Industries / Industrial equipment and services / Industrial services / Industrial cleaning / Tank cleaning
Inspection after cleaning
Once a tank is empty and clean, inspectors measure how much steel corrosion has taken from its floor, walls and roof, and decide whether it can go back into service and when it must next be opened.
6
Milestones achieved, from 1988 to 2025.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1988
A rebuilt tank splits open
On 2 January 1988 an oil storage tank of 4 million gallons at Ashland Petroleum's Floreffe terminal near West Elizabeth, Pennsylvania, collapsed as it was filled to capacity for the first time since being rebuilt. It had stood near Cleveland, Ohio, for over 40 years, was taken down and re-erected at Floreffe in 1986, and held No. 2 fuel oil. About one million gallons of oil reached the Monongahela River 25 miles upstream of Pittsburgh and contaminated the water supply of many towns. The National Bureau of Standards (NBS) found a brittle fracture that started at a flaw present before the rebuild, in steel too brittle at the shell's temperature of about 32°F (0°C) to stop a crack, steel that did not meet the toughness rules of the tank standard API 650. NBS wrote that neither inspection during building nor recurrent inspection in service can guarantee that no flaw exists.
LessonInspection cannot find every flaw, so a tank's steel must be tough enough to leak before it breaks.
Credit
investigated by the National Bureau of Standards (now NIST), report NBSIR 88-3792, June 1988
1991
A code for old tanks
In January 1991 the American Petroleum Institute (API) published the first edition of API 653, "Tank Inspection, Repair, Alteration, and Reconstruction", for tanks already in service. Earl Crochet, who took the first inspector exam in April 1993, writes that the Ashland collapse led directly to it, and that before it repairs followed the construction standard API 650 while inspection rested on personal experience. In 1999 the US pipeline safety regulator wrote its second edition into its rules for breakout tanks as an alternative to yearly inspection. The fifth edition of November 2014 has 189 pages and 14 annexes, against 78 pages in 1991.
Credit
American Petroleum Institute
1996
Magnets sweep tank floors
Magnetic flux leakage (MFL) inspection magnetizes the steel and senses the field that leaks out where metal has been lost, so a scanner pushed across a floor can find corrosion on the underside that no one can see. Eddyfi, which now makes such scanners, traces the method on tank floors to D. H. Saunderson's 1988 account of an MFE scanner, and writes that before it inspectors often took five thickness readings per plate, covering under 1% of a floor. In 1996 D. M. Amos of MFE Enterprises wrote that MFL had been used on tank floors for only about ten years, that tank floors are never flat, and that accept-or-reject rules based on signal height had proved unreliable for sizing defects, and he called for methods that miss fewer significant defects.
LessonOn a floor that is not flat, the height of a scanner's signal does not reliably say how deep a pit is.
Credit
D. M. Amos, MFE Enterprises, in Materials Evaluation
2002
Integrity tests become law
The US Environmental Protection Agency's oil spill prevention rule (Spill Prevention, Control, and Countermeasure), as rewritten on 17 July 2002 and amended in 2006 and 2008, requires facilities it covers to test or inspect each aboveground oil container for integrity on a regular schedule and after material repairs, with the frequency and type set by industry standards, and to keep records that allow comparison over time. It names visual inspection, hydrostatic testing (filling with water), radiography, ultrasound and acoustic emission as examples.
Credit
US Environmental Protection Agency
2014
A floor nobody inspected
On 9 January 2014 a mixture containing crude MCHM, a chemical used to clean coal, leaked through two holes about 0.75 and 0.4 inches across in the floor of tank 396 at Freedom Industries in Charleston, West Virginia, and reached the Elk River upstream of a water treatment plant's intake; water use was restricted for about 300,000 residents in nine counties, and the US Chemical Safety Board (CSB) estimated almost 11,000 gallons leaked. The holes were pitting corrosion from the inside of a tank installed in 1938, which rain likely entered through holes in its roof. The CSB found no record that Freedom or the previous owner had inspected the tanks, and no leak detection. The inspection made afterwards for the CSB found the leaking hole within 3 inches of the shell, in the zone API 653 calls critical, and could not classify the scanner's readings on the muddy floor.
LessonA tank that is never emptied, cleaned and inspected can corrode through from the inside with no warning.
Credit
US Chemical Safety Board, final report May 2017; inspection by Powers Engineering and Inspection, 2014
2025
Europe's tank guide adds robots
EEMUA, an association of plant owners and operators, first published Publication 159, its guide to the inspection, maintenance and repair of above-ground flat-bottomed tanks, in 1994. Edition 6, published in April 2025, adds guidance on new inspection methods including robots and drones, three levels of integrity assessment, guidance on change of service, and updated remaining-life calculations for pitting in tank bottoms. EEMUA says regulators worldwide use the guide as an example of good practice, and that it complements API 653.
Credit
EEMUA (Engineering Equipment and Materials Users Association)
Watch for
Longer gaps between inspections
Internal inspections are the costly ones, because the tank must be emptied and cleaned first. The US Pipeline and Hazardous Materials Safety Administration (PHMSA) proposed on 2 June 2026 to let operators of breakout tanks (tanks that absorb surges or hold oil between legs of a pipeline) set internal inspection intervals by risk-based inspection, as the fifth edition of API 653 allows. Its rules now bar that method and cap the interval at 20 years, or 10 when corrosion rates are not known; the proposal would cap the first internal inspection at 20 years and later ones at 25. PHMSA estimates savings of $24.5 million to $125.7 million a year at a 3% discount rate; API and the Liquid Energy Pipeline Association put them at $220 million a year, and the Pipeline Safety Trust asked for added conditions on robotic inspection. Comments closed on 3 August 2026, and no final rule had been published by mid-November 2026.
Achieved whena final rule takes effect.
Claimed by
PHMSA, June 2026
Open
01
Clean enough to scan
A public test of how much residue a tank floor can carry before floor scanners miss pits, so that cleaning crews and inspectors share a measurable definition of clean. Steel plates a quarter inch (6.4 mm) thick, the floor thickness of the Freedom Industries tank, get pits of known depth machined into the underside, in open plate, beside lap welds and within 3 inches (76 mm) of a mock shell wall, the critical zone of API 653. Each plate is scanned bare and then under measured layers of a stand-in residue (sand, clay and wax) and of loose scale. Raw scan files, pit maps and residue thicknesses are published together, which also gives software a set to learn from. It is built in a workshop, with no tank, oil or confined space.
The briefCredit · unclaimed02
Test the inspection gaps
A public table of measured tank floor corrosion, one row per internal inspection, drawn from API 653 reports that are already public or can be obtained under public-records laws: tank age, product, original floor thickness, minimum remaining thickness found, whether the floor had a liner or cathodic protection (a small electric current that slows corrosion from the soil side), and the next interval the inspector set. From it, compute how many tanks would have reached API 653's minimum floor thickness before 20, 25 and 30 years, the caps now under debate.
The briefCredit · unclaimed
Open 01 · the brief
Clean enough to scan
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test of how much residue a tank floor can carry before floor scanners miss pits, so that cleaning crews and inspectors share a measurable definition of clean. Steel plates a quarter inch (6.4 mm) thick, the floor thickness of the Freedom Industries tank, get pits of known depth machined into the underside, in open plate, beside lap welds and within 3 inches (76 mm) of a mock shell wall, the critical zone of API 653. Each plate is scanned bare and then under measured layers of a stand-in residue (sand, clay and wax) and of loose scale. Raw scan files, pit maps and residue thicknesses are published together, which also gives software a set to learn from. It is built in a workshop, with no tank, oil or confined space.
Why it is still open
In the 2014 API 653 inspection of the leaking Freedom Industries tank, the inspector recorded MFL indications on every floor plate of a muddy floor and wrote that they could not be classified unless the floor was blasted "immaculately clean"; the hole that leaked lay in the critical zone. Eddyfi, which makes Floormap scanners, writes that the industry has lived with "a significant mismatch between actual inspection coverage and reported inspection coverage" near the shell. Yang and co-authors (2023) tested an MFL sensor array on one 4 mm plate and published no data, and the largest open MFL data set, PipeMFL-240K, comes from pipelines. A web search found no public tank floor scan set.
Built with what exists
Steel plate pitted by a machine shop with CNC milling or electrical discharge machining (EDM, which cuts metal with sparks); a rented MFL floor scanner such as Eddyfi's Floormap X or an MFE model; an ultrasonic thickness gauge or phased-array probe to confirm each pit's depth; a depth gauge to measure the residue layers; Python with NumPy and PyTorch; Zenodo or Hugging Face to publish the files.
First step
Have ten pits machined into one 1.2 m by 2.4 m plate, rent a scanner for a day, and scan the plate bare and under 1 mm and 3 mm of residue.
Done looks like
At least 200 pits from 20% to 80% of plate thickness, a quarter of them in the critical zone and a quarter beside welds, each scanned under at least four residue conditions with at least two scanner models; detection rate and false calls published for each condition, with the thickest residue at which at least 90% of pits 40% deep are still found; all files under an open licence.
Open 02 · the brief
Test the inspection gaps
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public table of measured tank floor corrosion, one row per internal inspection, drawn from API 653 reports that are already public or can be obtained under public-records laws: tank age, product, original floor thickness, minimum remaining thickness found, whether the floor had a liner or cathodic protection (a small electric current that slows corrosion from the soil side), and the next interval the inspector set. From it, compute how many tanks would have reached API 653's minimum floor thickness before 20, 25 and 30 years, the caps now under debate.
Why it is still open
PHMSA's 2026 proposal notes that neither API 653 nor API RP 580 sets reliability targets, takes its caps of 20 and 25 years from intervals that appeared in editions of API 653, and asks for comment on whether to set minimum standards; industry's $220 million savings estimate rests on "an average assessment interval". Inspection reports are mostly private, but some are public, such as the 2014 API 653 reports on three Freedom Industries tanks posted by the CSB, which give a product-side corrosion rate of 0.01 inch a year for tank 396.
Built with what exists
The regulations.gov public API to pull comments and attachments in dockets PHMSA-2025-0019 and PHMSA-2025-1271; public-records requests to state tank programs; a model built for reading long documents, such as Meta's Muse Ember, to pull fields from scanned reports (see the Note "The jobs agents repeat"); an agent that cites the page each number came from, such as one built with Cursor's agent kit (see the Note "Findings that cite notes"); Python with pandas and lifelines, a library for survival analysis.
First step
Enter the three Freedom Industries reports, and download the attachments filed in the two PHMSA dockets.
Done looks like
At least 300 internal inspections from at least ten owners or agencies, each field linked to the page it came from; two people check 30 randomly chosen rows; the share of tanks that reached minimum floor thickness before 20, 25 and 30 years published with confidence intervals, by product and by floor protection.