The Vibration That Kept Almost Making Sense

The machine was one of the main engine’s seawater cooling pumps. It had run well for years with no problems. During a normal round, I noticed a small, uneven vibration near the pump. It was not strong. It was not constant. It came and went over the next few days, worse on some watches and better on others. The sound was rough and uneven, which made me think of cavitation, not a simple balance problem.

That guess was not random. This pump’s strainer had a history of getting dirty faster than it should, especially in warm water with more marine growth. A dirty strainer starves the pump of water, and that causes cavitation — the same rough, uneven sound I was hearing. I opened the strainer and checked it. It was dirtier than I expected. I cleaned it, started the pump again, and the vibration clearly got better. That was my first diagnosis, and it was correct.

I want to point out that this was not a quick guess. I had two years of history with this pump and this strainer. I already knew the fouling was happening faster than the schedule expected. When I opened the strainer, I found exactly what I expected to find. Every step of this first diagnosis was solid. That matters, because it explains why nothing that came next felt like a mistake in the making.

The First Two Times I Was Right, Which Was the Problem

About three weeks later, the vibration came back. Same rough sound. I checked the strainer first, since that had worked before. This time it was clean. So I looked further up the system, at the sea chest — the opening in the hull where seawater enters. We found marine growth partly blocking the grating there. This also restricts water flow into the pump, and causes the same kind of cavitation sound. We cleared it and cleaned the system. The vibration improved again. That was my second diagnosis. It was also correct — as far as it went.

I want to stop on that phrase, “as far as it went,” because it is the center of the problem. Both fixes were real. Both solved a real restriction in the water supply to the pump. Both produced a real improvement I could hear and feel. Neither one was the actual root cause. And the fact that both fixes worked is exactly what made the real problem so hard to see. Two correct diagnoses in a row do more than solve two real problems. They also build a story in your head. By the time the third event happened, I was not asking a new question. I was confirming the story I already believed.

There is one detail I skipped over at the time, and I have thought about it a lot since. After the sea chest cleaning, the improvement was real but not complete. A small, faint roughness remained. I explained it away as the pump needing a short settling-in period after the blockage cleared. I never went back to check if that small roughness actually went away, or if it just became quiet enough that I stopped noticing it. I will never know for sure. But I now believe that faint, unexplained sound was the bearing problem starting — showing up at almost the same time as a second, unrelated problem I was correctly solving.

The Third Time I Stopped Looking

The vibration came back a third time, about four weeks after the sea chest cleaning. This time it was weaker than before, which made it easier to ignore. By now I had a working theory, backed by two confirmed cases: this pump runs a little rough when its water supply is restricted, we clear the restriction, the vibration settles down. I checked the strainer out of habit. It was lightly dirty, nowhere near as bad as the first time. I cleaned it anyway and moved on. This time the vibration did not fully go away. It got quiet enough that I filed it under “this pump has always run a little rough” — a line I do not think I would have accepted from anyone else without pushing back.

That is the moment I think about most, because nothing about it looked careless from where I stood. I had checked something. I had found something, even if it was small. I had a two-time-confirmed explanation for why this pump behaves this way. What I did not do — what I did not even think to do — was ask what this leftover vibration would look like if my theory was simply wrong this time, instead of only partly right.

I also want to be honest about the ordinary pressure sitting under that decision. This did not happen in some perfect, calm moment of pure thinking. It was a busy stretch. Cleaning the strainer took time out of a watch that did not have much time to spare. And a known, already-solved problem is always faster to close out than an unknown one. None of that excuses skipping a real investigation. But it explains why the shortcut was so easy to take that week. I think this combination — a believable existing theory, plus normal time pressure — explains this kind of mistake far better than any story about carelessness would.

What the Spectrum Actually Showed

The fourth event happened about four weeks after the third. This time it was stronger than any of the earlier three. And it happened to land in the same week as a planned visit from a shore-based reliability contractor — a service the company had started earlier that year for quarterly checks. On a quieter week, I might have told them not to bother checking this pump, since I already “understood” its behavior so well. Their equipment was different from our basic handheld meter. Ours only gives one overall vibration number. Theirs could break the vibration down by frequency — what is called a spectrum.

The spectrum showed one clear, narrow spike. We worked out the bearing’s size and the pump’s running speed, and calculated where a specific kind of bearing fault — a defect on the outer ring — would show up. The numbers matched closely. This is the signature you get once a small pit or crack forms on the bearing’s outer ring, and the rolling balls hit it once every turn. It repeats at a fixed rate, which is why it shows up as a sharp spike instead of a spread-out noise like cavitation. I already knew this difference in theory. I had simply never pointed the right tool at this specific pump, because a basic overall number was all I usually reached for — and one overall number cannot tell two different problems apart when they happen to sound alike.

Why Being Right Twice Made the Third Diagnosis Worse

I do not think this story is really about guessing instead of checking. I checked every single time. What actually caught me out was something quieter: two correct diagnoses in a row gave their credibility to a third, unrelated event that just happened to sound similar. By the third time, I was not asking “what could be causing this vibration.” I was asking “which part of the water supply needs attention this time” — a much narrower question. And a narrower question can only ever return an answer from inside the theory I already had. A bearing fault on the other side of the pump was never going to show up as the answer to that question, no matter how carefully I asked it.

What makes this hard to write about is that I do not think trying harder, in the normal sense, would have caught it. I was already trying. I was checking real things and finding real problems every time. The failure was not in the checking. It was in never stepping back far enough to ask if the checking itself had narrowed down to fit a story that was only half true.

I have talked to other engineers about this since, and I now think it is not really a personal weakness. It is a structural one. People usually describe this kind of mistake as something that happens when you are not looking closely enough. That description let me avoid seeing it in myself for a long time, because I was looking very closely indeed. What it actually does — at least in a case like this — is narrow the question before you even start looking. Not “what is wrong here,” but “which known cause is it this time.” Any answer to that second question will always come from the list of causes you already believe in, no matter how carefully you search.

What I Changed on Every Ship Since

• Any repeating vibration or noise now gets an actual spectrum reading, not just one overall number, starting from the second time it happens — even when a believable, already-confirmed explanation exists. A repeat event will always come with a believable explanation on its side.
• I have pushed for the ship to have its own spectrum-reading equipment on board, instead of waiting for the contractor’s quarterly visit. The fault in this story sat in the data for weeks before anyone with the right tool happened to be there to see it.
• I now write down the character of a repeating problem over time — frequency details where possible, not just “fixed” or “not fixed” — so a pattern spread over months is visible as a pattern, instead of something I have to remember from scratch each time.
• Before I close out a repeating problem, I ask one direct question: what would this look like right now if my current explanation is wrong. If I cannot answer that, the diagnosis is not finished, even if the fix I already made was completely real.
• I now treat two correct diagnoses in a row as a reason to check more carefully the third time, not less. A good track record on one piece of equipment is useful information. It is also exactly the condition where this kind of mistake happens most easily.

What This Actually Costs

We caught it in time. The bearing was replaced during a planned stop, a few days after the contractor’s reading, and the pump was back in service well before anything close to failure. When we opened it up, the damage was worse than I expected. The outer ring showed wear across a visible section, not just one small pit. It was further along than I would have wanted to leave running for another round of “it’s always run a little rough.”

A seized seawater cooling pump on a main engine is not, on most ships, an immediate disaster — there is a standby pump for exactly this reason. But losing the main pump without warning, instead of on a schedule you chose, means relying on that standby at a moment you did not pick. The real cost of those eleven weeks was not money. It was three months of a real problem developing completely outside my awareness, on a machine I genuinely believed I understood. What ended it was not insight on my part. It was a contractor’s spectrum reader arriving on the right week, for reasons that had nothing to do with my own thinking. That is not a comfortable thing to build confidence from. It is exactly why I stopped waiting for a quarterly visit to be the thing that catches the next one.

I do not think there is a version of this story where I get to say I earned the good result. The bearing was found in time because of a scheduling coincidence, not because I finally saw past my own theory on my own. That is harder to accept than a story about a close call, because at least in a close call you can point to the moment you noticed something. There was no such moment here. There was only a contractor’s van arriving on schedule, and three months of a pattern I had built for good, correct reasons — a pattern that happened to be pointing at the wrong half of the problem the entire time.