Inspection, remaining-life estimates and uncertainty

An inspection method answers a bounded question

Visual examination can reveal accessible surface features under appropriate lighting, viewing conditions and operator capability. It cannot exclude every tight crack or buried defect. Penetrant methods depend on discontinuities being open to the surface and on compatible, sufficiently clean surfaces. An indication is evidence to interpret, not automatically a rejectable defect.

Penetrant inspection makes a surface opening visible through liquid entry and retention. After the excess at the surface is removed, retained liquid can emerge to form an enlarged visible or fluorescent indication, often assisted by a developer. A blocked opening defeats entry; a fully buried flaw has no path to admit the liquid. A visible indication is not a map of the complete buried shape or depth. These principles explain the surface-opening and cleanliness requirements.

Pulse-echo ultrasound asks another question. A probe sends sound into a suitable material and receives reflections from interfaces or discontinuities. Echo timing can locate a reflector along a known sound path; echo strength and absence also depend on coupling, orientation, scattering and geometry. A flat reflector that directs sound away from the receiver can escape the chosen view. A thickness reading is therefore not automatically a qualified search for every crack.

Other methods, including ultrasound or radiography, can address different questions, but geometry, material, flaw orientation and access affect capability. A successful check has a stated area, method, sensitivity, conditions and acceptance basis. “No reportable indication found in the inspected region” is a more defensible claim than “there are no flaws anywhere”.

Turn a return time into a bounded distance

For a straight normal sound path with known speed v and corrected round-trip time Δt, distance to the reflector is d=vΔt/2. The pulse travels down and back, hence the factor two. Speed has units of distance/time; multiplying by time gives total travelled distance. Instrument and coupling delays must have been accounted for before this simple equation applies.

To interpret d as thickness h, establish that the echo comes from the back face, not an internal reflector or another path. The figure contrasts two possible echo paths. Its simple planes and supplied sound speed isolate travel-time reasoning; the drawing is not an instrument setting or a complete flaw-evaluation method.

Build a transparent thickness trend

Consider an original paper model of uniform loss. A location was listed at 6.4 mm two years ago and 5.8 mm now. The average apparent loss rate is (6.4 − 5.8)/2 = 0.30 mm/year. If that rate remained constant, thickness would follow h(t) = 5.8 − 0.30t, with t in years after now.

The nominal crossing of a chosen 4.0 mm threshold is six years away. The figure adds an assumed current range 5.6–6.0 mm and possible future rates 0.20–0.50 mm/year. These are sensitivity bounds chosen for teaching, not a statistical confidence band or measured corrosion-rate distribution. Their earliest crossing is much sooner.

Keep uncertainties connected to action

The apparent historical rate includes measurement differences and any change in location or method. Two readings cannot establish constant future loss. Local pits, cracks, new chemistry, a coating change or cyclic loading may invalidate a uniform-loss model. The threshold itself belongs to a specific design assessment, not a universal safe thickness.

A remaining-life assessment combines relevant inspection evidence, loads, degradation mechanisms, model limits and consequence of failure. It may call for a different inspection or repair rather than more arithmetic. The simple crossing times here demonstrate sensitivity only; they neither authorise continued use nor set an inspection interval for equipment.

Link the information to its model before extrapolating

With nominal h0 = 5.8 mm, threshold 4.0 mm and rate 0.30 mm/year, crossing time is (5.8 − 4.0)/0.30 = 6 years. Calling this “six safe years” would add a safety conclusion the model does not establish.

Use the lower assumed current thickness and faster assumed rate: (5.6 − 4.0)/0.50 = 3.2 years. The upper-thickness, slower-rate combination gives (6.0 − 4.0)/0.20 = 10 years. The 3.2–10 year span is a set of scenario results, not a promised lifetime interval.

Before using any trend, reconcile measurement locations, methods and uncertainty. Then ask whether uniform loss remains an appropriate mechanism and what evidence the actual service decision requires.

In a separate invented parallel-sided coupon, supply sound speed v=6.0 mm/µs and a corrected back-face return time 2.0 µs. The distance is 6.0×2.0/2=6.0 mm. Reporting 12.0 mm would forget the return leg. A 1.0 µs echo instead corresponds to 3.0 mm along this path; without identifying its reflector it does not establish a 3.0 mm wall.

Now suppose a smooth accessible face has no visual indication. The concern is either a tight surface-opening discontinuity or a fully buried planar one. A compatible penetrant examination can address the first question through liquid entry; it cannot address the second through a solid covering layer. A suitable qualified ultrasonic arrangement may address the buried question, but direction and coverage matter. Repeated negative readings from the same unsuitable path do not resolve it.

For a changed uniform-loss exercise, take matched-location readings 6.2 mm three years ago and 5.6 mm now: apparent rate(6.2−5.6)/3=0.20 mm/year. With a separately supplied 4.4 mm model threshold, nominal crossing is(5.6−4.4)/0.20=6 years. If the earlier reading belongs to another spot, even that local historical slope is unsupported. A spatial difference must not be relabelled a time trend.

Try the reasoning

1. With supplied speed 5.0 mm/µs, a corrected return time 2.4 µs is converted to 12 mm thickness. Detect the error and state what else is needed.

   Show the answer and reasoning

   12 mm is the out-and-back path length. Distance is 5.0×2.4/2=6.0 mm. It represents thickness only if the straight-path echo is identified as the back face and the speed/delay assumptions are appropriate.

2. A penetrant result has no indication on an apparently clean face. Can it exclude a fully buried flaw? Explain using information formation.

   Show the answer and reasoning

   No. Liquid must enter an opening and later form an indication. A buried flaw lacks a surface entry path. Even surface flaws can be missed if an opening is blocked or conditions are unsuitable; the negative result remains bounded.

3. A planar reflector gives no echo to the original probe direction. Why can changing the direction add information rather than merely repeat the test?

   Show the answer and reasoning

   Reflection depends on orientation. Another suitable path can return energy that the first direction sends elsewhere. Coverage, calibration and qualified interpretation still matter; echo absence from one view does not uniquely imply absence of a flaw.

4. Use current thickness 5.6 mm, threshold 4.4 mm and possible future rates 0.15–0.30 mm/year. Calculate crossing scenarios and assess a 95% confidence claim.

   Show the answer and reasoning

   The faster-rate crossing is 1.2/0.30=4 years; the slower-rate crossing is 1.2/0.15=8 years. These chosen rate scenarios supply no probability model or justified 95% coverage. They are conditional crossings, not safe-use or inspection intervals.

5. A reading at spot A is 6.4 mm in year 0; a reading at spot B is 5.8 mm in year 2. A report uses them as local loss of 0.30 mm/year. What is missing?

   Show the answer and reasoning

   The report has not separated initial spatial variation from temporal change. It needs matched locations and comparable measurement evidence, or a justified spatial model. A true time trend at A or B cannot be obtained solely by subtracting these two different spots.

What to carry forward

Inspection narrows uncertainty within its capability. A trend crossing is a conditional model output, and service decisions need more than that output.