Industries / Construction and property / Engineering and architecture / Geotechnical engineering
Foundations
Foundation work chooses and checks what carries a structure's weight into the ground, from a concrete raft spread under a building to piles driven or drilled down to firmer layers.
6
Milestones achieved, from 1913 to 2016.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1913
A grain elevator tips
A new grain elevator at Transcona, near Winnipeg, Canada, failed in 1913 while it was being filled, because the load on its concrete raft exceeded the shear strength of the clay beneath. Later analyses found that the bearing capacity predicted by theory, 307.4 kPa, was close to the 296.9 kPa observed at failure. Crews then dug under the high side, sank 14 piers to bedrock and jacked the structure upright by 17 October 1914, raising it about 3.7 m; it is still in use.
LessonThe raft was likely designed to tolerate large settlement, but its exposure to a deep shear failure was "neither understood nor expected", in the words of a 1916 account by Allaire that the society quotes.
Credit
A. W. Skempton (1951) and J. Blatz and K. Skaftfeld (2003) for the analyses, as the Canadian Geotechnical Society reports
1942
Formulas no better than averages
From 1892 engineers often judged a driven pile's capacity from how far it sank under each hammer blow, using formulas such as A. Wellington's Engineering News formula, written for timber piles and drop hammers. In 1941 an American Society of Civil Engineers (ASCE) committee compared several formulas with many static load tests; in 1942 R. B. Peck pointed out that simply assigning every pile the tests' mean failure load, 91 tons, came statistically closer than any of the formulas.
LessonThe Federal Highway Administration (FHWA) still reports poor correlation and wide scatter for such formulas, and US bridge specifications bar them for piles above 600 kips (about 2,670 kN).
Credit
R. B. Peck; the ASCE committee on pile foundations
1960
Pile driving on a computer
E. A. L. Smith published a way to model the whole driving system (hammer, cushions, pile and soil) as linked masses and springs that a digital computer steps through after each blow. Texas highway researchers used it to find driving stresses and cut damage to concrete piles, and the FHWA funded programs built on it; its WEAP program, by G. G. Goble and F. Rausche (1976), lives on as the commercial GRLWEAP.
Credit
E. A. L. Smith
1972
Testing piles as they drive
At the Case Institute of Technology in Cleveland, a project funded from 1964 by the Ohio Department of Transportation and the FHWA, led by R. H. Scanlan and G. G. Goble, measured force and acceleration during each hammer blow and worked out the pile's capacity from them. Ohio began using the results around 1968, commercial use began in 1972 with the Pile Driving Analyzer, and FHWA demonstration projects in the 1980s spread the method across the US. In Sweden the Commission on Pile Research had measured long precast concrete piles during driving in 1959, which Massarsch and Fellenius call the first such measurements.
Credit
G. G. Goble, Frank Rausche and colleagues at Case; Pile Dynamics, Inc., founded in 1972
1983
A jack inside the pile
A bidirectional test casts a hydraulic jack cell low inside a pile and pushes the part above up and the part below down at the same time, so each part serves as the other's reaction. P. C. A. F. Elisio described such a test in Brazil in 1983, and Bengt Fellenius writes that the first commercial version was Brazilian; in the US Jorj Osterberg worked on his cell from 1984 and patented it in 1987, by the account of his former student John Schmertmann, and the two groups collaborated after Osterberg learned of the Brazilian device in 1988. Schmertmann put the record at 63,000 kips (about 280 MN), on a shaft 3 m (10 ft) wide and 67 m (220 ft) long for South Korea's Incheon Bridge in 2005, against 9,000 kips for the largest earlier test he knew of.
Credit
P. C. A. F. Elisio; Jorj Osterberg; Loadtest Inc
2016
A tower settles and leans
San Francisco's 58-storey Millennium Tower began sinking and tilting soon after it was finished in 2009, and by 2016 it had settled about 18 in. (46 cm) and leaned about 24 in. (61 cm) toward its northwest corner, by Engineering News-Record's (ENR) account. Simpson Gumpertz & Heger designed a fix of 18 new piles, 24 in. (61 cm) across and about 300 ft (91 m) deep to bedrock, tied into the mat foundation along two streets and jacked to take load off the old friction piles; the load was transferred in three stages in 2023. Monitoring reports for the owners' association, posted by the city, check settlement against a limit of 0.25 in. (6 mm) a year, and the April 2026 report lists only instrument faults as reportable.
LessonThe repair itself moved the building: the first six new piles caused unexpected settlement and a three-month pause while drilling methods were changed.
Credit
Simpson Gumpertz & Heger and engineer Ronald Hamburger; contractor Shimmick; Slate Geotechnical for the monitoring
Watch for
Grouted screw pile load record
Chance says its GroutForce screw displacement pile, a steel screw pile installed with a column of grout around its shaft, was installed 48 ft (14.6 m) deep in Vancouver, Canada, and load-tested to 676 kips (3,007 kN) in compression with 1 in. (25 mm) of movement, the highest load test result of any Chance product; the test stopped before the pile failed because the reaction anchors were pulling on the load frame. It also says the grout column lets the shaft be 50% or more shorter than other pile types for the same strength.
Achieved whenload tests run by engineers outside the company, in several soils, confirm the capacity, and a shorter GroutForce pile is shown to carry the same load as a conventional pile at the same site.
Claimed by
Chance, in an article it contributed to Pile Buck, July 2025
Open
01
Pile echo tests, ground-truthed
A public set of low-strain pile integrity tests, in which a tester taps the top of a finished concrete pile with a hand-held hammer and an accelerometer records the echo, each labelled with what was later found in that pile by coring, excavation or cross-hole sonic logging. Add a baseline model that reads the echo traces and scores it on the set.
The briefCredit · unclaimed02
A register of reused foundations
An open table of buildings put up on reused foundations, giving for each the original and new loads, the records that were found, the tests done (integrity tests, load tests), whether new piles had to be added, and how the building has performed since.
The briefCredit · unclaimed
Open 01 · the brief
Pile echo tests, ground-truthed
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public set of low-strain pile integrity tests, in which a tester taps the top of a finished concrete pile with a hand-held hammer and an accelerometer records the echo, each labelled with what was later found in that pile by coring, excavation or cross-hole sonic logging. Add a baseline model that reads the echo traces and scores it on the set.
Why it is still open
Reading these traces, called reflectograms, "remains a manual and expertise-intensive task", wrote the authors of a February 2026 study in Scientific Reports, which reviews many machine-learning models for the job. Its own data is available only on request, and a search found no public set of field records labelled with what was found in the piles.
Built with what exists
Field records from testing firms or highway agencies; Python with scikit-learn or PyTorch; the 2026 study's recurrent neural network as a baseline.
First step
Ask two highway agencies or testing firms for integrity records of piles that were later cored or dug out, and agree an open licence for them.
Done looks like
At least 500 field records from at least three sites, released under an open licence with each pile's verified condition, and a model's accuracy on held-out records compared with that of two experienced testers reading the same traces without seeing the answers.
Open 02 · the brief
A register of reused foundations
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open table of buildings put up on reused foundations, giving for each the original and new loads, the records that were found, the tests done (integrity tests, load tests), whether new piles had to be added, and how the building has performed since.
Why it is still open
Reuse often rests on archive research and due diligence because records of old foundations are limited, as Arup's Henry Tayler sets out in a 2020 guide for the Institution of Structural Engineers. At a March 2026 conference, DMT reported load tests on the driven precast piles of an 18-storey building from 2002 in Aarhus, Denmark, whose owners hoped to add six storeys: the piles reached failure before the planned maximum force, and DMT concluded that standard calculations did not match site conditions and that more capacity was needed. The EU-funded RuFUS project, led by Britain's Building Research Establishment, gathered case studies into a best-practice handbook and conference proceedings in 2006, but these are books for sale, fixed at that date. A search found no public register that collects such projects and their outcomes.
Built with what exists
Published case studies and conference talks; an AI model that pulls fields out of PDFs, such as the open-weights Mistral Large 4 (see the Note "A trillion open parameters"); a shared spreadsheet; the reuse checks in CIRIA's guide C653 as the list of fields.
First step
Enter the case studies presented at the 2026 Conference on Foundation Decarbonization and Re-use, starting with the Aarhus test.
Done looks like
50 projects from at least five countries, each with a source, and a published count of how many had load tests and how many needed new piles.