Surviving an NDT Audit in Steel and Primary Metals

What the auditor is actually testing in a metals plant

An audit of the NDT function in a steel mill, plate shop, forge or tube mill is not a technical examination of the metal. It is a test of whether the paper can be reproduced. The auditor assumes the product is probably fine and sets out to establish whether the controls that would have caught a problem actually operated, and whether the evidence that they operated still exists in a form a stranger can follow without help from the person who created it.

The method is consistent across client audits, registrar audits, class society surveys and regulator visits. The auditor takes a certificate and traces it backwards to the heat, the mill test report, the NDT record, the person who signed the accept call and that person's certification status on the date of signature. Then the auditor picks a piece of steel off the floor and traces it forwards to a certificate. Anything that fails in either direction becomes a finding, and the direction that fails more often is forwards.

Findings against personnel certification are the most damaging because they operate retrospectively. If a technician's certification had lapsed, every accept and reject that technician signed during the lapse is open to question, and the cost is not the correction. It is the extent of condition review that determines how many heats, how many certificates and how many customers are affected, and how many of them have to be told.

The written practice is the document that fails first

Under ASNT SNT-TC-1A, certification is an employer function performed against the employer's own written practice. The practice states which edition of SNT-TC-1A is invoked, which methods and levels are covered, the training hours, experience hours and examination requirements for each, the vision requirements and their intervals, the recertification interval and its basis, and every place where the employer's requirements differ from the recommended practice.

The failure modes are predictable. A practice copied from a template that names a superseded edition. A practice covering methods the plant does not perform while omitting one it does, with automated eddy current on a tube line the usual omission. Deviations from SNT-TC-1A taken in daily operation but never documented as deviations. No Level III approval signature, or a signature from a Level III whose own certification in that method has since expired. Where ISO 9712 is invoked instead, the structure is different because certification is issued by a certification body rather than the employer, and practices that blend the two schemes without saying which applies to whom are a reliable finding.

Rebuilding a written practice is not a drafting exercise. It has to describe what the plant actually does, because the auditor will read it and then walk out onto the floor to watch. ASNT Level III consulting on this document is normally the first item of work, and it almost always changes the personnel files as a consequence.

The evidence chain: heat number to certificate, in both directions

Traceability in primary metals breaks at predictable points. The first is marking transfer at the cutting table. When a plate is cut, the heat and plate identification has to be transferred to every piece before separation, and somebody has to record that it was done and by whom. Once the offcuts have been moved to the scrap bay the link is gone, and no amount of subsequent paperwork restores it honestly.

The second is the interface between the test laboratory and the certificate. Mechanical results, chemistry and NDT results are usually generated in different systems and the certificate is assembled from all three. Where the laboratory operates under ISO/IEC 17025, the scope of accreditation has to actually cover the test and the product form in question; accreditation for tensile and impact testing does not extend to ultrasonic examination of forgings. Where testing is subcontracted, purchaser approval of the subcontractor is frequently specified in the order and frequently never obtained.

The third is retention. Purchase orders commonly require records to be retained for longer than the plant's own retention policy, and nobody discovers the mismatch until a customer asks for a ten-year-old certificate. Holding certification, procedure and NDT records against the heat and the product form in a controlled inspection management system is what makes retrieval possible at that distance. A folder structure on a file server is storage, not a records system.

Product form drives the procedure, and this is where mills slip

Steel and primary metals is unusual in how tightly the examination method is tied to the product form. Plate is examined to ASTM A435 or A578 for straight beam ultrasonics and A577 for angle beam, with different acceptance levels inside A578 selected by the purchase order. Forgings go to A388 with their own calibration and scanning requirements, and to A275 for magnetic particle. Castings go to A609. Tubular product goes to E213, E309 or E570 depending on whether the mill line is ultrasonic, eddy current or flux leakage based.

The slip is procedural cross-contamination. A plate procedure used to make accept calls against a different specification's acceptance class. A forging examined with a calibration established for plate geometry. An A578 Level B disposition made from a procedure qualified only to A435. Each of these looks defensible on the shop floor, where the operator is competent and the equipment is calibrated, and each is indefensible in front of an auditor who simply reads the scope statement on the front of the procedure.

Procedures also have to state the acceptance criteria they are used against, not merely the technique. A procedure that says evaluate to customer specification pushes the decision onto the operator and hands the auditor an immediate finding, because the accept call cannot be reconstructed from the record without asking the person who made it what he had in mind that day.

Automated NDT on mill lines and the bracketing rule

In-line ultrasonic, eddy current and flux leakage systems on tube and pipe mills carry a verification requirement different in kind from manual NDT. Performance is demonstrated against a reference standard containing artificial discontinuities, typically notches of specified depth and orientation or drilled holes, and the specifications require the production run to be bracketed by acceptable checks at both ends.

The start-of-run check is almost always performed. The end-of-run check is the one that goes missing at a shift change or at the end of a campaign, and the consequence written into the specification is severe: product back to the last acceptable verification must be re-inspected. On a high-throughput mill that can be several hundred lengths, and the commercial choice between re-inspection and some other disposition has to be made by someone with the authority and the technical basis to make it defensible afterwards.

There is a related personnel issue that auditors find quickly. The person operating the line is often certified as an equipment operator rather than as a Level II in the method, yet is recording accept and reject decisions. Either the written practice covers that role explicitly, with defined training, examination and stated limits on interpretation authority, or the practice does not describe what the plant actually does and the auditor will say so.

What the Level III signs, and what he must not sign

The Level III signs the written practice and its revisions; NDT procedures and their qualification and demonstration records; technique sheets; the qualification determination behind each certification decision; examination papers and their grading; independent data review reports; and technical position letters used to answer an auditor's question on method, technique or acceptance criteria.

The Level III does not sign the mill test certificate, which is the manufacturer's declaration. He does not sign the quality management system's conformity declaration. He is not the API 510, 570 or 653 inspector of record and does not sign in that capacity, and Atlantis does not provide certification training for those API inspector credentials. He is not a process safety management auditor. And under SNT-TC-1A he does not certify your people: certification is issued by the employer on the basis of the Level III's qualification determination.

That last distinction puts plants in difficulty more often than any other. Certificates issued on a consultant's letterhead, for personnel employed by the mill, are a nonconformity under an employer-certification scheme and auditors know exactly where to look. The correct arrangement is that the outside Level III is retained by the employer, administers and grades the examinations, makes the qualification determination, and the employer issues the certification, with the retention agreement and the Level III's own credentials in the file. Where certification is intended to sit outside the employer entirely, ISO 9712 through a certification body is the route, and our NDT training and certification work supports both schemes.

Gap review before the audit versus remediation after a finding

A gap review before an audit and remediation after a finding involve much the same technical work and entirely different economics. Before the audit, sampling is adversarial by choice: you deliberately pull the oldest procedure, the newest technician, the heat with the messiest history and the mill line that had a shift change mid-run. Anything found is corrected under normal revision control, personnel are re-examined without any public record, and the plant re-baselines quietly.

After a finding, the clock belongs to the auditor. Major findings typically carry a short response window, corrections must be accompanied by root cause analysis and objective evidence of effectiveness, and the correction alone is never sufficient. The expensive obligation is extent of condition: identifying everything produced under the defective control, deciding whether it remains acceptable on the evidence retained, and determining who has to be notified.

There is also a reputational asymmetry that plants underestimate. A gap closed before an audit is invisible to everyone outside the plant. A finding closed after an audit sits in the registrar's file, is revisited at every surveillance visit for the remainder of the cycle, and client auditors ask about previous findings as a matter of routine when they arrive.

Representation on the day: technical questions only

During the audit itself the Level III's role is narrow and should stay narrow: answer technical questions on method, technique, procedure qualification, acceptance criteria and personnel qualification, and produce the evidence that supports the answer. He does not answer for the management system, for commercial arrangements, or for another function's records, and saying so plainly is not evasion.

The value is in preventing the two failure modes that turn a small question into a finding. The first is over-answering, where a plant employee volunteers history the auditor did not request and opens an entirely new thread that then has to be closed. The second is under-answering, where nobody in the room can explain why a DGS sizing curve was used instead of DAC, so the auditor writes it up as unsubstantiated rather than accepting a technique that was in fact correct.

Where a finding has already been raised and the accept and reject decisions themselves are in question, the work moves to independent review of the inspection data: re-reading the records, and where necessary the retained raw data, against the correct acceptance criteria, so that the extent of condition is decided on evidence rather than on assumption. If you have an audit date and want the gap review completed before it, start here.

Automated mill-line NDT and the bracketing sample the audit will ask for

Automated ultrasonic and eddy current testing on a rolling mill line runs continuously, generating a data volume that makes retaining every individual result impractical for most mills, so programmes retain summary pass/fail dispositions with periodic full-record captures instead. An audit tests exactly that compromise: whether the retained bracketing samples — the full records kept at defined intervals to demonstrate the automated system's ongoing performance — actually bracket the production run in question, or whether the interval was set so wide that a system drift between captures would go undetected until the next scheduled sample.

The finding that recurs is a bracketing interval chosen for line throughput convenience rather than derived from the system's demonstrated stability. A mill that captures full records once per shift on a line that can drift within an hour under certain alloy or gauge combinations has a bracketing scheme that looks compliant on paper and does not actually protect against the failure mode it exists to catch.

Where this becomes an audit finding with real exposure is the extent-of-condition question that follows: if a drift is discovered at one bracketing capture, every coil produced since the previous capture is potentially affected, and the mill needs a documented method for identifying and dispositioning that entire population — not just the coil that happened to fail the sample that caught the drift.

Request a consultation