UT thickness data: gauge to record without retyping

An ultrasonic thickness reading passes through more hands than any other number in an integrity program. The gauge shows 0.312. The technician writes 0.312 on a sheet, or says it to a helper, or types it into a tablet. The sheet goes to a data entry clerk who types it into a spreadsheet. The spreadsheet is imported into the inspection database, where the column mapping decides which CML it belongs to. Each step is a chance for 0.312 to become 0.321, or 0.372, or to land on the CML beside the one it was taken at. The corrosion rate computed from the wrong number is then trusted for years. This paper covers UT thickness data management from gauge to record: why the retyping step is the weak point, what the measurement itself must document under ASME Section V, and what the record has to carry so the reading stays the reading.

Why does retyping thickness readings corrupt corrosion rates?

Corrosion rate is a difference of two thicknesses divided by a time. Both thicknesses are small numbers with three or four significant figures, and the difference between them is often in the last digit. A transcription error of one digit is frequently larger than the real change in wall thickness between surveys.

Consider a CML that read 0.325 five years ago and 0.312 today. The true short-term rate is 2.6 mils per year. If today's reading is typed as 0.321, the rate becomes 0.8 mils per year and the remaining life triples. If it is typed as 0.302, the rate becomes 4.6 mils per year, remaining life halves, and the interval is cut in half with it. Neither error looks wrong on the screen. Both propagate into the interval, the RBI probability, and the turnaround scope.

The errors that transcription introduces are well known:

• Transposition. 0.312 becomes 0.321.
• Decimal drift. 0.312 becomes 3.12 or 0.0312, sometimes caught by validation, sometimes not.
• Unit confusion. Inches typed into a millimetre field, or a gauge set to millimetres read as inches.
• Wrong CML. The right number on the wrong row, the most damaging error because both CMLs now have a false history.
• Skipped readings. A row left blank and later filled from memory or from the neighbouring point.
• Rounding at entry. Four significant figures on the gauge, three on the sheet.

Detection is the real problem. A retyped reading looks exactly like a measured one. The error surfaces, if at all, as an implausible short-term rate at the next survey, by which time the wall has been assessed and the interval set on the false value.

The remedy is not more checking. It is removing the step. A reading that travels from the gauge to the record without a human typing it cannot be mistyped.

What does ASME Section V require of the measurement itself?

Getting the number into the record unchanged is only useful if the number was right when the gauge showed it. ASME Section V covers nondestructive examination methods, and its ultrasonic thickness measurement provisions set out what a defensible reading depends on.

• Calibration. The instrument is calibrated on reference blocks of known thickness, of the same or acoustically similar material, spanning the range of thicknesses expected. Calibration is verified at intervals during the survey and at its end, so a drift can be bounded to the readings taken between checks.
• Couplant. The couplant used is recorded, and the same couplant is used for calibration and measurement.
• Surface condition. Scale, coating, and roughness affect the reading. Coated surfaces need either coating removal or a gauge mode that ignores the coating layer, and the record should say which.
• Temperature. Sound velocity changes with temperature. Readings on hot equipment need a high-temperature transducer and a velocity correction, and the metal temperature should be recorded with the reading.
• Transducer and frequency. Recorded, because the choice affects the minimum measurable thickness and the response to back-wall corrosion.
• Personnel qualification. The technician's qualification to the employer's written practice, typically to SNT-TC-1A or an equivalent scheme, recorded against the survey.

Each of these is a field on the survey record, and several belong on every individual reading. A reading with no calibration record behind it is a number with unknown uncertainty. A reading with no temperature on a hot line is likely wrong by a known amount in an unknown direction.

The point for data management is that the measurement record and the reading are one thing. A system that stores the reading in the CML table and the calibration in a scanned PDF has separated the number from its basis.

How do modern UT gauges get data into the record?

Gauges from the major manufacturers offer three paths, and a program should know which it is using.

• Bluetooth. The gauge pairs with the tablet and sends each reading as it is taken, often with the gauge's own timestamp and settings. The inspection app receives the value and writes it to the active CML. This is the cleanest path because it is live: the technician sees the value land on the correct point while still standing at it.
• Keyboard wedge. The gauge presents itself as a keyboard and types the reading into whatever field has focus. It needs no special integration and works with almost any app, but it depends on the right field having focus at the moment the reading is sent, so the app should confirm the target CML before accepting the value.
• Data logger export. The gauge stores readings against IDs entered on the gauge, and a file is exported after the survey and imported into the system. This removes retyping but reintroduces the mapping problem: the IDs on the gauge must match the CML identities in the record exactly, and the import must map them without guessing.

Whichever path is used, the record should carry, for each reading: the value, the unit, the gauge model, the time from the gauge or the device, and the path it arrived by. A reading that arrived over Bluetooth is verifiably not retyped. A reading whose path is unknown might have been.

Device profiles matter here. A gauge integration written for one model may not handle another's framing or its units. A program that standardizes on a small set of gauges, and confirms the software has a profile for each, avoids the case where readings from one crew's instrument arrive as text in a notes field.

What must the CML carry for a reading to mean anything?

A thickness reading is a coordinate on a curve. The curve is the CML's history, and the CML's identity and design inputs are what make the curve useful. A reading stored without them is a number in a spreadsheet.

The CML should carry:

• Identity that persists. A CML ID tied to the component in the asset hierarchy, stable across surveys, contractors, and system migrations. The most common cause of a lost history is a CML renumbered between surveys.
• Position. Where on the component: course, elevation, clock position, distance from a reference weld. On a grid, the grid geometry and the reading's row and column.
• Design inputs. Nominal thickness, minimum required thickness and the method that produced it, design pressure, material, outside diameter and schedule, allowable stress, joint efficiency, corrosion allowance. These turn a reading into a remaining life.
• The reading history. Every reading with its date, technician, gauge, and conditions. Not the latest only.
• Derived rates. Short-term, long-term, and maximum corrosion rates, all visible. An implausible short-term rate is the first sign of a bad reading; hiding it behind a single governing rate hides the signal.
• Status. Active, retired, replaced, with the reason and date for any change.

Grid and scan data deserve a specific rule. A C-scan or a manual grid over a thinned area produces dozens or hundreds of readings. Typing the minimum into a single field discards the rest. The full grid is what a fitness-for-service assessment will need later, and it cannot be recovered without another entry. Store every point with its grid position.

How should thickness data be validated at entry?

Direct gauge input removes transcription error. It does not remove a mis-coupled probe, a reading through a lamination, a reading taken at the wrong point, or a gauge left in millimetres. Validation at entry catches what the gauge cannot.

• Range against nominal. A reading above nominal thickness plus a tolerance, or below the retirement thickness, prompts a confirmation before it is accepted.
• Rate plausibility. The implied short-term rate compared with the long-term rate and with the rates at neighbouring CMLs. A CML that suddenly corrodes ten times faster than its neighbours is either a real finding or a bad reading, and either way it deserves a second measurement now, while the technician is still there.
• Unit lock. The CML's unit is fixed and the gauge unit is checked against it.
• Repeat reading. For readings that fail plausibility, a second reading at the same point, recorded as a second reading rather than overwriting the first.
• Target confirmation. Before a reading is accepted, the app shows the CML it will be written to, so a keyboard wedge cannot type into the wrong row.

Validation should be visible in the record. A reading that was flagged and confirmed carries the flag and the confirmation, so the next inspector knows it was questioned and why it was kept.

What should the audit trail show for a thickness reading?

A reading is a fact about a wall on a date. It should never be silently changed. If a reading is corrected, the correction is a new event with the old value, the new value, the actor, the time, and the reason, and the history shows both.

The audit trail for thickness data should carry, for every reading: who recorded it and how it arrived (gauge over Bluetooth, keyboard wedge, import, manual), when it was recorded and when it synced, any validation flags and their resolution, and every subsequent view, edit, or export. For imports, it should carry the source file and the mapping that was applied, so a mapping error can be traced to every CML it touched.

Tamper evidence matters here more than anywhere else in the record, because thickness data is what an incident inquiry reconstructs. An audit trail in which each event is hashed together with the previous one lets an auditor verify that no reading was quietly edited or removed. It also settles disputes between contractors and owners about what was measured when.

How intelliSPEC handles it

intelliNDT and intelliINTEGRITY are built so the number the gauge showed is the number in the record.

• Gauge to record without retyping. Seven device profiles over Bluetooth or keyboard wedge, including the Evident 38DL Plus, the DeFelsko PosiTector UTG, and the Elcometer MTG for ultrasonic thickness, along with the DeFelsko PosiTector 6000 and Elcometer 224T for coating and profile measurement and the TSL 1153 RFID reader for tag identification. Readings enter the active CML directly.
• CMLs that carry the basis. Each CML point carries 64 fields, including nominal thickness, tMin and the method that set it, design pressure, material, outside diameter and schedule, allowable stress, joint efficiency, corrosion allowance, grid geometry, the latest thickness, and three corrosion rates. The CML lives on a component node in the asset graph, so its identity persists across surveys.
• Three rates, all visible. The corrosion summary computes short-term, long-term, and maximum rates and takes the maximum as governing; remaining life is (t minus tMin) divided by the governing rate, and the API 510 and API 570 interval is capped at half the remaining life. tMin routes by equipment type to ASME B31.3, ASME VIII Division 1, or API 653 and returns the code and formula with the number.
• Imports with the mapping recorded. Excel workbook import with AI-assisted column mapping and CSV import bring legacy survey data in, and the import is an audited event carrying actor, tenant, and request id.
• A hash-chained audit trail. Every create, read, update, delete, import, and export carries the actor, tenant, request id, IP, and user agent with a SHA-256 hash chain per tenant that a verify routine can recompute. Audit events are never auto-deleted.
• Offline at the CML. The full record, including prior readings and drawings with automatic CML point detection, is on the device with no time limit on the local store, and offline data is encrypted at rest.

For definitions of CML, tMin, and the corrosion rates, see the glossary. For how thickness programs fit a refinery or power plant, see oil and gas and power.