The Bearing That Failed for the Third Time
A rotary indexer on one of our machines kept failing. The bearing lasted three months, got noisy, and we replaced it. Three months later, same bearing, same noise. We replaced it again with a higher-grade bearing. Three months later, it failed again. Finally we stopped replacing parts and started asking why. The root cause wasn’t the bearing — it was that the indexer was running at 70% above its rated load because the customer had started running a heavier part through the station. The bearing was the symptom. The oversizing was the disease.
Machine failure root cause analysis is where most engineers fall into the trap of fixing the symptom instead of the cause. We see a broken part, replace it, and move on. But if you don’t ask why it broke, it breaks again — often with a worse failure the second time. This article is the process I use when a machine comes back with a problem, from the first phone call to the permanent fix.
The First Question: Is It Really Broken?
The customer calls and says “the machine is down.” Before you start troubleshooting, establish what “down” means. “It stopped” isn’t enough. You need to know:
- What is it actually doing? (Or not doing?)
- What was happening when it stopped? (Mid-cycle? At startup? After a changeover?)
- Has anyone touched anything? (Maintenance, changeover, a part jammed?)
- What does the HMI say? (There’s usually an alarm code or a fault message. Read it.)
- What parts are involved? (A specific station? A specific axis? A specific actuator?)
This sounds obvious. But half the service calls I’ve been on started with “the machine is broken” and ended with “the air supply was turned off.” The first five minutes of diagnosis is asking better questions, not opening panels.
Don’t Replace Parts Yet: The Five Whys
The instinct is to replace the broken part. Resist it. The broken part is evidence, not the problem. Use the “five whys” technique to dig past the symptom.
How the Five Whys Work
- Why did the machine stop? The gripper didn’t close, and the cycle timed out.
- Why didn’t the gripper close? The gripper cylinder didn’t extend.
- Why didn’t the cylinder extend? The air pressure to the station was low.
- Why was the air pressure low? The filter was clogged and the regulator couldn’t maintain pressure.
- Why was the filter clogged? It hadn’t been changed in 18 months, and the maintenance schedule didn’t include filter replacement.
The broken part (the gripper cylinder, which the customer replaced) was never the problem. The problem was a clogged filter and a missing maintenance step. Replacing the cylinder fixed it for a week. Fixing the filter and adding it to the maintenance schedule fixed it permanently.
You don’t always need five whys. Sometimes three gets you there. The point is to keep asking “why” until you reach a cause you can actually fix. If the answer is “the part failed,” you haven’t asked enough.
The Three Failure Categories: What You’re Actually Looking For
Most machine failures fall into one of three categories. Identifying which one you’re in determines where to look.
1. Wear and Tear (Expected but Too Fast)
A part wears out. Seals leak, belts stretch, bearings get noisy. The question isn’t “why did it wear?” — it’s “why did it wear faster than expected?”
- Is the part misapplied? (The 6208 bearing in the earlier example — wrong size for the load.)
- Is the environment harsher than expected? (More dust, more moisture, higher temperature.)
- Is the maintenance interval too long? (Grease not re-applied, filter not changed.)
- Is the duty cycle higher than designed? (The machine running 24/7 instead of one shift.)
2. Random Failure (Something Unexpected Broke)
A part that should last years failed in months. A bolt sheared. A tube cracked. A sensor failed early.
- Was it a defective part? (Single part failure, not a pattern.)
- Was there an overload event? (A jam, a crash, a voltage spike.)
- Was the installation wrong? (A bolt not torqued, a connector not seated.)
- Is there a design weakness? (Vibration loosens a fastener every time.)
3. Operational Failure (The Machine Did What It Was Told, but It Was Wrong)
The machine runs. It cycles. But the output is bad — wrong parts, missing operations, inconsistent results.
- Is the part variation outside the designed range? (The customer’s supplier changed the part geometry.)
- Is the sensor reading drifting? (A photo-eye covered in dust, a prox shifted.)
- Is the program behaving differently? (A parameter changed, a recipe was loaded wrong.)
- Is the machine not calibrated? (A vision system drifted, a linear axis lost its home position.)
Data Collection: Before You Touch Anything
When you’re on the floor (or talking to the customer), collect data before you start changing things. The first thing you touch can destroy the evidence.
What to Record
- The HMI alarm history. What was the last alarm? How many times has it occurred? The fault log tells you whether this is a new failure or a recurring one.
- The cycle count and runtime. How many cycles since the last maintenance? How many hours on this part? If a bearing fails at 10,000 cycles, it’s not normal wear. If it fails at 2,000,000 cycles, it might be.
- Visual evidence. Take photos before you start. A broken part on the floor tells you more than the part after you’ve handled it. A clogged filter photographed before cleaning tells the root cause.
- Measurements. If the pressure is low, measure it at multiple points. If the sensor isn’t triggering, measure the signal voltage. Don’t guess — measure.
The rule: If you don’t have the data, you don’t know the cause. You have a hypothesis. Test the hypothesis before you implement the fix. The fix that doesn’t address the root cause gets reversed later.
Common Failure Patterns: What to Look For
After enough machines, patterns emerge. Here are the ones I’ve seen repeatedly.
Pneumatic Failures
- Clogged filter: Slow cylinder motion, low pressure, inconsistent operation. Check the FRL filter first.
- Leaking fittings: Air pressure drops when a cylinder cycles. Soapy water on fittings finds the leak.
- Worn seals: Cylinder drifts or doesn’t hold position. Rebuild or replace the cylinder.
- Kinked tubing: A line that flexes every cycle fatigues and cracks. Look for tubing that bends at a fixed point.
Electrical Failures
- Loose connectors: Vibration shakes a connector loose. The machine works intermittently. Check connectors before blaming the PLC.
- Frayed cables: Drag chain cables fatigue at the flex point. Look for cables that bend at the same point every cycle.
- Sensor drift: A photo-eye gets covered in dust or oil. The sensor reads when it shouldn’t. Clean and re-teach.
- Voltage drop: A long wire run drops voltage under load. Sensors don’t switch reliably. Check voltage at the sensor, not at the PLC.
Mechanical Failures
- Loose fasteners: Vibration loosens bolts over time. A bracket shifts, a sensor moves, a rail shifts. Check fastener torque on anything that vibrates.
- Wear in moving joints:</strong) Bushings wear, linear guides develop play, bearings get noisy. The symptom is always “it’s not as accurate as it used to be.”
- Misalignment: A frame shifts, a rail settles, a pulley moves. The symptom is the same wear, but the cause is that something moved out of alignment.
- Contamination: Chips, dust, or oil get into bearings, guides, or gearboxes. The machine gets noisier and less accurate. Keep the wipers and seals maintained.
The Permanent Fix: More Than a Part Replacement
Once you’ve identified the root cause, the fix should prevent recurrence. A part replacement is a temporary fix. The permanent fix addresses why it failed.
Three Levels of Fix
- Patch: Replace the broken part. Gets the machine running today. Doesn’t prevent recurrence.
- Corrective: Address the immediate cause. If the filter clogged, change the filter and add it to the maintenance schedule. If a bolt came loose, add threadlocker.
- Preventive: Address the system that allowed the failure. If filters clog because the air supply is dirty, install a better filter upstream. If bolts loosen because of vibration, redesign the bracket to be stiffer.
Good engineering gets to level 2. Great engineering gets to level 3. The patch is what the customer asks for when they’re down. The preventive fix is what prevents the next service call.
Documentation: The Failure Report
When you’ve fixed it, write it down. A one-page failure report:
- What happened: The symptom, when it occurred, and what the customer saw.
- What we found: The root cause (from the five whys, not the symptom).
- What we did: The immediate fix (what was replaced or adjusted).
- What we’re changing: The permanent fix (design update, maintenance addition, procedure change).
- Follow-up: When will we check it again? Who is responsible?
This report goes into the machine’s file. The next time a similar machine has a similar failure, you don’t re-diagnose from scratch. You look up the last time this happened and apply the known fix. That’s how automation downtime reduction actually works — not by reacting faster, but by not repeating the same diagnosis.
A Troubleshooting Checklist
- What exactly is the machine doing (or not doing)? Get the alarm code.
- What was happening when it failed? (Cycle position, changeover, startup?)
- Has anyone changed anything? (Part, program, maintenance, setup?)
- What does the fault log show? Is this recurring or new?
- Follow the five whys: don’t stop at the broken part.
- Measure, don’t guess. Pressure, voltage, signal, torque.
- Categorize the failure: wear, random, or operational.
- Collect evidence (photos, measurements) before changing anything.
- Implement the corrective fix, not just the patch.
- Document the root cause and the preventive action.
The Bottom Line
Custom machine troubleshooting isn’t about being the fastest part-replacer on the floor. It’s about asking why until the answer is something you can fix. The bearing that failed three times wasn’t a bad bearing — it was an undersized bearing on an overloaded station. The gripper that wouldn’t close wasn’t a bad cylinder — it was a clogged filter with no maintenance schedule. Replace the part to get the machine running. Then ask why it broke. The answer is the fix that keeps it running.
