DFT on a Blasted Tank: Spot Averages, a High Reading, and a Profile Bias
A 20 ft by 20 ft area of a steel tank exterior carries an epoxy with a specified DFT of 6.0–10.0 mils. The inspector uses a Type 2 electronic gauge, checked each morning on certified shims on a smooth steel plate, where it reads within 0.1 mil. The steel was blasted to about 2.5 mil profile. Spot readings (three gauge readings per spot): Spot 1: 4.9/5.6/6.2. Spot 2: 7.1/7.6/7.4. Spot 3: 10.8/11.4/10.9. Spot 4: 8.0/8.4/8.1. Spot 5: 6.6/7.0/6.9. The contractor's QC reports an area average of 8.4 mils and asks the inspector to countersign. The inspector notes that the gauge has never been adjusted for blast profile, and that an uncoated blast panel of the same profile averages 0.6 mil.
1. Apply the acceptance rules and identify which spots fail and why.
Read the worked answer
Spot averages: 1 = 5.57, 2 = 7.37, 3 = 11.03, 4 = 8.17, 5 = 6.83 mil.
- Spot 1 fails: average 5.57 is below the 6.0 mil minimum. The single low reading of 4.9 is above the 80% floor of 4.8, so the secondary rule is not what fails it; the spot average is.
- Spot 2, 4 and 5 pass: all averages sit inside 6.0–10.0, and no single reading is below the 80% floor or above the maximum.
- Spot 3 fails: average 11.03 exceeds the 10.0 mil maximum, and every individual reading is high, so this is not a single outlier.
- Area average of the spot values is about 7.79 mil, inside the range, but an in-range average does not rescue failing spots. The area is not acceptable as it stands.
- The contractor's 8.4 mil figure does not match any computation from these readings, so the source data and the reading locations need to be reconciled before any countersignature. Ask which gauge, which spots and which time.
2. How does the 0.6 mil profile artifact change the numbers, and how defensible is the correction?
Read the worked answer
On blasted steel the gauge sits on profile peaks, so it reads coating plus trapped profile and is biased high, roughly by the peak height. Subtracting 0.6 mil from each spot average gives: Spot 1 = 4.97, Spot 2 = 6.77, Spot 3 = 10.43, Spot 4 = 7.57, Spot 5 = 6.23. The conclusion does not improve: Spot 1 now fails by a wider margin and Spot 3 still exceeds the 10.0 mil maximum, while Spots 2, 4 and 5 remain compliant. How defensible is the correction?
- Subtracting a reference panel average is a recognised approximation only if the panel has the same profile, curvature, orientation and substrate, and if it was measured with the same gauge and probe.
- On a curved shell, the reference panel must match curvature; a flat panel will not.
- Better evidence, if available: profile measured on the actual surface before coating, a gauge with documented profile adjustment, or destructive DFT by Tooke gauge at a representative location, with repair of the test site. Document the method chosen and the numbers both raw and adjusted, so a reviewer can follow the logic.
3. What do you tell the contractor, and what goes into the report?
Read the worked answer
Do not countersign the 8.4 mil average. Issue a nonconformance and state the disposition per spot.
- Low spot: if the area is still within the recoat window and the surface is properly prepared, additional coats can bring the DFT up; if the window has passed, abrade per the data sheet first. Re-measure afterwards with the same method.
- High spot: exceeding a specified maximum is a nonconformance. Whether it is a defect is an engineering question, for example flexibility, solvent entrapment on a high-build system, or stress in the film; refer it to the owner for a decision or concession rather than quietly accepting or rejecting it.
- Report contents: gauge make, serial and calibration and shim checks; the profile adjustment basis and the reference panel data; the time, location and value of every spot; computed spot averages; the raw and adjusted values; photographs; the NCR reference; and the name of the person notified. The tested judgement is that numerical compliance has three layers: individual readings, spot averages and area average, and that instrument bias must be dealt with before any of them are interpreted.