Industries / Construction and property / Heavy civil / Tunneling
Boring machines
A tunnel boring machine (TBM) cuts the whole round face of a tunnel in one pass with a rotating cutting wheel, carries the spoil away behind it and, in weak or wet ground, holds the ground up while the lining is built.
5
Milestones achieved, from 1882 to 2015.
3
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1882
A rotary borer in chalk
Captain Thomas English of Dartford, Kent, patented a rotary boring machine in 1880. Driven by compressed air, it pared the face of soft chalk with steel cutters on two revolving arms. Machines of this kind drove the first trial headings for a tunnel under the English Channel, about 7 ft (2.1 m) across, and by the end of 1882 the heading at Shakespeare Cliff near Dover had reached 2,040 yards (about 1,865 m). The British government, worried that a tunnel could carry an invading army, had the Board of Trade stop the work through the courts. Subterranea Britannica says the machine used was English's, though it is usually named after Colonel Frederick Beaumont; the Institution of Civil Engineers credits both men.
Credit
Thomas English; Frederick Beaumont; Sir Edward Watkin's Submarine Continental Railway Company
1971
Slurry holds up the face
In loose, wet sand and gravel the face of a tunnel flows in unless something holds it. John Bartlett of the consulting engineers Mott, Hay and Anderson patented a bentonite tunneling machine in 1964: a sealed bulkhead behind the cutting wheel keeps a slurry of bentonite clay and water under pressure against the face, so the crew behind works in ordinary air instead of compressed air. A prototype drove an experimental tunnel for London Transport at New Cross, southeast London, in 1971 and 1972, at up to 4 m per 10-hour shift. Wayss & Freytag of Germany used a similar machine, the Hydroshield, for a sewer under Hamburg in 1974, and slurry machines have since reached 17.6 m across, on the road tunnel from Tuen Mun to Chek Lap Kok in Hong Kong, which Ingenia, the Royal Academy of Engineering's magazine, calls the largest TBM ever built. Gary Brierley's history in Tunnel Business Magazine dates a slurry shield in Houston to 1960 (with no pressure on the face) and a Mitsubishi one in Japan to 1970, so which machine came first depends on the definition.
Credit
John Bartlett; Mott, Hay and Anderson; London Transport
1972
Boring while lining goes in
A double shield TBM has two steel shields that slide into each other like a telescope: the rear one grips the rock while rings of concrete lining are bolted up inside it, and the front one keeps cutting. Carlo Grandori's Italian contractor SELI won the Orichella and Timpagrande tunnels in Calabria in 1972, 4 km of 4.3 m tunnel in partly decomposed granite, as the first job for such a machine, developed with the Robbins Company. By the account of Tunnels & Tunnelling International, a SELI double shield on the Yindaruqin irrigation project in China, around 1991, advanced 1,200 to 1,300 m a month while installing a lining of six-sided segments.
Credit
Carlo Grandori (SELI); The Robbins Company
1992
Wrong machine for wet rock
Sweden's rail administration, Banverket (now Trafikverket), began twin rail tunnels through the Hallandsås ridge in 1992. The contractor Kraftbyggarna started with Hallborr, an open hard-rock TBM, which has no shield and needs rock that stands up by itself. The ridge's rock was cracked, broken and full of water, and Kraftbyggarna soon gave up on the machine, blasted about 3 km and left in 1995. Heavy water inflows followed, and in 1997 a toxic sealant leak stopped the work for years. Government-ordered studies from 1998 to 2000 concluded that the tunnels needed a watertight lining and a shielded machine. A Herrenknecht machine named Åsa, run by Skanska and Vinci, finished the second tube on 4 September 2013, and the two 8.7 km tunnels opened in December 2015.
LessonAn open machine needs rock that stands up and stays dry; where the ground may be broken and under water pressure, the machine and the lining have to keep the water out.
Credit
Banverket and Trafikverket; Kraftbyggarna; Skanska-Vinci; Herrenknecht
2015
One machine, four modes
Kuala Lumpur's first metro line runs through karst limestone, rock dissolved into caves and cracks, where slurry pressed against the face drains away into the voids and the ground above can sink. Herrenknecht and the contractor MMC-Gamuda developed the variable density TBM, which switches between earth pressure balance mode (holding the face with the dug soil itself) and slurry mode, and can thicken its slurry with limestone dust, a mixture worked out with Ruhr University Bochum, so that it does not escape. The first, S-774, 6.62 m across, launched in May 2013 and bored 4.4 km. The client, Mass Rapid Transit Corporation, marked the end of tunneling on the line on 21 April 2015, and the machine had won Technical Innovation of the Year at the 2014 International Tunnelling and Underground Space Awards in London.
Credit
Herrenknecht; MMC-Gamuda; Ruhr University Bochum
Watch for
Linghang under the Yangtze
China Railway Tunnel Group says Linghang, a shield machine with a 15.4 m cutterhead, bored a continuous 11.325 km under the Yangtze River from Chongming Island in Shanghai to a shaft near Taicang without any intermediate shaft, a world record for the longest continuous underwater drive by a 15-meter-class machine. About 1.8 km remained, to be finished by the end of 2026. In December 2025 the project's deputy equipment manager said the machine's system "enables autonomous operation through intelligent analysis and decision-making."
Achieved whenthe tunnel holes through and a body outside the builder, such as the International Tunnelling and Underground Space Association, confirms the length record, and a published log shows how much of the steering the system did without operators.
Claimed by
China Railway Tunnel Group, July 2026 (autonomy: December 2025)
A "smart brain" for TBMs
A Hunan University team led by Chen Renpeng, working with China Railway Construction Heavy Industry, says its operating system reads a TBM's sensors, forecasts ground settlement and warns if the machine may get stuck, and recommends settings. On a large-diameter machine in a Shanghai metro project it says the system assisted with over 90% accuracy in operational decisions and kept the machine within 30 mm of its line, and on a Guangzhou project it raised monthly progress by 5%.
Achieved whena paper or an outside party compares drives with and without the system on the same machine and ground, and states what counts as an accurate decision.
Claimed by
the research team, through the Xinhua news agency, January 2026
Prufrock's mile a week
The Boring Company says its Prufrock machines launch from the surface without a dug pit, build the tunnel lining while they keep mining instead of stopping every 5 ft (1.5 m), and run with nobody in the tunnel, controlled from an operations center in Bastrop, Texas. It is targeting 1 mile (1.6 km) a week and an all-in cost under $5 million a mile, and says two hard-rock versions are mining limestone in Nashville.
Achieved whena tunnel owner or regulator publishes measured advance rates and costs for completed drives.
Claimed by
The Boring Company, on its Prufrock page as of November 2026
Open
01
An open cutter wear table
A public table of disc cutter use on finished hard-rock TBM drives, gathered from papers, theses and project reports. Disc cutters are the steel rings on the cutting wheel that crack the rock, and changing them stops the machine. Each row gives the tunnel, machine diameter, cutter size, rock type and strength, abrasiveness test results such as the Cerchar abrasivity index and the Cutter Life Index, and cutters changed per kilometer or per cubic meter. Then score the published wear models, those of the Colorado School of Mines (CSM), the Norwegian University of Science and Technology (NTNU) and Gehring, against it.
The briefCredit · unclaimed02
Settlement models across cities
A test of whether a model that predicts how much the ground surface sinks above a TBM tunnel, trained on one city's records, still works in another city.
The briefCredit · unclaimed
Open 01 · the brief
An open cutter wear table
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public table of disc cutter use on finished hard-rock TBM drives, gathered from papers, theses and project reports. Disc cutters are the steel rings on the cutting wheel that crack the rock, and changing them stops the machine. Each row gives the tunnel, machine diameter, cutter size, rock type and strength, abrasiveness test results such as the Cerchar abrasivity index and the Cutter Life Index, and cutters changed per kilometer or per cubic meter. Then score the published wear models, those of the Colorado School of Mines (CSM), the Norwegian University of Science and Technology (NTNU) and Gehring, against it.
Why it is still open
Writing in the Journal of Mining and Environment in 2020, Zahiri Galeshi and colleagues note that because TBM makers treat the data as sensitive and wear is hard to measure while boring, "the researchers have a limited access to the cutter wear data"; their own study used more than 1,000 cutter records from one tunnel, Kani-Sib in Iran. Model checks found in this session use one or a few tunnels each, and a GitHub search found code for monitoring cutter wear on one machine but no table across projects.
Built with what exists
Open-access papers and theses (NTNU Open, Politecnico di Torino's thesis archive, open journals); an AI model that reads PDFs and pulls out figures, with a person checking each row (see the Note "Seven more flagged"); Python with pandas.
First step
Enter the Kani-Sib figures and two more tunnels from open papers, and code the NTNU model for one of them.
Done looks like
At least 30 drives from at least eight countries, each row linked to its source, a second person agreeing with 19 of 20 rows picked at random, and a published table of each model's median error in cutters per kilometer.
Open 02 · the brief
Settlement models across cities
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A test of whether a model that predicts how much the ground surface sinks above a TBM tunnel, trained on one city's records, still works in another city.
Why it is still open
Open records are rare. Settlement data from Guangzhou Metro Line 9, 17 variables of geology and machine settings, were published in Data in Brief in 2020 under a CC BY license. A 2025 study of Bangkok's MRT Blue Line in the journal Underground Space posted its dataset and code on GitHub, with no license stated. A French student project on GitHub used synthetic data because, its description says, "the real data is confidential." A 2023 chapter by R. P. Chen, N. Zou, H. Z. Cheng and H. N. Wu trained a model on shield-tunnel records pooled from several Chinese cities, and its summary says the pooled model beat models trained on one city each; its records were not released. No study found in this session trains a model in one city and tests it on another city's public records.
Built with what exists
The two datasets; Python with scikit-learn and LightGBM.
First step
Map both datasets' columns onto one list of variables, and ask the Bangkok authors for a license to reuse their data.
Done looks like
Published errors in millimeters for a model trained in each city and tested in the other, set beside the error within each city, with the shared variable list and code released.