Root Cause Analysis for Repeat Machine Failures

Replacing a failed part gets the machine running. It does not explain why the part failed. If the same pump eats a bearing every five months, the bearing is not the problem. It is the symptom, and each repair just resets the clock. Finding the real cause is one of the most useful skills a maintenance technician can build, because each repeat failure you end gives back hours for every year that machine stays in service. Reminder: all teardown and inspection work starts with lockout and stored energy released. Electrical diagnosis stays with a qualified electrician.

What You Will Be Able to Do

• Use failure records to pick which problem to work on first.
• Save the evidence before it gets cleaned or thrown away.
• Read common bearing damage and link it to a cause.
• Trace a failure through its physical, human and system causes.
• Choose a fix that addresses the cause, and prove it worked.

Finding the Bad Actors

You cannot run a full analysis on every failure. Start where it pays. A bad actor is a machine or part that takes a large share of your downtime or repair effort. A simple Pareto table finds them. It ranks problems from largest to smallest.

Worked example. Twelve months of work orders for one area:

The total is 173 hours. P-204 and Conveyor 3 together make up 60 percent, so they go first. Note that the most frequent failures are not always the costliest. Conveyor 3 fails more often, but P-204 costs more downtime.

This only works if the records are good. If work orders say "fixed it" or use no failure codes, you cannot rank anything. Good records start with the person holding the wrench.

Save the Evidence

Evidence is often lost in the first hour of a repair. A failed bearing goes into the scrap bin, and the housing gets washed. After that, all you have is an opinion.

Before and during teardown:

1. Take photos before you clean anything. Show the part, its position and the area around it.
2. Write down the conditions. Note the time, load, speed, temperature, alarms and any recent work on the machine.
3. Take a sample of the oil or grease before you drain or wipe it.
4. Mark the orientation. Note which side of a bearing faced up, or which way the load ran, so damage can be matched to the load.
5. Measure. Check the shaft and housing fits, runout, alignment and bolt torque before you disturb them, where you can do so safely.
6. Keep the parts. Tag them with the asset, date and work order, and keep them dry. Do not wash off the deposits. Coat them only if your procedure says so.

Note what you see. Leave the conclusions until you have more information.

Reading a Failed Bearing

A bearing keeps a record of how it was treated. ISO 15243 is the international standard that classifies bearing damage by how it looks and what causes it. Its main groups are fatigue, wear, corrosion, electrical erosion, plastic deformation, and fracture and cracking. You do not need to memorize the standard. You need to learn the common patterns.

The location of the marks matters as much as how they look. A load zone in the wrong place tells you the force was going somewhere it should not be. Your bearing supplier can often look at a failed bearing, and a lab can go further if the machine is important enough.

The Layers of a Cause

Most failures have three layers. If you stop at the first one, the problem returns.

The system layer is not about blame. If a good person could make the same mistake tomorrow, the system is still broken. A fix at this layer protects every machine, not only the one in front of you.

A Worked Investigation

The problem: pump P-204 has lost its drive-end motor bearing seven times in about three years, at about five months each. Each repair replaced the bearing, and each time the pump ran well at first.

Step 1: State it with numbers. "Drive-end bearing on P-204 fails about every 5 months. Design life is several years. The non-drive-end bearing has never failed. The twin pump, P-205, has no bearing failures."

The comparison is valuable. Same motor, same fluid, same grease and same crew, yet one pump fails. That rules out a lot.

Step 2: Look at the evidence. The failed bearing shows a wear path angled across the race. There is no discoloration. The grease looks fine. That points to misalignment, not to lubrication.

Step 3: Ask why, with proof at each step.

A soft foot is a machine foot that does not sit flat on its base. Tightening the bolt then bends the frame. To check, loosen one foot at a time with a dial indicator on it, or use feeler gauges at the foot. Many alignment procedures aim for about 0.002 inch of movement or less at any foot. Use the maker's or your site's limit. Pipe strain means the piping pulls on the pump, so loosening a flange lets the pump move.

Step 4: Find the fix at each layer.

• Physical: correct the soft foot with proper shims, and rework the pipe support so the flange lines up with no force.
• Human: align to tolerance, with the final reading recorded.
• System: add soft foot, pipe strain and final alignment readings to the repair task, with limits. Check the other pumps for the same issue.

Before you fit a new bearing, also check that you are not installing into a damaged housing or a worn shaft. Reusing them may guarantee the next failure.

Common Repeat Failures

Warning: Do not reset a breaker or swap a fuse again and again to keep a machine running. Repeat trips mean something is wrong, and a qualified electrician has to find out what. Never replace a fuse with a larger size.

Choosing a Fix That Lasts

Some fixes hold, and some only look like action.

A shorter PM interval on a part that should last for years is a warning sign. It may keep you running, but it spends labor and parts to live with a problem you could end.

Changes to a machine's design, guards, or safety functions need the right engineering review. Do not change a pressure system, a lifting device or a safety device on your own.

Proving the Fix Worked

Set your measure before you close the job.

Worked example. P-204 had 7 failures in 35 months of running, so its MTBF (mean time between failures) was 35 / 7 = 5 months.

1. Just after the repair, take a vibration baseline. Radial and axial readings should be clearly lower than before. Record final alignment and soft foot.
2. At 5 months, the old failure point, there should be no bearing tones in the spectrum.
3. At 12 months, there have been no failures. That is 2.4 times the old MTBF. The fix is looking real, but it is not yet proof of a long life.
4. Over the next years, keep trending it. Compare downtime in the next Pareto.

If the problem returns, the analysis was wrong or incomplete. Go back to the evidence. There may have been more than one cause.

Close the loop by updating the CMMS. Add a failure code, attach the photos and the report, change the PM tasks and the repair task, and tell the crews on other shifts. Share what you found with people who work on similar machines.

Common Mistakes

Check Your Understanding

1. Asset A has 12 short failures for 15 hours of downtime. Asset B has 3 failures for 45 hours. Which do you work on first if downtime is the main cost?

Asset B. It has a larger share of the downtime, even with fewer failures. Do look at whether Asset A's failures have a simple shared cause.

2. A failed bearing shows a band of wear running at an angle across the race, and the grease looks good. What is the likely cause?

Misalignment. Check alignment and soft foot before you fit a new bearing.

3. Two identical fans run in the same room. One breaks belts every 2 months, and the other never does. What does that tell you?

The cause is something that differs between them, such as sheave alignment, a worn sheave, or how it was installed. Shared items, such as the belt brand and the room, are less likely.

4. A VFD-driven motor keeps failing bearings, and the races show washboard lines. What do you do?

Those marks point to electric current through the bearing. Save the bearing, and have a qualified electrician or engineer look at grounding and the drive. Replacing the bearing alone will repeat the failure.

5. A pump had 6 failures in 30 months. After a fix, it has run 15 months without a failure. Is it proven?

Not fully. The old MTBF was 5 months, so 15 months is three times that, which is a good sign. Keep trending until the data shows a long life.

Key Takeaways

• Use a Pareto of real work order data to choose where to start.
• Save the part, the lubricant and the photos before anyone cleans up.
• A failed bearing shows how it was treated, so read the pattern before you fit a new one.
• Work through physical, human and system causes. Stopping early brings the failure back.
• Compare a failing machine with a healthy twin to rule out causes fast.
• Prefer fixes that remove the cause, and prove them with data over time.