How to Validate an API 653 Tank Inspection Report Before You Accept It
An API 653 report review checks whether the inspection extent matched the interval clause, whether the floor and shell data support the minimum remaining thickness calculation, and whether the next-inspection date is arithmetic rather than habit. Many reviews confirm the report is sound. A purchaser or owner can then rely on it, which is the point.
API 653 tank reports are the most disputed inspection record in US downstream because so much money hangs off one number: the next internal inspection date. That date is the output of a chain of bottom thickness readings, a corrosion rate, a minimum remaining thickness calculation under 4.4.5.1, and the interval caps in 6.3.2, 6.3.3 and 6.4.2. Break any link and the date is unsupported. A validation walks the chain in the order the standard builds it, then states whether the report's conclusion survives. Owners commission this before a turnaround budget is locked. Purchasers commission it during diligence, when the tank's remaining life is priced into the deal and the only evidence is a PDF from a contractor they did not hire. The review does not re-inspect the tank. It tests whether the existing report can carry the weight placed on it.
Source: API Std 653 clause structure, interval logic and the minimum remaining thickness expression were verified against a published API 653 evaluation output (Codeware worked example, dated 11 November 2022) and corroborated against independent API 653 interval summaries in August 2026. Settlement provisions checked against the same output and independent Annex B commentary. Documentation requirements from 29 CFR 1910.119(j)(4)(iv), read via osha.gov in August 2026.
| Link in the chain | API 653 clause | What the report must show | Common failure |
|---|---|---|---|
| External inspection interval | 6.3.2 | The corrosion-rate-derived interval set against the five-year ceiling, with the lesser governing | Five years asserted with no calculation behind it |
| External thickness inspection interval | 6.3.3 | The derived interval set against the fifteen-year ceiling, with the lesser governing | The ceiling quoted as though it were the interval |
| Internal inspection interval | 6.4.2 | The derived interval against the applicable ceiling, extended where a release prevention barrier is documented | A longer ceiling claimed with no barrier evidenced |
| Bottom minimum remaining thickness | 4.4.5.1 | Max Or = (min(RTbc, RTip) - MRT) / (StPr + UPr), with every rate traced to readings | Corrosion rates assumed rather than measured |
| Critical zone thickness | 4.4.5.4 | The critical zone evaluated on its own minimum, separate from general bottom plate | Critical zone absorbed into a bottom-wide average |
| Bottom plate minimum thickness | Table 4.4 | MRT at the end of the chosen interval satisfies the table | Table entry read against the wrong interval length |
| Settlement evaluation | Annex B (B.2.2.4, B.3.4) | Cosine curve fit reaching an R-squared of at least 0.9, or the alternative evaluation run; edge settlement against Bew, Be and Ba | R-squared below 0.9 with no alternative evaluation performed |
Did the inspection extent match the interval requirement?
The first question is not what the inspector found. It is what the inspector was required to look at. API 653 runs three separate interval clauses, and each one takes the lesser of a calculated value and a fixed ceiling: 6.3.2 for external inspection by an authorised inspector, 6.3.3 for external ultrasonic thickness measurement, and 6.4.2 for internal inspection. A report that satisfies one clause while quietly missing another produces a valid document about the wrong scope.
This mismatch is easy to see once the reviewer lines the dates up. A published API 653 evaluation output shows the pattern clearly: the thickness interval computed as the lesser of 14.262 years and the fifteen-year ceiling, giving 14.262 years; the external interval as the lesser of 7.131 years and the five-year ceiling, giving five years; the internal interval as the lesser of 12 years and the applicable ceiling, giving 12 years. Three different next dates, from one dataset, under three clauses.
Owners get caught when a contractor performs an external inspection and the report language implies the tank is cleared overall. It is not. The external inspection satisfies 6.3.2 and nothing else. The external inspection scope in particular is worth reading closely, and is covered in atmospheric storage tank external inspection intervals.
Floor scan coverage and how MFL results get verified
API 653 does not name a floor scanning technology. What it requires is bottom thickness evidence good enough to support the minimum remaining thickness evaluation, and that evaluation needs both a top-side and an under-side corrosion rate. Magnetic flux leakage scanning is the usual way to get there. So the reviewer's first question is coverage: what proportion of the floor plate area was scanned, which plates were excluded, and on what basis. A coverage claim that exists only as a shaded sketch fails this question.
The second question is detection verification. An MFL crawler is set up against a reference plate carrying machined defects of known depth, and the report should record that setup, the scan speed used and the operator who ran it. Scan speed belongs in the record because MFL sensitivity is qualified at a speed band, and a crawler driven outside that band is not the instrument that was verified.
The third question is prove-up. MFL ranks indications; it does not measure remaining thickness. Every indication that drives a repair decision or feeds an under-side corrosion rate needs a follow-up ultrasonic reading or pit gauge measurement recorded against the same indication number. A floor report that converts MFL severity percentages straight into thickness values, with no prove-up data behind them, has fabricated the input to the whole remaining life calculation.
The settlement survey: method, point count, and the cosine fit
Settlement evaluation is where reports most often go quiet, because the arithmetic is unforgiving. API 653 Annex B evaluates shell settlement by fitting an optimum cosine curve to the elevation measurements and testing the fit. The threshold is explicit: an R-squared of at least 0.9. When the elevations do not fit, the standard does not permit the inspector to shrug. A published evaluation output shows exactly this case, recording R-squared of 0.5014 and stating that evaluation under B.2.2.5 must be performed instead.
Point count and spacing are checked next. Annex B sets a minimum of eight measurement points with a maximum spacing around the circumference, so tank diameter drives how many stations the survey needed. The same published output used 16 points on a 12.192 m tank. A survey run at eight points on a large-diameter tank has under-sampled the circumference, and the resulting curve fit describes a shape the tank may not have.
Edge settlement is evaluated separately under B.3.4, against the allowable values Bew, Be and Ba selected for the geometry. Two conditions read together matter here: where edge settlement exceeds 75 percent of the allowable and is greater than 50.8 mm, follow-up magnetic particle or penetrant examination of the shell-to-bottom and bottom welds is called for. A settlement section that reports numbers without stating whether that follow-up was triggered is incomplete. Where the survey itself is in question, lidar scanning during a refinery turnaround produces an elevation dataset that can be re-evaluated later rather than trusted once.
Shell thickness data quality
Shell thickness data is only useful if the same location is being measured each time. The reviewer checks that readings are tied to identified condition monitoring locations rather than to a course and a compass bearing, that the CML identifiers match the ones used in the previous survey, and that the report names the instrument and the surface preparation. Two surveys taken 300 mm apart on a corroding course produce a corrosion rate that is an artefact of where the technician stood.
Repeatability shows up in the numbers. A published API 653 evaluation illustrates the structure: CML A44-MP2, previous measured thickness 6.35 mm on 11 November 1990, current measured thickness 6.1 mm on 11 November 2020, 30 years between measurements. Same identified location, two dates, one difference. That is the shape a reviewable dataset takes, and it is the shape a corrosion rate can legitimately be built from.
Minimum required thickness is the other half. The report must state the required thickness it compared against and where that requirement came from, whether interpolated from the API 650 shell design tables or calculated for the course. Where the bottom is concerned, the general bottom plate minimum comes from Table 4.4 and the critical zone is evaluated separately under 4.4.5.4, on its own minimum. Averaging the critical zone into the bottom hides the location that governs.
The corrosion rate basis is the whole argument
Every disputed tank report comes down to the corrosion rate, because the rate divides into the remaining metal and sets the date. The report has to state which basis it used. A short-term rate is computed between the two most recent surveys and reacts fast to a change in service. A long-term rate is computed from the original thickness and smooths across the tank's life. A published evaluation output declares its choice on the face of the report: corrosion basis, short-term corrosion rate.
Declaring the basis is not enough on its own; the choice has to be defensible against the service history. A tank that changed product, changed water bottoms management, or had its cathodic protection interrupted has a short-term rate that means something and a long-term rate that conceals it. A reviewer reads the rate selection against what the tank has actually done since the last internal inspection, and flags a long-term rate applied over a known service change.
The arithmetic itself is trivially checkable, which is why it is checked. Corrosion rate equals the previous thickness minus the current thickness, divided by the years between measurements. In the published example, 6.35 mm minus 6.1 mm over 30 years gives 0.0083 mm per year. A reviewer recomputes every governing rate from the tabulated readings. Rates that cannot be reproduced from the report's own data table are the single most common finding in tank report validation.
Does the next-inspection date follow from the data?
This is the test the whole review builds toward. Remaining life is the current thickness minus the required thickness, divided by the corrosion rate. In the published example, 6.1 mm minus 6 mm divided by 0.0083 mm per year gives 12 years. That 12 years is then set against the ceiling in 6.4.2 and the lesser governs, producing an internal inspection date twelve years out rather than at the ceiling. A report that names a date without showing this division has asserted a conclusion.
The bottom evaluation under 4.4.5.1 is the same idea expressed as a permitted interval. The expression is Max Or = (min(RTbc, RTip) - MRT) / (StPr + UPr), where RTbc and RTip are the minimum remaining thicknesses from bottom-side and internal corrosion after repairs, MRT is the minimum remaining thickness required at the end of the interval, and StPr and UPr are the top-side and under-side corrosion rates. The result is compared against the interval the owner wants. In the published example the bottom plate permits 40 years against a specified 30, the critical zone permits 50, and the annular plate permits 37.68, so 30 years is acceptable on all three.
Notice what governs there: the annular plate, at 37.68 years, is the closest to the specified interval, and it is the component that would fail first if corrosion rates were slightly worse. A validation identifies the governing component explicitly. A report that presents only the most favourable of the three has chosen its own answer. Where the arithmetic will not close at the interval the owner needs, the routes forward are a shortened interval, a repair, or a risk-based interval built under RBI programme design.
Why tank reports are the most disputed inspection record in US downstream
Three forces meet on one document. The owner wants the longest defensible interval because an internal inspection means taking the tank out of service, cleaning it, and losing storage capacity through a season. The inspection contractor wants a report that will not be second-guessed. The insurer, the purchaser or the regulator wants evidence, and receives a PDF. Nobody in that triangle has an incentive to check the arithmetic, which is precisely why it goes unchecked.
The dispute usually surfaces late. A tank changes hands, a terminal is acquired, a claim is filed, or a corporate audit samples the file. At that point the report is years old, the inspector has moved on, and the raw data may not have been transferred with the report. Validation done at the time of issue costs a fraction of validation done under commercial pressure, and produces a record that travels with the tank.
The remedy is procedural rather than technical. Owners who require a records package alongside the report, with thickness tables, settlement elevations, floor scan data and prove-up readings attached, remove most of the dispute in advance. That practice, and the general standard a report has to meet, is set out in what makes an NDT report defensible.
When the review confirms the report
Confirmation is the frequent result and it is a usable one. The extent matched the clause, the coverage was documented, the corrosion rate reproduces from the tabulated readings, the settlement fit met the threshold, and the next date falls out of the arithmetic. The reviewer writes that finding against the clause numbers, and the owner now holds a checked conclusion instead of an unchecked one.
A confirmed report does specific work afterwards. It supports the turnaround budget that assumed the tank stays in service. It survives a purchaser's diligence question without a renegotiation. It answers an insurer asking how the interval was set. And it protects the original inspector, whose judgement has now been independently corroborated by someone with no commercial stake in the outcome.
The honest position is that you cannot tell which outcome a review will produce until it runs. Reports that look thin often survive intact because the underlying data was sound and the write-up was terse. Reports that look polished sometimes carry a corrosion rate nobody can reproduce. That uncertainty is the reason to validate, not a reason to wait.
Who does API 653 report review, and what a review package should contain
The reviewer needs competence in the standard, independence from the contractor who wrote the report, and access to the underlying data rather than the summary alone. Independence is the requirement owners most often skip. A review performed by the same firm that ran the inspection is an internal check, and it carries no weight with a purchaser, an insurer, or an audit team asking whether the original conclusion was right.
Send the reviewer more than the report. Thickness tables with CML identifiers and dates, settlement elevation data with station positions, floor scan coverage maps and prove-up readings, the previous inspection report for rate comparison, and any repair records since. A review built on the summary tables alone can check internal consistency and clause application, and it cannot check whether the summary reflects the data.
Atlantis performs independent API 653 report review for owners, purchasers and terminal operators alongside API 653 tank inspector services, and treats validation as a separate engagement from inspection so the independence holds. Pricing is affordable, accessible and fully customisable, quoted on request against your report inventory. Start with a report list at contact, or read the wider approach at report validation.
How often does a storage tank need inspecting under API 653?
Three clocks run at once. Clause 6.3.2 sets the external inspection by an authorised inspector, 6.3.3 sets external ultrasonic thickness measurement, and 6.4.2 sets internal inspection. Each takes the lesser of a corrosion-rate-derived interval and a fixed ceiling, so a corroding tank is inspected sooner than the ceiling allows. The clause-by-clause logic is worked through in API 653 tank inspection intervals explained.
What does an API 653 tank inspection report review cost?
The variables are the number of tanks, whether raw data accompanies the report or only summary tables, whether a settlement survey and floor scan are included, and whether the review has to rebuild a corrosion rate from prior reports. A single-tank purchase diligence review is a contained piece of work. A terminal-wide programme audit is not. Atlantis is affordable, accessible and fully customisable, and quotes on request against your report list.
What are the requirements for an API 653 inspection report?
The report has to carry its own evidence: thickness readings tied to identified condition monitoring locations, the corrosion rate basis stated and calculated, the minimum remaining thickness evaluation under 4.4.5.1, settlement data with the fit statistic, and the next-inspection date derived from those inputs. Where the tank sits inside an OSHA process safety programme, 29 CFR 1910.119(j)(4)(iv) additionally requires the date, the person, the equipment identifier, the test description and the results.
Who does tank inspection report review near me?
Report validation runs from the document package, so the reviewer works remotely and proximity matters only when you add a site verification visit. The requirements that matter are independence from the inspection contractor who wrote the report and current competence in the standard. Atlantis provides API 653 tank inspector services and independent report review for US owners, purchasers and terminal operators.
Can a purchaser have a tank inspection report reviewed before closing?
Yes, and this is the most common purchaser use. The seller supplies inspection reports, the buyer prices remaining life into the offer, and the buyer has no relationship with the contractor who produced them. A validation states whether the reported next-inspection date follows from the data, and whether the inspection extent matched the clause. That converts a PDF of unknown provenance into a checked input for a valuation.
Does a review overturn the original API 653 inspector?
Usually it confirms them. Independent review of a competent report finds the extent correct, the corrosion rate defensible and the arithmetic sound, and says so. That outcome is worth having, because the owner now holds a checked conclusion rather than an unchecked one. Where the review does differ, the finding is written against a clause and a number so the original inspector can answer it directly.