Industries / Industrial equipment and services / Industrial services / Industrial cleaning
Chemical cleaning
Chemical cleaning dissolves scale, rust and other deposits from boilers, heat exchangers and pipes by filling them with, or pumping through them, acids, chelants (chemicals that grip metal ions and hold them in solution) or solvents.
4
Milestones achieved, from 1932 to 2010.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1932
An acid that spares steel
Hydrochloric acid dissolves scale, but it also attacks the steel underneath. John J. Grebe and Ross T. Sanford of Dow Chemical patented adding a small amount of an arsenic compound, an inhibitor that slows the acid's attack on metal, so that oil wells could be treated "without material injury to the casing"; the patent was granted on 13 September 1932. In a 1960 review in the journal Corrosion, the consulting chemist Charles M. Loucks described how contract chemical cleaning of plants grew after inhibited acids were introduced for cleaning.
LessonLoucks warned that acid cleaning is hazardous enough, to people and to equipment, that letting the job on the lowest bid, without both sides agreeing on what is to be done, invites trouble.
Credit
John J. Grebe and Ross T. Sanford, Dow Chemical
1953
Power stations switch to acid
I. Davies of Britain's Central Electricity Generating Board (CEGB) wrote in 1966 that until about 1953 its new boilers got only a simple alkaline boil-out before they were first put into service. After that the CEGB adopted acid cleaning, first on boilers and later on the heaters that warm water on its way to the boiler. Cleaning with acid after a plant had entered service was, he wrote, "still controversial."
Credit
the CEGB, as reported by I. Davies in Anti-Corrosion Methods and Materials (1966)
1987
Solvents for nuclear steam generators
In a pressurized-water nuclear plant, the steam generators are large heat exchangers whose tubes collect deposits and sludge. In 1978 the Electric Power Research Institute (EPRI) began a program, sponsored by the Steam Generator Owners Group, to remove that sludge chemically with as little corrosion as possible. It produced three solvents: one for iron oxides and one for metallic copper, applied in turns with water rinses between, and one for corrosion products in the narrow gaps called crevices. A 1987 paper by P. M. Olson and C. A. Swift reported that two US utilities had used the EPRI process "with very positive results," and that several European utilities had used a process from Kraftwerk Union (KWU). Westinghouse engineers later reported that the EPRI solvents, combined with pulses of nitrogen, removed more than 38,000 lb (about 17 metric tons) of deposits from the steam generators of Virginia Power's two Surry units in 1994.
Credit
EPRI and the Steam Generator Owners Group; Kraftwerk Union; Westinghouse
2010
Pipes cleaned with fuel gas
On 7 February 2010, at the Kleen Energy power plant under construction in Middletown, Connecticut, workers were cleaning debris out of new fuel pipes with a "gas blow": forcing natural gas through them at about 650 psi (45 bar) and venting it to the open air near the main building. Just over 2 million standard cubic feet (about 57,000 m³) of gas was released over the morning. It found an ignition source and exploded, killing six workers and injuring at least 50. The US Chemical Safety Board (CSB) noted a similar gas blow that ignited at Calpine's Wolfskill Energy Center in California in 2003, without injuries, and recommended cleaning with air, or with pigs (plugs pushed through the pipe) driven by air. The National Fire Protection Association (NFPA) wrote a new standard, NFPA 56, in less than 24 weeks; it prohibits using flammable gas to clean piping, and its permanent 2014 edition was approved in July 2013.
LessonClean new pipe with air, not with the fuel it will carry.
Credit
investigated by the CSB; NFPA
Watch for
A reactor system cleaned whole
Holtec International says the work to restart the Palisades nuclear plant in Michigan, shut down in 2022, included "a first-ever comprehensive system cleaning" that "significantly reduced dose rates," meaning radiation levels, in sensitive areas of the plant. The cleaning was followed by passivation, which brought the reactor's main water circuit to normal temperature and pressure to restore protective surfaces on the metal. A July 2026 Holtec update lists the job as primary system chemical decontamination and passivation. Holtec gives no dose figures and does not say in what sense the cleaning was a first.
Achieved whenHoltec or the US Nuclear Regulatory Commission publishes dose rates measured before and after the cleaning.
Claimed by
Holtec International, March 2026
Open
01
Score solvent picks against experts
A public test set of published chemical cleaning jobs. Each case gives the deposit's lab analysis (for example, the shares of iron oxide, copper, calcium, silica and organic matter), the metals of the equipment, the cleaning solution the engineers chose and whether it worked. Then a scored test of whether a software tool or AI model, given only the analysis and the metals, picks the same family of solution: inhibited hydrochloric acid, acid with fluoride for silica, a chelant such as ammoniated EDTA, an organic acid such as citric or formic, or an organic solvent.
The briefCredit · unclaimed02
Know when cleaning is done
A tool that reads the measurements taken during a chemical clean (iron and copper dissolved in the solution, acid strength, pH and temperature) and says when the deposit is used up and the solution can be drained, tested on a public set of cleaning curves.
The briefCredit · unclaimed
Open 01 · the brief
Score solvent picks against experts
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test set of published chemical cleaning jobs. Each case gives the deposit's lab analysis (for example, the shares of iron oxide, copper, calcium, silica and organic matter), the metals of the equipment, the cleaning solution the engineers chose and whether it worked. Then a scored test of whether a software tool or AI model, given only the analysis and the metals, picks the same family of solution: inhibited hydrochloric acid, acid with fluoride for silica, a chelant such as ammoniated EDTA, an organic acid such as citric or formic, or an organic solvent.
Why it is still open
The choice rests on a lab analysis of the deposit and on rules of thumb, such as using an organic solvent when a deposit is more than 30% organic. Hexxcell and Dow presented a modeling method in 2019 that predicts how well given solvents will clean refinery heat exchangers from a few lab-measured values, but it is a company's method. A May 2026 project on GitHub suggests acid strength and a split between alkaline and acid stages for cleaning geothermal wells from the brine's chemistry, by fixed rules, without a test against engineers' choices. A search found no public set of cases against which any selection tool can be scored.
Built with what exists
Papers from the Heat Exchanger Fouling and Cleaning conferences, which are free to download at heatexchanger-fouling.com, and patents with worked examples; a document-reading AI model to copy each deposit analysis and chosen solution into a table, with a person checking every row (see the Note "Seven more flagged"); a model that gives probabilities for fixed-answer questions such as "which family of solution," for example Gemini Decision 1 (see the Note "A thousand questions, one pass"); a spreadsheet; Zenodo to publish the set with a DOI.
First step
Go through the papers listed on heatexchanger-fouling.com and list every one that reports both a deposit analysis and the cleaning solution used.
Done looks like
At least 100 cases from at least 20 sources, each linked to its source. Two chemical cleaning engineers who have not seen the published choice pick a solution family for 50 of the cases, which sets the level of agreement between people. Any tool's agreement with the published choices is reported beside that level.
Open 02 · the brief
Know when cleaning is done
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A tool that reads the measurements taken during a chemical clean (iron and copper dissolved in the solution, acid strength, pH and temperature) and says when the deposit is used up and the solution can be drained, tested on a public set of cleaning curves.
Why it is still open
Engineers watch the iron in the cleaning solution rise until it levels off, as the iron-against-time curves in a 1996 patent by Wayne W. Frenier of HydroChem Industrial Services show. King Saud University patented a sensor in 2019 that raises an alarm when a layer of magnetite, a black iron oxide, on a steel probe has dissolved during acid cleaning; the patent calls continuous monitoring of iron in the cleaning fluid time-consuming and laborious. A search found no published, tested rule for calling the end point from logged readings, and no public set of logged cleans.
Built with what exists
WebPlotDigitizer, a free tool that reads data points off published graphs, to pull iron-against-time curves from patents and papers; Python with SciPy to fit curves that level off; household-safe bench runs, such as rusty steel coupons in citric acid descaler with iron test strips read every 30 minutes.
First step
Digitize the iron-against-time curves in US patent 5,529,637, which has 23 sheets of drawings.
Done looks like
At least 50 published curves digitized and shared; a rule that, from the first part of each curve, predicts the time at which the iron stops rising to within 1 hour on at least 80% of curves it was not built on; 10 bench runs logged and published.