Why Most Rotating Equipment Failures Start With Small Numbers

I've pulled bearings that looked perfect when we installed them.

Faces were clean. Clearances were in spec. Installation was by the book.

Three months later, they failed.

Same with seals. Same with couplings. Same with pump performance.

The problem wasn't the parts. The problem was that we missed something small. A couple of mils of soft foot. A pump running 15% below BEP. Uneven torque on two bolts. Shims that weren't seated flat.

Small numbers. Easy to ignore. Easy to rationalize.

"It's within tolerance."

"Close enough."

"We'll catch it next time."

Except we don't catch it next time. Because by then, the bearing's already running hot or the seal's already leaking.

Most equipment failures don't announce themselves. They start quietly. A mil here. A degree there. A few PSI of pressure instability that "doesn't look like a problem."

Then three months later, you're pulling the machine apart.

Take soft foot. Industry standard says anything under 2 mils is acceptable.

I've seen guys check soft foot, get a reading of 1.8 mils, and call it good.

"We're within tolerance."

Technically, yes. But here's what happens: You torque the hold-down bolts. That 1.8 mils of soft foot lets the frame distort just enough. The bearing bores shift—maybe half a mil. The shaft moves. Alignment changes.

Not much. Just enough.

Six weeks later, vibration starts climbing. Coupling shows wear. The inboard bearing runs 15°F hotter than it should.

We had three months to fix 1.8 mils. Instead, we spent two days replacing a bearing.

I aim for 0.001 inches or less now. It takes an extra 10 minutes with shims. But I don't get called back.

Most seal failures I see aren't seal problems.

They're operating point problems.

A centrifugal pump is designed to run at Best Efficiency Point (BEP). That's the flow rate where hydraulic forces balance out. Impeller loads are even. Pressure is stable. Everything works the way it's supposed to.

Run that same pump at 60% of BEP, and everything changes.

Low flow creates recirculation at the impeller eye. Recirculation creates pressure fluctuations. Those fluctuations hammer the seal faces. The seal tries to track the movement, can't keep up, and starts leaking.

People pull the seal, see the damage, and blame the seal.

But the seal wasn't the problem. The pump was running off its curve.

I've seen pumps eat three seals in six weeks because they were operating at 45% flow. We moved the operating point back to 80% BEP. The same seal design ran for 18 months.

The seal didn't change. The operating conditions did.

Before you blame the seal, check the curve. Pull the flow reading. Check suction and discharge pressure. Plot it. If you're below 70% BEP or above 120% BEP, you're outside the stable range.

That means suction recirculation, shaft deflection, and cavitation. And that kills seals faster than any mechanical issue.

I watched a technician align a motor once. Alignment looked good on the laser. Vertical and horizontal readings were in spec.

Then he torqued the bolts. Started at one corner, went around in a circle. Tightened each one to full torque before moving to the next.

I asked him to check the alignment again after the torque. He didn't want to—said it was already good.

I insisted. We measured again.

Vertical alignment had shifted 3 mils. Horizontal had moved 2 mils. Both were out of spec now.

Torquing pulls the machine. If you don't follow a cross-pattern and tighten in stages, you distort the frame and shift alignment.

It's in every alignment manual. But people skip it when they're in a hurry.

I follow a star pattern now. Snug all bolts first. Then go around again, tighten to 50%. Then 75%. Then full torque. Then verify alignment.

Takes five extra minutes. Saves me from coming back two weeks later when vibration shows up.

Here's one most people don't think about: shim seating.

You stack shims under a foot. Looks good. Bolt it down.

But if those shims aren't flat, if there's dirt between them, if they're bent—they compress unevenly when you torque. The machine settles. Alignment shifts.

I've seen alignment move 2-3 mils after startup just because shims settled.

Now I clean every shim before I stack them. Flat file the foot pad if there are burrs. Use no more than three shims per stack. Check that they're seated properly before I torque.

Small step. Big difference.

Every time I get called to troubleshoot a "bad bearing" or a "bad seal" or "high vibration," I check the small numbers first.

How much soft foot?

Where's the pump running on its curve?

What was the torque sequence?

How many shims are under each foot?

Was alignment verified after torque?

Nine times out of ten, the failure traces back to something we could've measured and fixed during installation.

We didn't fail because the bearing was bad. We failed because we accepted 2 mils of soft foot. Or we skipped the post-torque alignment check. Or we didn't verify the operating point.

Small numbers. Easy to miss. Easy to justify.

But they add up.

Here's what I check now on every alignment job:

Not "under 2 mils." Aim for 1 mil or better. Shim it out. Verify it. Don't accept the tolerance as the target.

Pull flow, suction pressure, and discharge pressure. Plot it on the curve. If it's outside 70-120% BEP, fix the operating conditions before you blame the equipment.

Star pattern. Three stages. Verify alignment after final torque. Every time.

Three shims max. Flat and clean. Seated properly. No dirt, no burrs, no bent shims.

Measure alignment again after you torque. If it shifted, correct it. Don't walk away assuming it held.

This takes an extra 20-30 minutes per job.

But I don't get called back to fix the vibration. I don't pull bearings that failed early. I don't chase seal leaks that shouldn't be leaking.

The tolerances exist for a reason. But the tolerance isn't the target.

If you're sitting at 1.8 mils of soft foot and calling it good because you're "within spec," you're gambling that thermal growth, settling, and operating loads won't push you over.

Sometimes you win that bet. Sometimes you don't.

I'd rather just fix the 1.8 mils.

Precision isn't about being perfect. It's about respecting the numbers that matter.

Soft foot, operating point, torque sequence, shim quality—these aren't theoretical concerns. They're the difference between equipment that runs for years and equipment that fails in months.

Most teams have the skills. Most teams have the tools.

What separates high-reliability teams from average ones is whether they take the time to check the small numbers.

Because those small numbers? They're not small at all.

See you next Friday.

What's your experience with tolerances? Do you aim for the minimum spec or do you go tighter?