What Does "Good" Actually Look Like in Inspection?

A friend of mine, a true gentleman and an experienced API inspector, posed a simple but powerful question on LinkedIn last week:

"What does good actually look like in inspection?"

The person who asked it was Paul Evey, and his post can be found here. It gets directly to the heart of why mechanical integrity (MI) continues to fail across refineries and process facilities.

As an engineer, my inspection friends will tell you that engineers have a tendency to over-analyze. In this case, I would argue that over-analysis is exactly what is required. Because "good inspection" is not intuitive, and it is certainly not defined by tradition, habit, or audit checklists.

What a Good Inspection is Not

A good inspection is not defined by:

• How many CMLs were taken
• How many scans were performed
• How thick the inspection report is
• How many boxes were checked

Those are measures of activity, not assurance.

A good inspection is defined by one thing:

Did we meaningfully reduce uncertainty about the damage mechanism?

If the answer is no, the inspection - no matter how expensive or comprehensive it looked - did not accomplish its purpose.

What a Good Inspection Must Answer

A good inspection program must be able to answer these questions:

At a minimum, the inspection report must include:

• A clear description of how much of the asset was examined
• Proper documentation of the inspection technique used, allowing it to be repeated
• A description of damage type, location, and dimensions
• An estimated or calculated damage rate

Without this information, the report may look professional, but it cannot be used to quantify inspection effectiveness.

What We Often Get Instead

In many facilities, inspection programs deliver:

• 100% scan coverage with a tool that cannot see the damage
• Spot UT where corrosion is highly localized
• Visual inspection where damage is subsurface

That produces confidence, not knowledge of the actual condition. Confidence, when it is unearned, can be deadly.

What "Good" Would Actually Look Like

A real inspection program would work like this:

Step 1 - Identify Damage Mechanism Morphology

Not just "corrosion," but:

• Pitting vs. general thinning
• Internal vs. external
• Under-deposit vs. flow-accelerated
• Wet H2S vs. amine vs. naphthenic vs. MIC

Each of these produces very different flaw shapes and very different inspection challenges.

Step 2 - Let Damage Morphology Drive NDE Physics

Inspection methods must be chosen based on what they can actually detect, not what is convenient or cheap:

• Shear-wave vs. straight-beam UT
• Guided-wave vs. conventional UT
• Digital radiography vs. film
• PEC vs. MFL vs. IRIS vs. ACFM

Physics matters. If the tool cannot "see" the flaw morphology, coverage is meaningless.

Step 3 - Prove Damage Mechanism Detectability

We must be able to answer:

• What size and type of indication is dangerous?
• What size and type can this method detect?
• What probability of detection has been demonstrated?

In reality, we often do not know these numbers and rely on engineering judgment. If that judgment is provided by inexperienced personnel, we are not estimating we are guessing.

Coverage is Not the Same as Detection

"Good" inspection is about reducing uncertainty, not about achieving coverage metrics. The biggest lie in modern inspection is:

"We had 100% coverage."

Coverage of what? With what sensitivity? For what damage mechanism? You can have 100% coverage and a 0% probability of detecting the flaw that will cause a loss of containment.

That is how HF leaks, exchanger ruptures, and refinery fires keep happening - even though "everything was inspected."

A Conversation the Industry Needs to Have

Paul Evey asked an important question. The industry needs to answer it honestly. If we truly want to prevent loss-of-containment events, we must stop mistaking inspection activity for inspection effectiveness and start demanding evidence that our programs actually reduce uncertainty about failure.

I invite the industry to have this conversation. What are your thoughts? Comment below, join the conversation on LinkedIn, or contact me directly. I am excited to hear your thoughts.

Be the first to comment

There are no comments for this article.