Industries / Construction and property / Commercial and residential construction / Flooring
Preparing the subfloor
Preparing the subfloor is the work of making the base under a floor dry, flat, clean and sound enough to hold the finished floor, whether that base is a concrete slab or wood panels on joists.
7
Milestones achieved, from c. 25 BCE to 2018.
0
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
c. 25 BCE
A floor in layers
The Roman architect Vitruvius set out how to build a floor from the ground up. The soil is checked to be solid and rammed where it is loose. On timber floors no wall may touch the underside of the boards, and two nails go through the board edges into every joist so that no edge warps; he warns that oak warps when damp and cracks the paving above. Rammed layers of rubble and lime mortar follow, and only then is the paving laid "to rule and level" and rubbed smooth.
Lessona floor fails from below, so the base is made solid before anything goes on it.
Credit
Vitruvius, De architectura, Book 7
1972
Gypsum poured over wood
Gyp-Crete, a gypsum underlayment, came into use. An underlayment is a thin layer poured or laid over the subfloor to give a flat base, and this one went over the wood subfloors of apartment buildings. Maxxon, which makes it, says Gyp-Crete started the gypsum underlayment industry and has been the standard over wood subfloors in multifamily building since 1972.
Credit
Maxxon
1978
Self-leveling cement arrives
ARDEX K 15, a cement-based underlayment that flows out flat under its own weight, reached the US market in 1978. ARDEX Americas calls it "America's first self-leveling cement underlayment" and says it was faster than any other product then on the market.
Credit
ARDEX, a group that traces its roots to NORWAG-Werke, founded in 1949 in Witten, Germany
1998
Calcium chloride test standardized
A 2010 trade report says calcium chloride had been used since the 1940s to judge whether a slab was dry enough for resilient flooring such as vinyl. ASTM published the method as F1869 in 1998. A dish of anhydrous calcium chloride, a salt that soaks up water, sits under a sealed dome on the slab, and the water it gains is reported as pounds given off by 1,000 square feet of concrete in 24 hours.
Credit
ASTM International Committee F06
2002
Probes inside the slab
ASTM F2170 measures relative humidity (RH) inside the concrete with a probe set in a drilled hole, instead of the vapor leaving the surface. Wagner Meters' summary of the standard says the hole goes to 40% of the slab's thickness for a slab drying from one side, or 20% for a slab drying from both sides, and a 2010 trade report said F2170 had been replacing calcium chloride as the preferred test since 2002. The standard says too much moisture after the floor goes down can cause "debonding, peaking" and microbial growth. Failures still happen: an FCNews column (2015) put the yearly cost of moisture-related flooring failures in the US at $300 million to $1 billion, and CTLGroup, a testing firm, puts it at $2.4 billion for commercial owners in North America (2023).
astm.orgwagnermeters.comrandrmagonline.comfcnews.netctlgroup.com
Credit
ASTM International Committee F06
2010
Ruled out for lightweight concrete
ASTM changed F1869 to bar the calcium chloride test on all lightweight aggregate concrete, trade press reported in October and November 2010. The same reports said that both the science behind the test and its reliability "have remained problematic".
Lessona surface test that suits one kind of slab can mislead on another, so the test has to match the concrete.
Credit
ASTM International Committee F06
2018
Readings in one day
A revised F2170 cut the wait before an official reading from 72 hours to 24. A February 2018 release from Wagner Meters, a probe maker, says ASTM made the change after an interlaboratory study that Wagner commissioned in 2014, which found readings at 24 hours statistically equivalent to readings at 72 hours.
Credit
ASTM International
Open
01
An open slab-drying record
A public dataset of in-slab relative humidity, logged on real slabs from the day the concrete is placed to the day flooring goes down, with the facts that drive drying: thickness, mix, vapor retarder, the date heating or cooling started, and room temperature and humidity. On top of it, a scoreboard that tests any drying forecast, from a simple curve fit to a machine-learning model, on slabs the forecast has not seen.
The briefCredit · unclaimed02
Moisture readings without drilling
A model that estimates the in-slab RH an F2170 test would report from readings that need no hole: a handheld impedance moisture meter reading (it senses the top inch of concrete), surface temperature, room conditions, and slab age and thickness. It is trained and tested on paired readings taken beside real F2170 probes.
The briefCredit · unclaimed
Open 01 · the brief
An open slab-drying record
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public dataset of in-slab relative humidity, logged on real slabs from the day the concrete is placed to the day flooring goes down, with the facts that drive drying: thickness, mix, vapor retarder, the date heating or cooling started, and room temperature and humidity. On top of it, a scoreboard that tests any drying forecast, from a simple curve fit to a machine-learning model, on slabs the forecast has not seen.
Why it is still open
The standard test gives a snapshot, not a forecast: ASTM F2170 says results show the slab's condition "only at the time of the test and in the specific locations tested". Forecasts exist in private hands; a 2024 US patent to Egate Smart Building Innovation Oy covers predicting drying from wireless sensors in and around the slab. A search found lab calibrations and models, but no public field dataset and no published field accuracy for any forecast.
Built with what exists
Logging RH probes made for F2170 holes, such as Wagner Meters' Rapid RH L6 sensors with cellular remote monitoring; a room temperature and humidity logger; Python with pandas and scikit-learn; Zenodo or GitHub to publish.
First step
Write a one-page data format, ask two testing firms or flooring contractors for anonymized F2170 logs they already keep, and set probes in one slab now being built, with the owner's permission.
Done looks like
Logs from at least 30 slabs on at least 5 sites, released openly, with each forecast scored by its error in days in predicting when a slab reaches the RH limit in the flooring maker's instructions, against a straight-line extrapolation as the baseline.
Open 02 · the brief
Moisture readings without drilling
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A model that estimates the in-slab RH an F2170 test would report from readings that need no hole: a handheld impedance moisture meter reading (it senses the top inch of concrete), surface temperature, room conditions, and slab age and thickness. It is trained and tested on paired readings taken beside real F2170 probes.
Why it is still open
The guide for impedance meters, ASTM F2659-23, says it "is not intended to provide quantitative results as a basis for acceptance" of a floor, and sends users to F1869 or F2170. F2170 needs a hole drilled to 40% of the slab's thickness and a 24-hour wait. A search found no study that pairs handheld readings with F2170 results and tests a predictive model.
Built with what exists
Any commercial impedance meter used to ASTM F2659; F2170 probes; an infrared thermometer; scikit-learn, for example gradient-boosted trees; a spreadsheet or phone form for logging.
First step
On one slab that already has F2170 probes, take impedance readings in a small grid around each probe every time the probe is read.
Done looks like
At least 200 paired readings from at least 20 slabs, published; on slabs held out from training, the share of estimates within 3 RH percentage points of the F2170 reading and the share of wrong pass or fail calls against a stated limit. A negative result counts if the data are released.