The Vibration You Can’t See That Snaps Crankshafts

The crankshaft broke clean across the main journal, about two years into the set’s life. A 1-megawatt machine, beautifully maintained — oil samples like clockwork, load bank every year, a load bank test just six months earlier that it had passed without a shrug. Nobody had abused it. The maintenance file was boring in all the right ways. And one Tuesday the engine simply seized, and when they opened it up, the crank was in two pieces.

I wasn’t the failure investigator on that one. I read the report, the way you read an obituary of someone you almost knew. Torsional vibration. The one failure mode that leaves no early symptoms you can hear, and the one report everybody files without reading.

Here’s the thing nobody tells you at the quoting stage: a diesel engine doesn’t just shake. It twists. Every power stroke from every cylinder hits the crankshaft as a hammer blow — a pulse of torque, thousands of times per minute. The shaft twists a little with each blow, then springs back. Most of the time, nothing comes of it. But every rotating system has a natural frequency — a speed at which it wants to twist on its own — and if those hammer blows land at exactly that frequency, the twisting adds up instead of cancelling out. Resonance. The shaft starts winding up like a rubber band. And steel, given enough cycles, gives up. Fatigue cracks don’t care how clean your oil is.

You can’t feel it. You can’t hear it. It still breaks things.

That’s what makes torsional vibration the industry’s unloved problem. Lateral vibration — the shaking you feel in your boots — is honest. It rattles the foundation, it shows up on a vibration meter, and everybody respects it. Torsional vibration lives inside the shaft. A normal vibration survey, the kind most maintenance teams run, never looks at it. The sensor sits on the bearing housing and measures how the outside moves. The danger is in how the inside twists. Different axis. Different instrument. Most of the time, a different engineer entirely.

The engine manufacturers know this. Every big engine maker publishes torsional vibration data for its engines — a stack of charts and limits that ships with the engine documentation, assuming anyone reads it. The coupling vendors, for their part, refuse to pick a coupling for a large set without a torsional analysis. They’ve been burned too many times. When you see an elastomeric coupling with torn rubber blocks at 800 hours, the coupling isn’t the disease. It’s the symptom.

When does this actually matter to you?

Honest answer: if you’re buying a 100 kVA standby set for a small building, it probably doesn’t. Standard single-bearing gensets in that size range have been built the same way for decades, and the manufacturers did the math once, long ago. Your engine, your alternator, your standard coupling — the combination is known-good.

The danger zone starts where the machines get big and the combinations get custom. A 1.5 MW set with a two-bearing alternator. A 2.5 MW prime-power machine with an unusual driveline — a PTO driving a fire pump off the front of the crank, or a set ordered with a coupling the catalog doesn’t usually pair with that engine. Anything that runs at variable speed. Anything where somebody swapped a component for a “same spec” alternative that isn’t actually the same where it counts.

I’ve watched this happen in paralleling projects too. Two sets run fine alone. Tie them together on one bus, and now each engine’s torque pulses talk to the other through the electrical connection, and the coupling between them absorbs conversations it wasn’t designed for. The sets sync up fine. The electrical guys sign off. Six months later the couplings are warm, then hot, then torn. Nobody at the sync panel ever saw it coming, because electrically everything looked beautiful.

My rule of thumb: if the set is big enough that the factory orders a special coupling for it, it’s big enough to deserve a torsional vibration analysis. If the engine maker says the TVA is required for the warranty on your driveline configuration, believe them. They wrote that clause with someone else’s crankshaft in mind.

What a proper torsional analysis actually checks

A TVA is not a vibration measurement. It’s a calculation — a mathematical model of the whole rotating line, built from what engineers call the mass-elastic data: the inertia of the flywheel, the crankshaft, the coupling halves, the alternator rotor, plus the torsional stiffness of each piece between them. Feed in the engine’s firing order and the torque pulse shape, and the model spits out the system’s natural frequencies and how the operating speed relates to them.

What you’re reading for, if you ever get your hands on one of these reports, is embarrassingly simple. First: where are the critical speeds — the speeds where resonance happens — and how far are they from your running speed? A healthy analysis shows the operating point sitting in a quiet zone, with comfortable margin on both sides. A bad one shows a critical speed sitting right at, say, 1500 rpm, which is exactly where your 50 Hz set lives all day. That set is a fatigue accident billing by the hour.

Second: the coupling. Every flexible coupling has a torsional stiffness and a damping rating, and the analysis uses both to predict how much the coupling will twist and how hot it will get doing it. Rubber-element couplings are wonderful at absorbing shock and at moving a resonance away from the running speed — engineers call that detuning, and it’s the whole reason the coupling exists. Steel disc or grid couplings are stiffer and last longer but offer less damping. Picking wrong in either direction buys you a different kind of trouble.

Third: shaft stress. The report estimates the alternating torque in each shaft section at resonance and compares it against the fatigue limits of the steel. This is the line in the report that, read properly, tells you whether the crankshaft has a long career or a countdown.

A good TVA is boring to read. Quiet zones everywhere, margins generous, coupling temperatures sane. That’s the point. The excitement belongs in the movies.

The coupling is the cheapest insurance on the machine

Here’s where the industry quietly splits into two camps. The experienced packagers treat the coupling as a tuned component — chosen from the TVA for that specific engine-alternator combination, stiffness and damping matched, sometimes with a different element after the analysis says the first choice put a resonance in the wrong place. The cheap packagers treat it as a commodity: this engine frame takes this coupling part number, same as last time. Same engine, different alternator, different flywheel inertia, different driven equipment — doesn’t matter, bolt it up.

I’ve seen the torn-rubber-coupling photo from the second camp more times than I can count. Warm rubber. Cracked rubber. Rubber blocks that look chewed. And every time, somebody’s first instinct is to blame the coupling brand. Honestly? The brand is usually fine. The system told the coupling to live at a resonance, and the coupling died doing what the math said it would.

An old vibration consultant I know has a line he repeats at every site visit: “The coupling is the fuse.” He means it kindly. A coupling failing early — visibly, messily, cheaply — is the system screaming that the torsional picture is wrong, at a price point you can survive. The alternative is the shaft absorbing that same energy silently until one Tuesday it doesn’t. Given the choice between replacing a two-thousand-dollar coupling every couple of years and replacing a crankshaft once, the fuse is the better deal. But the real fix is the analysis, done before the steel gets ordered.

So what should a buyer actually do?

Three things, in order of cheapness. First: at the quoting stage, ask whether your configuration needs a torsional vibration analysis. Big set, unusual driveline, PTO equipment, anything non-standard — ask. The packager either has one, will commission one, or gets very quiet. The quiet ones are telling you something.

Second: read the report, or have someone you trust read it. You don’t need to follow the math. You need the three answers above: critical speeds clear of the operating range, coupling chosen from the analysis, shaft stresses inside limits. If the report shows a critical speed near running speed with a note saying “acceptable,” make somebody explain acceptable to you in plain language. In my experience, that note is where the trouble lives.

Third: after commissioning, check coupling temperature and condition at every service for the first year. A coupling running hot or shedding rubber dust is the fuse talking. Listen to it. A quick infrared check during the monthly run takes thirty seconds and has caught more torsional problems than most people would believe.

Torsional vibration will never be the exciting part of buying a generator. Nobody puts it in the brochure. But the crankshaft in that two-year-old set didn’t know that, and the maintenance file didn’t save it. The analysis would have. If you’re specifying a machine big enough to matter, talk to our engineers about the driveline before you sign — it’s a cheap conversation, and it’s a lot cheaper than the other one.