels and the system can operate at speeds of up to 96.561 km/h (60 mph). Tunnel shutdowns can be drastically reduced while tunnel safety is increased because of the high rate of speed in which the system can operate. The photorealistic texture of the 3D-models generated also allows for the identification and analysis of small material damage to the tunnel (e.g., cracks > = 0.3 mm wide).

This article illustrates the technology and 3D results of this innovative system and is based on measurements performed in the subway tunnels of the Bay Area Rapid Transit System (BART) in San Francisco, CA (total length 66.23 km (4.1 miles).

Advantages of subway tunnel scanning and monitoring

Subway tunnels undergo assessment testing in distinct time intervals. This is primarily conducted manually by surveying critical cracks and other features of the tunnel, thus reducing its usability and reliability. Traditionally, tunnels must be closed for surveys and inspections, which is difficult to do in busy cities and on modern subway systems. Therefore, shutdowns usually occur at night, when traffic is reduced. On larger projects, inspection crews may work several days or weeks at night, thus requiring resources and leading to higher costs because the work is performed outside of regular work hours.

Modern scanning systems can significantly reduce tunnel closure times because they employ enhanced survey speeds. Because this requires less effort than traditional modes of measurement, the structural health and monitoring of the tunnel can be intensified. More frequent measurements increase the overall safety of subway tunnel systems. With 3D models, the tunnel operator receives precise and comprehensible data for determining needs like maintenance, revision and construction (Kontrus and Mett, 2019).

Thus, 3D tunnel scanning is economical and efficient. Time, personnel and tunnel closures can be reduced or even avoided. The time it takes to conduct measurements and the preparation needed are short as well. Another positive aspect of using digital scanning systems is that most of the inspection time is shifted from the tunnel to a virtual 3D environment in an office. Cracks, tunnel installations and other issues can be analyzed independently from the often rough and sometimes dangerous tunnel environment. The faster tunnel scans can be conducted, the better for tunnel operating companies.

The main task of tunnel scanning systems is the continuous survey of tunnel surfaces and the subsequent 3D reconstruction of tunnel structures. In addition, modern scanning systems provide information about the surface of the tunnel in the form of high-resolution photos that enable recognition of even the finest cracks (> = 0.3 mm) and the classification of tunnel objects, such as construction joints and further installations. Scanning systems used for tunnel monitoring There are currently three approaches for the 3D measurement of tunnel structures and the monitoring of their surfaces.

The oldest and most common approach is laser scanning. Laser data images are available in gray scale. Thus, laser scanners cannot capture tunnel surfaces in true colors or red, green or blue (RGB) values. In laser scanning the measurement of crack profiles of relevant crack widths (0.3 to 1 mm) is possible because of the intensity and contrast differences in the images. Established laser scanning systems can achieve geometrical accuracies of up to 5 x 5 mm for the measurement of clearance profiles, etc.

The second approach is the use of photogrammetric systems, where 3D geometry and photo surfaces are created from the photos of digital cameras. Photogrammetric systems can be used for 3D reconstruction (e.g., tunnel construction) (Bauer et al., 2015) and can operate at high speeds (Mett et al., 2019). However, when the measurements are conducted at speed, a bright illumination and flash technique are required in these inspections. Therefore, such devices are not well established in the market. In general, only a few photogrammetric systems are currently used in the field of tunnel construction. These devices are stationary and operate from a fixed point without movement. One such example is the 3GSM shapemetrix (with commercially available cameras) and the Dibit Handheld 3D-complete system. The shapemetrix TBM is used for documenting the digital rock-face for tunnel boring machine (TBM) rock excavation.

The third approach is a hybrid system that combines geometric data from the laser measurements and photo textures received from digital cameras. These can either be operated as stop-and-go systems, which scan one tunnel section from a constant position and then proceed to a further position (Fig. 1), or as kinematic systems, which scan the tunnel while continuously moving. Some wellestablished systems include the SPACETEC TS3 and the Dibit LSC 4100-SRMF2. Both are configurable for road and track use, achieve geometric accuracies of around 10 x 10 mm and photo resolutions of up to 1 x 1 mm. Both systems reach speeds of up to 4 km/h (2.5 mph) at walking speeds. The newly developed Dibit dynamic system contains industrial cameras and light-emitting diode (LED) flash technology for the 3D measurement and monitoring of tunnel surfaces at walking speed.

Dibit high-speed system for monitoring subway tunnels

The Dibit high-speed 3D measuring system FSC 6100- SRmF10 is able to measure tunnel structures at speeds of up to 96.561 km/h (60 mph). It can be operated in various configurations in subway, rail and road tunnels. The scanner consists of a photogrammetric unit with highspeed cameras and LED flash technology, a laser and an optional thermal imaging unit.

In this measuring system, the longitudinal axis is arranged horizontally, which means that cameras and lasers are aligned orthogonally to the tunnel surface to be measured (Fig. 2). The high-performance cameras are installed in a helical arrangement, which enables 360-degree coverage of the surface. The flash modules are designed with such an intensity that the overall illumination of a two-lane railway/subway tunnel is guaranteed.

The focus ranges of the lenses are set by default in such a way that tunnel surfaces in an area between 2 and 6 m (6.6 and 19.7 ft) can be shot with sharp photo resolution. The focus area can be adjusted for special tunnel cross-sections or for short or long distances to the tunnel surface.

The FSC 6100-SRmF10 system can be flexibly arranged and operated in different configurations during measurement in subway tunnels. If the entire tunnel space is to be covered 360 degrees all around, the system is attached to a carrier vehicle on a specially developed, extendable and height-adjustable support arm around 2.5 m (8.2 ft) above the track bed.

For the BART project, Dibit USA used a hybridvehicle measuring system that can drive on the road and on tracks. This so-called railway truck was operated at speeds of up to 50 mph (Fig. 3).

The BART project

Dibit Measuring Technique USA, Inc. was contracted to provide a detailed inspection of approximately 40.23 km (25 miles) of the tunnels within the BART subway system in San Francisco and Oakland, CA. In February 2020, Dibit scanned 19.3 km (12 miles) of M-Line BART tunnels. This inspection was done using a system that combines Lidar scanning and photogrammetry to create a high-definition and accurate 3D model of the tunnel, which can be used to assess deficiencies, spalling, cracks, etc. These data were acquired with a customized cart that can be pushed along the rails at about 1 mph. The subway tunnel was accessed via a station platform that was used as a staging area.

Once the scanning system was assembled and the last train for the night had passed through, the system was lowered on a rail cart and onto the track. After a quick system calibration, the data acquisition was initiated. This scanning took eight four-hour nightshifts to complete.

In July 2020, Dibit scanned 13 additional miles of M-Line, R-Line and the Oakland Wye BART tunnels using the FSC 6100-SRmF10, a new photogrammetric system. In the week prior to the scanning, Dibit and BART employees worked to mount and secure the 159- kg (350-lb) scanning device onto the back of a high-rail vehicle. Each night, once the operating window began, the high-rail vehicle was set on the rails at a maintenance way and driven to the desired track area for scanning. The scanning was performed during track shutdown and took five four-hour nightshifts to complete.

The final submittals to the client included a highresolution 3D point cloud and tunnel maps showing the lining deficiencies (Fig. 4). The client also received the Dibit8 mapping software to classify features such as corrosion, leakage and cracks as tunnel information system (TIS) objects. Any tunnel features like emergency doors, lights or power supply cables can be mapped in the software (Fig. 5).

Software computation of 3D tunnel models

Processing and analysis of the 3D data are performed with Dibit8 tunneling software, which is designed for use with the high amount of measurement data generated from the high-speed FSC 6100-SRmF10 system. The database part of the Dibit-TIS is a core component of the software. In combination with the Dibitviewer, 3D tunnel data can be analyzed and visualized. Within the scope of the BART project, Dibit-TIS enabled the recording and mapping of components (blocks, niches, etc.), installations (lamps, traffic control systems, etc.) and damaged areas (cracks, spalling, etc.).

To analyze the data, images are drawn either manually on the 2D orthophoto of the tunnel or on the photo-textured, high-resolution 3D tunnel data (Fig. 5). The software allows the user to make linkages with measurement or inspection images and can then determine needed inspection or remediation protocols (e.g., injection protocols of crack remediation).

Changes in the tunnel’s surface can also be made visible by overlaying photorealistic tunnel images of different epochs (or phases, i.e., recording times) in the Dibit-viewer. Spatial and temporal changes can be quantified (4D change detection) and serve as a basis for subsequent inspection and rehabilitation activities. One of the features of the Dibit-TIS is its ability to capture the structured, spatially thematic assignment and visualization information found in complex tunnels. The information gathered is assigned to object classes and layers. Depending on the task involved, the data can be systematically included in tunnel analysis and exported by means of automated reports or used for next steps and planning.

The objects are saved with coordinate information in Dibit-TIS, and can be annotated in the form of open and closed polylines. The marking of objects as surfaces, circles and points is also possible. The positions, lengths, areas and various other parameters of individual objects can be exported in tabular form from the TIS, and thus can be further statistically evaluated.

The measurement data can also collect images of objects with freely selectable descriptive attributes, e.g., water occurrences, the appearance of sintering, crack width and much more. This allows for thematic filtering and individual aspect evaluation.

It is also possible to link a large number of different sensor data with the spatial tunnel models. In the future, these could include Georadar data, thermographic data, and multi- or hyper-spectral data, as well as conventional manual measurements and structural information. The Dibit8 software processed the BART measurement data with true-color 3D point clouds and/or textured 3D mesh models that were exported in various data formats (e.g., E57, LAS, OBJ) or ortho image data (i.e., TIFF, JPG) for further analyses in CAD (computeraided design) or BIM (building information modeling) software (Mett et al., 2019).

Conclusion

This high-speed tunnel monitoring project conducted on the BART was the first time this type of work was done in the United States. The FSC 6100-SRmF10 defines new technical standards regarding measurement velocity, measurement accuracy and image resolution. Use of this high-measurement capability has the potential to reduce and minimize tunnel closure times.

With the help of Dibit8 software, tunnel characteristics such as cracks, surface damage and tunnel installations, for example, can be analyzed in a virtual 3D environment. The results are valuable information for objective tunnel analysis conducted by engineers. By comparing tunnel measurements of different epochs, change detection (i.e., the growth of cracks) can be performed and developed for future maintenance and rehabilitation work. Furthermore, this digital tunnel surveillance project had a positive effect on tunnel safety and the proper operation of the BART system.

References

Bauer, A., Gutjahr, K., Paar, G., Kontrus, H., and Glatzl, R. (2015). Tunnel Surface 3D Reconstruction from Unoriented Image Sequences. Austrian Association for Pattern Recognition (OAGM) Workshop. 28.05.-29.05.2015. Salzburg, Austria.

Kontrus, H., Mett, M. (2019). High-speed 3D tunnel inspection. 7 S. Proceedings of the Rapid Excavation and Tunneling Conference (RETC) 2019. 16.06.-19.06.2019, Chicago, IL. SOC FOR MINING METALLURGY. ISBN: 978-0-87335-470-7.

Mett, M., Kontrus, H., and Holzer, S., (2019). Dibit TIS – Das „Proto“- BIM für den Tunnelbau. Proceedings of the 20th international geodetical week Obergurgl. 10.02.-16.02.2019 Obergurgl, Austria. Edited by K. Hanke and T. Weinold. Arbeitsbereich für Vermessung und GEOinformation. Universität Innsbruck.

Mett, M., Kontrus, H., Eder, S. (2019): 3D tunnel inspection with photogrammetric and hybrid systems. 10 S. Proceedings of the 14th International Conference on Shotcrete for Underground Support (ECI SUS XIV), Nong Nooch Gardens – Pattaya, Nov. 17-20, 2019. Thailand.