API 653 — Tank Inspection, Repair, Alteration, and Reconstruction

What API 653 covers, and the boundary it draws

API 653 provides minimum requirements for maintaining the integrity of welded or riveted, non-refrigerated, atmospheric-pressure aboveground storage tanks after they have been placed in service. Its jurisdiction runs from the foundation through the bottom, shell, structure, roof and appurtenances, and out along each nozzle to the face of the first flange, first threaded joint or first welding-end connection. Past that face, the piping belongs to API 570 and the relevant ASME B31 code. Drawing that line on a plot plan before a turnaround is the cheapest hour anyone spends.

What it does not cover matters as much. It is not a design standard: new tank design and fabrication remain with API 650. It does not cover pressure vessels, underground tanks, or low-pressure storage designed to API 620, though its principles are widely borrowed. It does not cover the design of the foundation, only the evaluation of what that foundation has done to the tank. And it does not address process piping, dike integrity or spill containment beyond their effect on the tank shell and bottom.

Where a calculated result under API 653 is unacceptable, the code does not dead-end. It permits a fitness-for-service assessment to API 579-1/ASME FFS-1, a reduction in maximum fill height, a change of service to a lower specific gravity, or a shortened inspection interval. Choosing among those options is an engineering decision that has to be written down and signed, and it is exactly the point at which most owners bring in ASNT Level III consulting to author the evaluation and the supporting NDE technique.

Three clocks, not one: how the interval structure works

Operators commonly speak of a tank being "on a five-year cycle", which conflates three independent requirements. The first is routine in-service monitoring from grade by operations personnel, recommended at intervals not exceeding one month, needing no certification and catching the things that actually leak: seal weeps, shell distortion, foundation washout, settled areas, blocked drains and failing coatings. The second is the formal external inspection by an API 653 authorised inspector, due at the lesser of five years or RCA/4N. The third is shell thickness measurement, which has its own clock entirely.

That third clock is the one most often missed. External ultrasonic thickness measurement is due at five years when the corrosion rate is unknown, and at the lesser of RCA/2N or fifteen years once it is established. Because a five-year external visual inspection can be closed out without a single UT reading, a programme can look compliant for a decade while the shell corrosion rate remains genuinely unmeasured. Auditors find this by asking for the UT data set, not the inspection report.

Interval arithmetic collapses without clean historical thickness data at fixed, repeatable locations. Readings taken at approximately the same place produce corrosion rates that swing by a factor of two between campaigns, and the interval swings with them. Tank programmes that hold up under audit keep numbered CMLs with photographs and elevations, and store them in something more durable than a spreadsheet — the case for an inspection data management system is made almost entirely by API 653 interval calculations.

Shell evaluation: the arithmetic that decides fitness

The minimum thickness for a shell course is calculated as tmin = 2.6 × D × (H − 1) × G / (S × E), with D the tank diameter in feet, H the height in feet from the bottom of the course containing the corroded area to the maximum design liquid level, G the specific gravity of the stored product, S the allowable stress and E the joint efficiency. The (H − 1) term is the one-foot method carried over from API 650. Getting H wrong by measuring from the tank bottom rather than from the bottom of the affected course is the single most common calculation error.

API 653 allows higher allowable stresses than API 650 permits for new construction, because the tank exists and has been proved by service. For the bottom course, S is the lesser of 0.80 times the yield or 0.429 times the tensile strength; for upper courses, the lesser of 0.88 times yield or 0.472 times tensile. E is the original joint efficiency, though 1.0 may be used where the corroded area sits clear of welds by the margin the code specifies. Applying API 650 stresses to an in-service evaluation is conservative but expensive, and it has retired sound plate.

Averaging is where evaluations quietly go wrong. Thickness within a corroded area may be averaged, but only over a defined critical length L = 3.7 × √(R × t2), where R is the tank radius in inches and t2 the least thickness in the area. Averaging across an arbitrary patch of plate because it "looks like one area" inflates the result and produces a fitness conclusion that will not survive review. When a third party recalculates the same data set and reaches a different answer, the difference is nearly always averaging length or the value of H.

Bottom plates, MRT and the critical zone

Bottom evaluation is governed by minimum remaining thickness, and MRT is a forward projection rather than a measurement. You start from the least remaining thickness found, subtract the topside corrosion rate multiplied by the proposed interval, subtract the underside corrosion rate multiplied by the same interval, and compare what is left against the floor: 0.100 in. for a bottom with no release prevention barrier, 0.050 in. where an RPB with leak detection is installed. A bottom measuring 0.140 in. today can fail a twenty-year interval and pass a ten-year one, on identical data.

Underside corrosion rate is the term nobody can measure directly, and the honest answer is that it comes from history, from the soil-side environment, from cathodic protection performance and from previous bottom replacements. Where a tank has no CP, no liner and no prior internal, assuming a low underside rate to justify a long interval is an assumption an auditor will attack immediately. Magnetic flux leakage floor scanning finds the pit population, but a prove-up UT programme on the deepest indications is what turns a scan into a defensible number.

The critical zone deserves separate attention: the annular ring of bottom plate within three inches of the shell. Minimum thickness there is the smaller of one half the original bottom plate thickness or 0.100 in., excluding corrosion allowance, and lap welds are not permitted in that band under the current construction rules. Corrosion in the critical zone is structurally serious because it sits where the shell-to-bottom fillet transfers load, and it is where the tank actually tears when it fails.

Repair, alteration and reconstruction

Repairs must be authorised before work starts, performed to written procedures, welded by welders qualified to ASME Section IX, and examined to the criteria the code specifies for each repair type. Lap-welded patch plates on the shell, insert plates, hot taps, bottom overlays and annular ring replacement each carry their own dimensional limits, minimum spacing from existing welds, corner radii and examination requirements. The recurring finding is not bad welding; it is repairs executed as maintenance work orders, with no authorisation record and no examination package.

An alteration changes the tank's physical dimensions or configuration and pulls in design review. Reconstruction — dismantling and reassembly, including relocation — is the most heavily controlled activity in the standard, requiring plumbness, roundness, peaking and banding checks, full examination of the reassembled seams, and a hydrostatic test. Owners who buy a used tank and move it are performing a reconstruction whether or not anyone uses that word in the purchase order.

Examination of repair welds runs on vacuum box testing at a partial vacuum of at least 3 psi (approximately 21 kPa) for bottom seams, magnetic particle or liquid penetrant for shell repairs, and radiography or ultrasonics where the code or the referencing specification calls for volumetric coverage. Because these reports are written under time pressure at the end of a turnaround, they are also where documentation defects concentrate, which is why many owners route the completed package through independent report validation before the tank returns to service.

Settlement, distortion and brittle fracture

Annex B handles settlement, and it separates uniform settlement, planar tilt and out-of-plane distortion. Uniform settlement stresses nothing; planar tilt matters mainly for the floating roof and the shell-to-bottom junction; out-of-plane settlement is the one that induces shell stress, and it is evaluated by fitting a cosine curve through elevations measured at equally spaced points around the circumference and comparing the deviation from that fitted curve against a permissible value that scales with plate height, radius and material properties. Edge settlement near the shell is assessed separately, using the breakover distance and depth.

The measurement itself decides the answer. Optical levelling at eight points around a 150-foot tank produces a curve fit with too few degrees of freedom to be meaningful, yet it is still common. Denser point sets change conclusions, and 3D laser scanning services now supply shell verticality, roundness and a full settlement profile in a single occupation of the site, which also gives the repair contractor an as-built to work from rather than a set of hand-recorded elevations.

Section 5 addresses brittle fracture, and it becomes live whenever a tank changes service to a colder product, a heavier specific gravity, or a higher fill height than its original basis. The evaluation depends on plate thickness, material toughness data or its absence, minimum design metal temperature and the tank's operating history. A tank that has held its current product for thirty years without incident carries a strong service-proof argument; the same tank switched to a new duty does not, and the evaluation has to be redone rather than assumed.

The findings that recur in tank-programme audits

Five findings account for most of what auditors write. First, the external visual clock is current but the shell UT clock has expired, because nobody separated them. Second, the internal interval was set at twenty years with no defensible underside corrosion rate behind it. Third, MRT was computed on today's thickness with no projection to the next inspection. Fourth, repairs were executed without written authorisation or a completed examination record. Fifth, corrosion rates were computed from thickness readings taken at locations that do not repeat between campaigns.

A sixth is subtler and increasingly common: the inspection report is technically correct but is not the document the calculation was based on. Spreadsheets circulate, get revised, and diverge from the report of record, so the number defended in the audit cannot be reproduced from the report on file. The fix is not more inspection; it is a single system of record where the readings, the calculation inputs, the resulting intervals and the next-due dates live together and the report is generated from them.

Where tanks fall inside an OSHA Process Safety Management covered process, the mechanical integrity element imposes its own requirements for written procedures, trained personnel, inspection and test records with results, and correction of deficiencies before further use. API 653 supplies the technical criteria; PSM supplies the management system that has to be shown to be operating. Owners handle this with a defined mechanical integrity software workflow so that deficiency correction and interval compliance are both evidenced without a manual reconciliation before every audit.

Where API 653 hands off to other documents

API 653 does not stand alone. API 650 supplies the construction basis and the design formulas it modifies. API 651 governs cathodic protection and API 652 covers bottom linings. API 575 offers guidance on inspection practice for atmospheric and low-pressure tanks. API 579-1/ASME FFS-1 handles the assessments API 653 declines to prescribe, including local thin areas, blisters, weld misalignment and crack-like flaws. API 580 and API 581 supply the risk-based methodology that can extend intervals up to, but never beyond, the twenty-year ceiling.

On the execution side, ASME Section V provides the NDE methods and ASME Section IX the welder and procedure qualification. Personnel performing NDE are qualified to SNT-TC-1A or an equivalent scheme under the owner's written practice, and the authorised inspector holds API 653 certification. These are separate qualifications doing separate jobs, and conflating them — treating a certified tank inspector as automatically qualified to perform and interpret ultrasonic examination, or the reverse — creates a personnel-qualification finding that is trivially easy for an auditor to prove.

The practical consequence for an owner is that a tank programme is really a document-control problem wearing inspection clothes. Thickness data, corrosion rates, calculation revisions, repair authorisations, examination records, settlement surveys and next-due dates all have to reconcile to one another on demand. Getting the technical criteria right is the smaller half of the work. If you want a second set of eyes on either half, start with a consultation rather than at the next turnaround.

Frequently asked

Who is allowed to perform an API 653 inspection?

External and internal inspections must be performed by, or under the direction of, an API 653 certified authorised inspector. NDE such as ultrasonic thickness measurement, magnetic flux leakage floor scanning, magnetic particle and vacuum box testing is performed by technicians qualified under the owner's written practice to SNT-TC-1A or an equivalent scheme. The two qualifications are distinct, and each has to be evidenced separately in the inspection record.

Is API 653 mandatory, or is it a recommended practice?

API 653 is a standard, not a recommended practice, and it is written in mandatory language. Whether it is legally binding depends on jurisdiction: some US states adopt it directly for storage tank programmes, EPA SPCC rules require integrity testing in accordance with industry standards, and many owner specifications and insurers invoke it contractually. In practice, once your written procedure cites API 653, you are held to it in full.

What happens if the measured thickness is below the calculated minimum?

Nothing automatic. The code permits several routes: a fitness-for-service assessment to API 579-1/ASME FFS-1, a reduction in maximum fill height, a change of service to a lower specific gravity product, repair or plate replacement, or a shortened inspection interval. What is not permitted is continued operation at the original conditions without a documented, engineered and authorised evaluation supporting the decision.

Does a risk-based inspection assessment let me skip the internal inspection?

No. An RBI assessment conducted in line with API 580 and API 581 can extend the internal inspection interval and can justify a different mix of examination techniques, but it cannot push the interval past the twenty-year ceiling and it cannot remove the requirement. The assessment itself has to be documented, reviewed by a competent team, and reassessed on a defined cycle rather than performed once and cited indefinitely.

How do API 653 and API 570 divide responsibility at a tank nozzle?

API 653 jurisdiction ends at the face of the first flange, the first threaded joint, or the first welding-end connection on each nozzle. Everything downstream is piping, inspected under API 570 to the applicable ASME B31 code. Splitting the scope explicitly on a marked-up plot plan before a turnaround prevents the classic gap where both programmes assume the other covered the nozzle-to-piping transition.