Too little time/products
Statement 11:
We get too little time and/or products for reliability testing
A majority of 74% of the participants agreed
This is one of the most heard complaints. The interesting fact is that there was no difference between the top-downers and bottom-uppers. This is remarkable because it is essentially the top-down who decides on number of products and time to test.
Both have their influence on Time to Market and cost price which are more important. They think!
If in the overall budget a reliability track is part of it then one sees that:
- Reliability starts often too late. It is a misconception that reliability testing is limited to final products. Also with a car one does not wait until the complete car is finished before reliability testing starts. That is already started long before with parts.
- Development planning delays. Because Time to market is important the reliability test period is shortened in order to reach a certain milestone.
An example.
- Warranty period 1 year, being 2000 working hours
- No more than 5% warranty calls accepted
- Available test time 1000 hours (6 weeks)
- Number of products 50
- Confidence level of the test 90%
Assuming a certain failure distribution the reliability statistics creates next test table
| Number of failures | ||||||
|---|---|---|---|---|---|---|
| Units | 0 | 1 | 2 | 3 | 4 | 5 |
| 50 | 1796 | 3064 | 4236 | 5374 | 6500 | 7623 |
| 100 | 898 | 1524 | 2096 | 2645 | 3180 | 3709 |
| 150 | 599 | 1014 | 1393 | 1754 | 2106 | 2451 |
| 200 | 449 | 760 | 1043 | 1312 | 1574 | 1830 |
What does it mean? With 50 samples one needs 1796 hours test time and no failure may occur
- This can not be done within the time frame of 1000 hours.
- So, what next? Take 100 units then only 898 hours needed.
- And what happens if we find a failure?
- Bad luck, continue the test until 1524 hours, so one needs an extra 3 weeks.
- But if another failures occurs you have to continue until 2645 hours
- With 200 samples you may have 2 failures within that 1000 hours of test time
The conclusion is that it is not so easy to take just 50 units and test until 1000 hours.
From a reliability point of view it is strongly advised to have at least 5 failures to get a little impression. But it can not be done within the available timeframe. For that you need 350 units. Is it possible to have so many units? Sometimes it looks like a mission impossible
In reality reliability testing is more complicated than this simple example, but it is just for the picture. There are more options possible, a task for reliability engineering to do it as efficient as possible.
Why did 74% agree with this statement? Simply because one is confronted later on with unexpected failures. The idea is that if one had enough test units and enough time they would have found earlier in the design cycle, saving a lot of money.