MTTF population

Ikigai

Registered
Hi all,

I am trying to calculate the MTTF of our electromechanical system after gathering field failure data over the years.

The systems are being installed gradually so it could be 1000 installs in year 1, 1500 in year 2, etc... This means that the population is growing steadily and the failures are below 0.5% per year.

I have the install date of each system and I have the failure date of the failed units, in this case I will consider that once a system has failed its non-repairable. The current population is 5000 which are still alive and going, and 200 which have failed.

My questions are:
Will MTTF make sense in this case, or will it be just a confusing metric?
For calculating MTTF, do I take the operational hours of the whole population up to today, or only the population that failed and the amount of failures in that sample?

Thank you in advance for any help.
 

Miner

Forum Moderator
Leader
Admin
I would counsel against using MTTF/MTBF as a reliability metric. It is very misleading and only applies to random failures. If your product has early life or wear out failures, then MTTF/MTBF metrics are no longer valid. The MTTF of a North American male is approximately 900 years. Know anyone that old? Failure Rate and FIT suffer from the same problem as they are merely reciprocals of MTTF/MTBF (FIT is the PPM version of Failure Rate). The easiest metric to understand is the L10 life (i.e., the point at which 10% of the product has failed). You can use whatever percentage is most appropriate for you (e.g., L1, L5, L10, etc.). You can also use the reliability metric R (e.g., R(10,000 hrs) = 0.98, or R(.95) = 100,000 cycles). It is very accurate but takes a little more explanation.

Use the entire population of interest. The units that have not yet failed are called "censored." They do provide useful information and must be included. If you exclude them, your reliability metric will look worse than it actually is. When you analyze this data, recognize that you are dealing with "arbitrary censoring." That is, the product is being put into service at staggered times, not simultaneously. At any point in time the non-failed product will have different run times.
 

Ikigai

Registered
The MTTF of a North American male is approximately 900 years. Know anyone that old?
Exactly, the figure I calculated doesn't make sense and its far greater than any current survival period.

Thank you for the recommendations, I will try to calculate the L10 life and the reliability metric and see how good it fits along with the other metrics.

I am curently monitoring the failure rates for each year based on the number of failures and the surviving population which builds an accurate image. I have also built the first part of the Bathtub curve but at the moment I cannot see any data indicating that any part of the population has reached the end of its useful life, so the other part of the curve should be visible at a later stage.
 

d_addams

Involved In Discussions
Exactly, the figure I calculated doesn't make sense and its far greater than any current survival period.

Thank you for the recommendations, I will try to calculate the L10 life and the reliability metric and see how good it fits along with the other metrics.

I am curently monitoring the failure rates for each year based on the number of failures and the surviving population which builds an accurate image. I have also built the first part of the Bathtub curve but at the moment I cannot see any data indicating that any part of the population has reached the end of its useful life, so the other part of the curve should be visible at a later stage.
Mean time to failure is not the mean time of the failures you have observed, it is the mean time of failure for ALL the units in service. So the claim of 'the calculated figure doesn't make sense as it is far greater than any current survival period' sounds misinformed about what MTTF is. Unless over half of your products have failed already, the estimated MTTF should be much higher than the average time to failure of the units that have already failed.

perhaps check out Accendo Reliability to start on your journey of reliability engineering and life analysis.
 
Top Bottom