The failure that keeps coming back: when 5 whys aren't enough

It's the fourth time this year the same pump has stopped with the same symptom. The team replaces the bearing, aligns it, retightens the base, the pump runs again, and three months later it's down again. When a failure comes back, the previous analysis stopped halfway.

The 5 whys work very well when the failure has a straight-line chain of causes. Recurring failures usually have more than one cause at the same time. When you follow a single thread of whys, you fix one of them, the failure goes away for a while and comes back when the others line up again.

The three layers of a failure that comes back

• Physical cause: what broke and how. E.g.: bearing seized from lack of lubrication.
• Contributing cause: what sped it up or made it easier. E.g.: misalignment that raises the load and the heat.
• Latent cause: the failure in the management system that let the others happen. E.g.: there's no alignment procedure, and the lubrication plan doesn't cover that bearing housing.

If the action only tackles the physical cause (replace the bearing), the failure comes back. If it tackles the contributing cause (align), it takes longer to come back. Only by tackling the latent cause (procedure and plan) does it tend to go away for good.

Ishikawa: open up before you narrow down

The Ishikawa diagram, or fishbone, forces the team to look at six categories before picking a hypothesis. For pump B-11, which had its bearing replaced four times in ten months:

Category	Hypotheses raised
Machine	Cracked base, causing misalignment after startup
Method	Alignment done "by eye", with no procedure
Material	Different bearing brand in the last two replacements
People	Replacement done on the night shift, by a technician with no alignment training
Environment	Floor washing water reaching the bearing housing
Measurement	No vibration or temperature measurement after the replacement

Ishikawa only raises hypotheses. The next step is to look for evidence for each one: measure the alignment, inspect the base, open the bearing that came out (see why bearings fail early), check the purchase records for the parts. In the example, three were confirmed: cracked base, alignment with no procedure and water in the bearing housing. For each confirmed hypothesis, apply the 5 whys until you reach the latent cause.

Simplified FMEA: what to tackle first

When the equipment has several failure modes in its history, FMEA (failure mode and effects analysis) helps you prioritize. For each mode, give scores from 1 to 10:

• S (severity): how much the failure hurts.
• O (occurrence): how often it happens. Use the work order history, not a hunch.
• D (detection): how hard it is to notice before it fails. 10 = you only find out when it breaks.

Multiply: RPN = S × O × D. The highest number comes first.

Failure mode (pump B-11)	S	O	D	RPN
Bearing seized from misalignment and water	7	8	6	336
Cavitation from low tank level	6	3	7	126
Mechanical seal leak	6	5	3	90
Coupling element wear	5	2	4	40

FMEA also shows where to reduce the risk: lower the occurrence (fix the cause) or improve detection (inspection, measurement). A high D score calls for an inspection routine that catches the failure earlier.

Fault tree: when two things have to happen together

Some failures only happen when two conditions meet. The fault tree starts from the event and works down, separating what's "or" from what's "and":

• Pump B-11 stopped, which happens when
  • the motor tripped on overload, OR
  • the bearing seized, which only happens when there's
    • higher than normal load (misalignment from the cracked base), AND
    • compromised lubrication (water in the bearing housing).

The tree explains why the pump sometimes lasted months and sometimes weeks: it depended on both conditions showing up together. It also shows that, on an "AND" branch, removing one of the conditions is enough to cut that path; on an "OR" branch, you have to deal with each cause separately.

Which one to use

Situation	Tool
Failure with an apparently single cause	5 whys
The team doesn't know where to start	Ishikawa, then 5 whys on each confirmed hypothesis
Several failure modes on the same equipment	Simplified FMEA
Failure that only shows up sometimes, when conditions combine	Fault tree

Close the loop

1. One action per confirmed cause, with an owner and a deadline, including the latent cause (procedure, plan, training).
2. Record the analysis with the equipment history, so the next team doesn't start from scratch.
3. Follow up: did the time between failures on the equipment go up? Did it come off the repeat list? The % sick units calculator shows whether the fleet as a whole improved.

In taskfield, every corrective work order stays in the asset history with symptom, cause and fix, and return work orders are linked to the original. A failure that comes back shows up right away, not at the year-end meeting.