Applying Weld Acceptance Criteria as an Inspector

A new inspector knows what the flaws look like. A senior inspector knows which rule applies to this weld on this job, what to do with a reading that sits right on the limit, and which decisions are not theirs to make. Acceptance criteria are the written limits that decide whether a weld passes. They always come from the governing code or the contract documents, never from memory and never from the last job you worked on.

The limits in the examples here are made up for practice, on purpose. Do not use them on real work. Read the actual table in the edition your job names, and keep in mind that reading about inspection does not qualify you to accept or reject welds. On code work, that belongs to a qualified inspector.

What You Will Be Able to Do

• Trace a weld to the code, edition and loading that govern it.
• Apply zero-tolerance, size, length and frequency limits correctly.
• Handle a measurement that falls right at a limit.
• Time an inspection around delayed cracking.
• Separate the inspector's decision from the engineer's.
• Decide what to do when the normal method does not fit.

Find the Rule Before You Look at the Weld

Start at the top of the chain and work down. Each level can make the one below it stricter.

Different kinds of work use different documents. AWS D1.1 covers structural steel, AWS D1.5 covers highway bridges, ASME B31.3 covers process piping, and API 1104 covers pipelines. Each has its own limits, written for its own risks. The same undercut can pass under one and fail under another, so a number you remember from one code proves nothing under a different one.

Some tables also change between editions. Clause and table numbers move as well, so cite the edition you actually used.

How Limits Are Written

Criteria come in a few types, and each one has its own trap.

Read the words around the number. "In any 12 inches" means any 12 inches you can find, not the first 12, and not the 12 that happen to look best. "Exceeding" and "at least" tell you whether a reading right at the number passes.

Worked Example: The Sliding Window

Here are made-up project limits for practice only:

• Count every visible pore 1/32 inch or larger.
• The sum of the pore diameters must not exceed 1/4 inch in any 1 inch of weld.
• The sum must not exceed 1/2 inch in any 12 inches of weld.

A 30 inch fillet shows these pores. Each position is measured from the start of the weld.

The 1 inch check. The tightest cluster is at 3 to 3.75 inches. That adds up to 2 + 3 + 2 = 7/32 inch, which is under 8/32, or 1/4 inch. This passes.

The 12 inch check, starting at the beginning. From 0 to 12 inches, the sum is 2 + 3 + 2 + 4 + 2 = 13/32 inch. That is under 16/32, or 1/2 inch. It seems to pass.

The 12 inch check, sliding the window. Move the window to run from 3 to 15 inches. Now it holds all seven pores: 2 + 3 + 2 + 4 + 2 + 3 + 1 = 17/32 inch. That is over 16/32, so the weld is rejected.

Anyone who only checked the first 12 inches would have passed that weld. Start a window at each flaw, and check where the flaws bunch together. Then write down the window you used.

Borderline Readings

Most welds are clearly good or clearly bad. Your skill shows on the ones that are not.

1. Clean it. A reading through slag or spatter means nothing.
2. Find the worst point. Measure there, not at an average spot.
3. Use a gauge that can read the limit. A gauge marked in 1/32 inch steps cannot prove a value at 0.031 inch. Use a finer tool, such as a dial depth gauge, or take a replica of the surface.
4. Repeat it. Take the reading more than once, and have another qualified person check it if you can.
5. Do not round in the weld's favor. A reading of 0.033 inch is over a limit of 0.031 inch.
6. Record the actual number. Write down what you measured, not "OK."

If you cannot prove a weld meets the limit, it has not met the limit. Send it for repair, or take it to the person with authority to decide. Keep your measuring tools in calibration, because your reading is only as good as the tool. That also means checking that a gauge still reads true before a critical call.

Timing the Inspection

Hydrogen cracking can show up hours or days after welding. A weld inspected too soon can pass and then crack later, so some rules build in a wait.

• AWS D1.1 lets visual inspection begin once the weld has cooled to ambient temperature.
• For certain high-strength steels, such as ASTM A514, A517 and A709 Grade HPS 100W, D1.1 bases acceptance on an inspection made at least 48 hours after the weld was finished.
• Other standards set their own delays. TWI, a welding research institute, reports waits from about 16 to 48 hours in various codes and specifications, some based on the steel, the thickness or the heat input.

Build that wait into the job schedule. An early inspection on one of those steels does not count, however good the weld looks. Also plan for what comes after you: inspect welds before they are painted or covered, and catch root work before the next layer buries it.

Who Decides What

Your authority is real, and it has limits. Respect both.

An inspector compares the work to the rules. An engineer may know a weld with a small flaw is still safe for its load, and may accept it. That is an engineering decision, and it has to be documented. If you hear "the engineer said it's fine," ask for it in writing before you change anything on your report.

On many structural jobs, two kinds of inspection happen side by side. The contractor's people do quality control, and an inspector working for the owner verifies the work. The governing documents spell out who does what. Know which role you are in, and keep your independence. Your job is not to design the repair, run the crew or help the schedule.

When the Normal Method Does Not Fit

Advanced work is often about the case the table does not cover cleanly.

The rule here is simple. If you cannot inspect a weld as the procedure requires, do not quietly change the method. Report the limit and get a decision.

Checking Other People's Work

Senior inspectors review other inspectors and welders as well as welds.

• Sample their measurements. Re-measure a few welds that were already accepted. If the readings differ, find out why before you trust the rest.
• Audit the preconditions. Confirm that the WPS matches the joint, that the welder's qualification covers it and has not lapsed, and that the preheat was really reached.
• Look for trends. A rise in rejects from one welder, one shift or one station usually has a cause. It may be training, a machine or a batch of filler.
• Review reports as well. A report with no measurements, no edition or no weld ID cannot be relied on.
• Expand when required. If partial inspection finds flaws, the specification may require more testing. Follow it, even when the schedule is tight.

Treat all of this as checking the system, not hunting for someone to blame. People stop raising problems if every finding becomes a fight.

Common Mistakes

• Applying the last job's limits. A different code or edition can change the number.
• Checking only the first window. The worst stretch of weld may not start at zero.
• Rounding toward a pass. A reading over the limit is over the limit.
• Using a gauge too coarse for the limit. You cannot prove what your tool cannot read.
• Inspecting too early. Delayed cracks can appear after you sign.
• Accepting a verbal "use as is." Get the engineer's decision in writing.

Check Your Understanding

A foreman says the job uses the same undercut limit as the last project, which was under a different code. What do you do?

Check the contract documents for the code and edition, then read the limit from that table. A limit from another code does not apply here.

A new inspector checks every weld by counting flaws in the first foot only, then writes accepted. Why is that habit a problem?

A limit worded for any stretch of weld applies wherever the flaws bunch together. The worst stretch may start anywhere, so begin a window at each flaw and record the one you used.

You and a second inspector land on opposite sides of a limit at the same spot. What should happen next?

Clean the spot, confirm that both tools read true, and measure the worst point again. If you still cannot prove the weld meets the limit, treat it as not meeting it and take it to the person with authority to decide.

An engineer tells you on the phone that a rejected weld can stay. Do you change your report?

No. Keep the finding as it is, and get the engineer's decision in writing, so that it can be attached to the record.

Key Takeaways

• Criteria come from the governing code, edition and contract, not from memory.
• Know the limit type, and slide length windows along the whole weld.
• Measure the worst point with a tool fine enough for the limit, and do not round toward a pass.
• If you cannot prove a weld meets the limit, it has not met it.
• Delayed cracking can require a wait, and an early inspection may not count.
• The inspector applies the rules. Only the engineer or owner can accept a deviation, in writing.