BRENNEN

Industries / Construction and property / Heavy civil / Paving

Quality testing

Quality testing checks the finished pavement in the field: how dense it was rolled, how smoothly it rides and how much it bends under a load.

5

Milestones achieved, from 1952 to 2019.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1952

    A beam measures the bend

    The Benkelman beam is a long lever with a dial gauge. Its tip goes between the rear dual tires of a loaded truck, typically 80 kN (18,000 lb) on a single axle, and as the truck drives away the gauge records how far the pavement springs back, a reading called deflection. A. C. Benkelman developed it while assigned to the WASHO Road Test of the Western Association of State Highway Officials; Pavement Interactive gives the year as 1952. It was used at the AASHO Road Test in the late 1950s and by the Asphalt Institute for overlay design in the 1960s and 1970s. FHWA notes that it is cheap but slow and labor-intensive, and that it gives one reading rather than the shape of the dip around the load.

    fhwa.dot.govpavementinteractive.org

    Credit

    A. C. Benkelman

  2. 1976

    Dropped weight replaces the truck

    A falling weight deflectometer (FWD) drops a weight onto a plate on the road and records with a row of sensors the shape of the dip the blow makes, from which engineers work out the stiffness of each layer. The first Dynatest FWD was built in Denmark in 1976 and reached the United States in 1981, and KUAB has made FWDs since 1976. FHWA says the FWD "emerged in the 1980s as the worldwide standard for pavement deflection testing." Its Long-Term Pavement Performance (LTPP) program has collected FWD measurements on more than 900 pavement sections since the late 1980s.

    fhwa.dot.govfhwa.dot.gov

    Credit

    Dynatest; KUAB

  3. 1986

    One scale for roughness

    Roughness readings from different instruments could be compared only through imprecise conversion formulas. In May and June 1982, teams from Brazil, Britain, France and the United States ran their instruments over 49 test roads in Brasília, from asphalt to earth, in the International Road Roughness Experiment, and Belgium's road research center joined the analysis. From the results, World Bank Technical Paper 45, published in January 1986, proposed the International Roughness Index (IRI). It is computed from a measured road profile as the total up-and-down travel of a standard model car's suspension per distance driven at 80 km/h, given in m/km, and most roughness equipment can report it.

    documents1.worldbank.org

    Credit

    Michael W. Sayers and Thomas D. Gillespie of the University of Michigan Transportation Research Institute; Cesar A. V. Queiroz of Brazil's Road Research Institute; the World Bank and GEIPOT, Brazil's transport planning agency

  4. 2007

    Swapped cores cost $8 million

    Many state contracts raise or cut an asphalt crew's pay by the density of cores, cylinders drilled out of the new pavement and tested in a laboratory. In December 2007 Gohmann Asphalt and Construction of Clarksville, Indiana, paid $6.7 million to the US government, $1.1 million to Kentucky and $362,165 to Indiana, without admitting wrongdoing. The claims were that on Interstate 64 and other work from 1997 to 2006, its workers checked spots in advance with density gauges made by Troxler Electronic Laboratories and, when state examiners were not present, swapped cores they knew would fail for passing ones. The US and Indiana said swapping touched 10% to 40% of inspections on 132 contracts, where density set 40% of pay in Kentucky and 35% in Indiana. It came to light through a suit by a former crew supervisor.

    Lessona test result is only as good as the custody of the sample.

    enr.com

    Credit

    Paul Roederer, the whistleblower; the US Attorney for the Western District of Kentucky

  5. 2019

    Radar maps the density

    A dielectric profiling system (DPS) runs radar antennas along the new asphalt without touching it and reads the surface's dielectric value, an electrical property that changes with how much air the mix holds. A few cores or laboratory specimens convert the readings into density. Instead of a handful of cores at random spots, it gives a continuous density record of the whole lane, including the joints between lanes. AASHTO published a provisional standard for it, PP 98-19, and FHWA lists pilot projects in Alaska, Florida, Maine, Minnesota, Nebraska and Ohio and a pooled study with nine states. FHWA cites research that each 1% of air voids above 7% costs about 10% of pavement life.

    fhwa.dot.govinfotechnology.fhwa.dot.gov

    Credit

    SHRP2 project R06C (Stephen Sebesta, Tom Scullion and Timo Saarenketo, 2013); Minnesota DOT, which hosts the pooled study TPF-5(443)

Watch for

  1. Radar calibrated without cores

    Michigan DOT engineers, writing with two authors from Geophysical Survey Systems, a ground-penetrating radar maker, say DPS readings can be calibrated from as few as two laboratory-compacted pucks of the paving mix instead of field cores. Across 29 datasets from 22 Michigan projects, puck calibrations varied by less than 0.5% air voids in 28, while more than half of the core calibrations varied by more than 1.0%. They trace the outliers to surface moisture during testing and recommend pucks as the main reference, with cores kept for checks.

    Achieved whena transportation agency accepts or pays for paving on DPS density calibrated from pucks, or an outside group repeats the comparison on its own projects.

    api.crossref.org

    Claimed by

    Fawaz Kaseer and Ethan Akerly (Michigan DOT), Roger Roberts and David Cist (Geophysical Survey Systems), March 2026, in Transportation Research Record

Open

  1. 01

    Match cores to their holes

    A check that each core tested in the laboratory is the one drilled at the chosen spot. The inspector photographs the pavement surface at each marked spot before drilling, the laboratory photographs the top of each core on arrival, and software matches the stone pattern in the two photos, which differs at every spot, and flags any core that matches no photo.

    The brief
    Credit · unclaimed
  2. 02

    Copied numbers in test reports

    A screen that reads a season of asphalt test reports and flags results copied from earlier reports, exact or near-exact repeats of whole sets of numbers for different samples, scored on reports with planted copies.

    The brief
    Credit · unclaimed

Open 01 · the brief

Match cores to their holes

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A check that each core tested in the laboratory is the one drilled at the chosen spot. The inspector photographs the pavement surface at each marked spot before drilling, the laboratory photographs the top of each core on arrival, and software matches the stone pattern in the two photos, which differs at every spot, and flags any core that matches no photo.

Why it is still open

Custody of cores rests on people and seals. New Jersey's construction handbook requires an inspector to witness the coring, record the location and seal the core box with numbered department seals, after which the contractor delivers the box to the laboratory; it treats a form filled in for an unwitnessed core as falsifying a public record. At Gohmann, swaps on 132 contracts came to light only through a former employee's suit. Searches of state core-custody rules and of research on matching pavement images found no use of photo matching.

nj.govenr.com

Built with what exists

A phone camera with a printed scale and color card; OpenCV feature matching (SIFT or ORB keypoints); a vision model as a second matcher; old asphalt slabs from a recycler, cored or sawn into pieces on private ground; Python.

  1. First step

    Photograph 20 marked spots on one old asphalt slab, saw out the pieces, photograph them indoors, and test whether ORB matching pairs them correctly.

  2. Done looks like

    200 spots photographed in place and 200 pieces photographed again under different light and angle, with 30 pieces secretly swapped or relabeled by a second person. Published: the share of true matches found, the share of swaps caught and the false alarms, with every photo.

Open 02 · the brief

Copied numbers in test reports

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A screen that reads a season of asphalt test reports and flags results copied from earlier reports, exact or near-exact repeats of whole sets of numbers for different samples, scored on reports with planted copies.

Why it is still open

In March 2026 two Ohio asphalt companies agreed to pay $30 million, without admitting liability, to settle claims that from 2012 to 2025 they submitted mix design results "using data copied from prior mix tests," along with false quality-control results, on Ohio DOT work. Private individuals brought the claims under the False Claims Act's whistleblower provisions. The nearest screening work found, a 2020 Boise State University thesis by Mostofa Najmus Sakib, used Idaho's hidden record of every data entry to flag results changed after the fact, after an FHWA case study had found that 74% of the Idaho Transportation Department's test results did not match the contractors'. That method needs an edit log and does not look for copies.

enr.comscholarworks.boisestate.edu

Built with what exists

Test reports obtained from a state transportation department under its public-records law; pdftotext, with a language model and a person checking samples to read scanned sheets; Python and pandas for exact and near-duplicate matching across reports; 200 made-up reports to test on first.

  1. First step

    Write 200 made-up reports in one state's format with 20 planted copies, build a first exact and near-match screen, and file a records request for one district's reports from a year.

  2. Done looks like

    1,000 real reports with 50 planted copies made by a second person, some exact and some with one or two numbers changed. Published: the share of planted copies found, false flags per 1,000 reports, and the screen's code.