Where should the inspection team look first?
Corrosion under insulation hides. Stripping and screening every insulated circuit is not an option, so someone has to decide the order. This workbench takes the circuit list, its operating context and the inspection history a plant already holds, and turns them into a ranked, explainable look-here-first list.
It is a decision-support layer, not a replacement for the inspector's judgment or for API 583 practice. Every score shows what drove it, every weight can be changed, and the backtest below shows how the order would have performed against known findings.
Circuit attributes you already track: material, operating temperature, insulation and jacketing, coating, environment, damage points, dates, prior findings.
A ranked list with priority tiers and the three drivers behind each score, plus a first-look suggestion per circuit.
Hide the findings from one historical campaign, rank the circuits blind, and count how many real findings sit at the top of the list.
Ranked listForty-eight insulated circuits on one sample unit, ordered by CUI likelihood.
Move a weight and the order, the tiers, the backtest and the plan all update together. Click a row for the full record.
How the score is built
Each circuit gets a 0 to 1 reading on eight factors drawn from API RP 583 and the CUI damage-factor logic in API 581. The weights start from published guidance. In a pilot they are recalibrated from your own findings.
Temperature bands used
Carbon steel: susceptible from about −12 °C to 175 °C, worst between roughly 50 °C and 120 °C. Austenitic stainless: chloride stress-corrosion cracking risk from about 50 °C to 175 °C. Cyclic or intermittent service raises the reading. Ranges follow API RP 583 guidance and are adjustable per site.
| # | Circuit | Score | Tier | What drives it | Last insp. |
|---|
BacktestRank blind, then check the order against what the last campaign actually found.
This sample unit has … circuits with CUI in its history. The curve shows how quickly each ordering reaches them.
Share of known CUI findings reached, by share of circuits inspected
Hover for values. A curve that climbs early means the crew meets the real problems first. Ordering by time since last inspection is the common default when no risk model is in place.
What the inspection team getsA first-look plan for the top of the list, in language a crew can act on.
The workbench decides the order. Your inspection procedures decide the method. The suggestions here are illustrative.
| # | Circuit | Why it is here | Suggested first look |
|---|
See the unit in 3DEvery reading point on the rack, and a probe that works the list in order.
A clay model of the same 48 circuits. Click any pipe for its record, run the sweep to watch the findings appear in rank order, pull the insulation layers apart, and print the spec sheet.
Open the reading simulationOne anonymized export, or one page on how you plan today.
Either is enough to pressure-test the approach. No meetings are required. Questions go by email to a named technical contact and are answered in writing within a day.
A historical, anonymized circuit export
One unit or one campaign, 150 circuits or more, with the findings included. We hide the findings, rank blind, and send back a written scorecard like the panel above, using your numbers instead of sample ones.
How prioritization happens today
A short description of the current planning cycle (RBI software, spreadsheets, engineer judgment) and a sample planning list. We map where a ranked, explainable list would slot in and what it would change.
| Column | Example | Needed |
|---|---|---|
| circuit_id | P-310-014, or a hash you generate | Required |
| material | CS, SS, low alloy | Required |
| op_temp_c | 88, or a band such as 50–120 | Required |
| service | Sour water | Required |
| cyclic_service | Y / N | Helpful |
| insulation_type | Calcium silicate | Required |
| jacket_type | Aluminum | Helpful |
| jacket_condition | Good / Fair / Poor at the last walkdown | Required |
| coating | None / Organic / TSA, with year if known | Required |
| insulation_year | 2004 | Helpful |
| last_inspection_date | 2019-06 | Required |
| environment_tags | cooling-tower drift; wash-down zone | Helpful |
| damage_points | 3 supports; 1 dead leg; 2 penetrations | Helpful |
| prior_cui | Y / N | Helpful |
| criticality | 1 to 4, or your RBI consequence class | Helpful |
| finding | CUI found Y / N, max wall loss % | Required for the backtest |
circuit_id,material,op_temp_c,service,cyclic_service,insulation_type,jacket_type,jacket_condition,coating,insulation_year,last_inspection_date,environment_tags,damage_points,prior_cui,criticality,finding
Not needed: client or site names, P&IDs or drawings, coordinates, cost data, or any personal information. Circuit IDs can be hashed on your side before export.
Four steps, by email, in about three weeks.
Nothing here needs a meeting. If the blind ranking does not beat the order you would have used anyway, the report says so and we stop there.
- STEP 1You · about an hourSign the NDA, send the export
Or the one-page workflow description. Name one technical contact for written questions.
- STEP 2Clear Anvil · one weekRank blind, then unblind
Findings are held back, circuits are ranked from attributes and history, then the order is checked against what was actually found.
- STEP 3Clear Anvil · a few daysWritten scorecard
Capture at the top of the list versus your current order, the drivers that mattered at your site, and where the model was wrong.
- STEP 4Both · your callSmall paid pilot, or stop
If it earns its place: one live planning cycle on one unit, with your planners in the loop. Not a large software commitment.
Send one export.
Get a written scorecard back.
Reply to the thread or write to the address below. No meetings needed unless you want one.
jamil@clearanvil.com