A failure is a function that is not fulfilled. That sentence explains why the third step of the FMEA method comes before the fourth, and why teams that skip it end up with rows like “crimp not OK”.
Step 3: write the function
For each process step the team writes what the step must achieve, with a requirement that can be measured. Then it does the same for the work elements.
In the hose assembly cell H2, step 40 looks like this:
- Step 40, crimp fittings: fix both fittings to the hose. Requirement: crimp diameter 24.6 mm plus or minus 0.15 mm, ferrule at the stop.
- Crimper: close the dies to the programmed diameter. Requirement: the program matches the order.
- Die set: form the ferrule evenly. Requirement: the die set matches the ferrule and is serviced every 20,000 crimps.
- Setter: select program and die set and confirm the first part.
The requirements come from the drawing and the specification. A characteristic that the customer has marked as special must appear here. If it does not, the FMEA has a hole before the failure analysis begins.
A product characteristic is measured on the part: diameter, length, angle. A process characteristic is set on the process: test pressure, hold time, cuts per blade. The control plan will later check both.
Step 4: turn the function around
Now each function is reversed. What if it is not fulfilled? The answer is a chain of three links:
- Failure mode: how the step fails, in the words of the requirement. “Crimp diameter too large.”
- Effect: what happens then. “Fitting pulls off under pressure, risk of injury.”
- Cause: why it could happen, at the work element. “Wrong crimp program or die set selected.”
A few patterns help to find failure modes: the function is not done at all, done too much or too little, done on the wrong part or with the wrong setting, done but with damage, or it changes over time.
Three effects, one severity
The effect is described for three parties: your own plant, the customer plant and the end user. The worst one sets the severity.
Take the fitting that is not pushed on to its insertion depth. In your own plant it may pass the proof test, because the crimp still holds for ten seconds. At the customer plant nothing is visible and the hose is installed. In the field the hose comes off under load and oil sprays. The third effect counts.
If you cannot judge what happens at the end user, ask the customer. Their answer belongs in the FMEA with a date.
The cause test
A cause is specific enough when you can name an action against it.
| Weak | Better |
|---|---|
| Operator error | Similar fittings in neighboring bins |
| Machine problem | Die wear and pressure drift, no check in the run |
| Not cleaned | Cleaning shot skipped under time pressure |
One failure mode can have several causes. Each cause gets its own row, because each needs its own action.
Where the rows come from
The complaints and rejects of comparable processes: every one of them is a row. The FMEA of the sister process, used as a list of ideas and never as a copy. And the question to the setter: what goes wrong today that nobody writes down?
The result for cell H2
The team found 24 causes in 7 steps: 4 at cutting, 2 at cleaning, 4 at inserting the fittings, 5 at crimping, 3 at the proof test, 3 at capping and labeling and 3 at packing. Eight of them have an effect that can hurt someone or break a legal requirement.
Beside each cause the team wrote what prevents it today and what would detect it. Often the honest answer was “training” and “the operator would see it”. The next part shows what such answers are worth.