Exchanger bundle inspection inside the window

On most process units, the exchangers set the pace of the turnaround. There are more of them than there are columns, they share a small number of cleaning pads, test rings, and tube inspection crews, and any one of them can move from "clean, test, return" to "retube" in a single hydrotest. The unit cannot start up with a bundle on the pad. This paper covers what API 510 asks of an exchanger, how to choose a tube inspection method, and what the bundle record must carry so the plug-or-retube call is made on evidence and the next event starts from a real history.

Why exchanger bundles decide the critical path

A column has one internal inspection, one set of trays, and one manway sequence. A unit may have forty exchangers, each with a pull, a hydroblast, a visual, a tube inspection, a test, and a return, and all forty compete for the same three or four shared resources. When a bundle fails its test after cleaning, the recovery options (plug, retube, or install a spare) each have a different lead time, and the decision has to be made by someone who can see the tube results and the plugging history.

The failure mode is not usually a lack of effort. It is a sequence that puts low-criticality bundles on the pad first, a tube inspection crew waiting for cleaning that is behind, and a plug-or-retube decision that waits for a report to be typed. Each of those is a record and scheduling problem before it is a metallurgical one. The TEMA designation (front head, shell type, rear head, for example AES or BEU) tells the planner what the pull involves: a fixed-tubesheet bundle does not come out at all, a floating-head bundle needs the floating head cover and backing device removed first, and a U-tube bundle limits which methods can inspect the bend.

What does API 510 require for a shell-and-tube exchanger?

API 510 treats the exchanger as a pressure vessel with two pressure boundaries, shell side and channel side, each with its own design pressure, thickness monitoring, and corrosion rate. The inspection program has to cover:

• Thickness monitoring at CMLs on the shell, channel, heads, and nozzles, with corrosion rates and remaining life computed for each boundary.
• Internal inspection of the shell and channel when the bundle is out, covering the shell interior, pass partition plates, gasket surfaces, nozzle bores, and any lining or cladding.
• The bundle as part of the internal condition: tubesheet faces and ligaments, tube ends and rolls or welds, baffles, tie rods and spacers, impingement plates, and the tubes themselves.
• Interval setting on the same basis as any vessel: the internal or on-stream interval is limited to half the remaining life or ten years, whichever is less, with RBI under API 580 and 581 permitted to adjust the interval where the assessment supports it.
• Pressure testing where the repair or the inspection plan calls for it, and documentation of any repair or alteration under the applicable code and jurisdiction.

The tube inspection itself is not prescribed method by method in API 510. It is the owner's inspection plan, informed by the damage mechanisms expected for the service, that decides how many tubes and by which technique. That is why the damage mechanism record for the exchanger matters more than a generic percentage.

Which tube inspection method fits which bundle?

There is no single tube method. The choice follows tube material, the mechanism you are looking for, and how clean the tubes can be made:

• Eddy current testing (ECT). Fast, good at detecting and sizing pitting, wall loss, and cracking in nonferrous tubes (admiralty brass, copper-nickel, austenitic stainless, titanium). Ineffective on ferromagnetic tubes without a saturation probe.
• Remote field testing (RFT). For ferromagnetic carbon and low-alloy steel tubes. Detects general and localized wall loss, less sensitive to small pits than ECT is on nonferrous.
• Magnetic flux leakage (MFL). Also for ferromagnetic tubes, quick screening for wall loss, weaker on gradual thinning.
• IRIS (internal rotary inspection system). Ultrasonic, gives a true wall thickness reading around and along the tube on any material. Slower and requires the tube to be clean to bare metal, which decides how much hydroblasting time it costs.
• Near-field or partial saturation techniques for finned or duplex tubes where the standard methods struggle.

A common plan is a screening method on a high percentage of tubes (ECT or RFT) with IRIS on a sample of the worst indications and at known problem rows, such as the inlet rows under the impingement plate or the tubes at the baffle windows where vibration damage concentrates. The plan should name the method, the tube count, the sampling logic, and the acceptance threshold before the bundle is pulled. The method choice and the readings themselves live in intelliNDT on the same asset record the visual inspection writes to.

What should the bundle record carry?

A bundle record that says "inspected 20 percent, 6 tubes plugged, returned to service" is a summary of an inspection, not an inspection record. The record that supports the next decision carries:

• The tube map. Every tube position keyed to the tubesheet layout, with the method used, the result, the maximum wall loss percentage or thickness, and the axial location of the indication (inlet end, baffle, U-bend, outlet end).
• Plugged tubes. Which tubes, when, why, and the running count against the plugging allowance set at design.
• Tubesheet and joint condition. Ligament cracking, erosion at the face, tube-to-tubesheet joint type and condition.
• Internals. Baffles (wear at the tube holes, corrosion), tie rods and spacers, impingement plate integrity, sealing strips, pass partition plates.
• Shell and channel. Thickness readings at the CMLs, gasket surface condition, nozzle bores, lining.
• Cleaning and test. Cleaning method and result, test-ring hydrotest pressure and result, leaking tubes identified.
• Disposition. Plug, retube, return as is, or replace, with the approving engineer and the reasoning.
• Damage mechanism. Under-deposit corrosion on cooling water, inlet erosion, vibration fatigue at baffles, chloride cracking of stainless, dealloying of brass, or whatever the evidence supports.

The tube map is the field that pays off across events. Wall loss that has migrated from the inlet rows into the bundle interior over three turnarounds is a trend that no single event's summary reveals.

How do you plan bundle work so it fits the window?

The pull is rarely the constraint. The cleaning pad, the test ring, and the tube inspection crew are, and they are shared. A planning sequence that holds up:

1. Rank bundles by consequence and by likelihood of a retube, using the tube map history and the damage mechanism record. The bundles most likely to need a spare or a retube go to the pad first, so a long-lead decision is made with the most window left.
2. Set the tube inspection plan per bundle before the event: method, tube count, sampling logic, acceptance criteria, and the person who will make the disposition.
3. Pre-stage spares and plugs for the high-likelihood bundles, and confirm the retube lead time with the shop before the unit comes down.
4. Track the bundle through its steps as an activity with daily progress: pulled, cleaned, inspected, tested, dispositioned, returned. A bundle sitting between steps is the most common silent delay.
5. Record the delay event when a bundle misses its slot on the pad or the ring, with the cause, so the next event's sequencing is based on where the queue actually formed.

The visual and internal inspection findings belong in intelliINSPECT, where they are on the asset and not in a contractor's report folder.

Plug or retube: how do you make the call with the record?

The plugging allowance is set at design as a fraction of the tube count and reflects the thermal and hydraulic margin. Approaching it is one trigger for retubing, but it is not the only one. The decision is better framed as three questions the record should be able to answer:

• Is the loss localized or general? Six tubes plugged for inlet erosion under a failed impingement plate is a localized problem with a localized fix. Six tubes plugged for general thinning across the bundle means more will fail before the next event, and the count will not stay at six.
• Will the mechanism continue? Vibration damage at a baffle will keep going until the baffle spacing or the flow changes. Plugging buys one run, not a solution.
• What did the trend do since the last event? If the maximum wall loss moved from 30 to 55 percent in one run, the corrosion rate is telling you what the next run will look like.

The answers come from the tube map history and the damage mechanism record, not from this event's plug count alone. Where the disposition changes the exchanger's design basis, or defers a known defect, it should go through the integrity engineer and be recorded as such, with the remaining life and next interval recomputed in intelliINTEGRITY.

How intelliSPEC handles it

The exchanger is an asset node in the graph with the bundle and its components under it, so the tube map, the shell CMLs, the visual findings, and the NDT readings all write to one record. CML points carry the fields the remaining-life calculation needs (nominal thickness, tMin and its method, design pressure, material, allowable stress, latest thickness, and three corrosion rates), and the corrosion summary computes short-term, long-term, and maximum rates with remaining life as (t minus tmin) divided by the governing rate. The API 510 interval calculation caps the next internal at half the remaining life.

Instrument readings from UT gauges enter the record over Bluetooth or keyboard wedge without retyping, and photos bind to the component at capture. Damage mechanisms are first-class records, so "under-deposit corrosion, cooling water side" is a field the API 580/581 RBI scoring reads, not a phrase in a paragraph. intelliPROJECTS tracks the bundle as an activity with daily progress and records delay events against it, and the plug-or-retube disposition is a recorded workflow transition with the approving engineer's identity in the hash-chained audit trail. A completed record pins the form configuration it was executed with, so the tube map from this event reads the same way at the next one.