Are you searching for MTBF?

I would ask why would you do that, yet I probably know.

You are looking for reliability information about a component or system. You want to know something about the expected failure rate or durability. Will it last long enough to meet your design and customer requirements?

Or, you have heard of MTBF and want to understand the acronym and metric. Maybe how to calculate the value from test results or field data.

What ever the reason, let’s get something clear. MTBF is not reliability. MTBF is the inverse of the failure rate, it is an average of times to failure, and it is rarely useful.

Reliability and MTBF

Reliability is the probability of successful operation over a specific duration within a specific environment.

There are four elements of a reliability statement.

• Function
• Environment
• Probability
• Duration

MTBF is only one way to state the probability. Along with the reliability function of the exponential distribution (assuming no other time to failure type information is available) we can estimate the probability of success given MTBF.

If I want to know how many units will fail in a month or year, again using the exponential distribution we can estimate these values. As you would expect we would expect different values over different time frames.

When searching for MTBF, you are searching for only one of the four elements needed to understand reliability.

Why limit your self?

After finding MTBF values?

Once you are successful with your search, then what? Are you done?

Got Reliability?

I would suggest asking a few more questions from the source of the MTBF value.

• How was MTBF determined?
• What failure mechanisms are expected?
• If tested, do the failures occur from the failure mechanisms expected?
• What other failure mechanisms and associated stresses are important to consider?
• What function(s) were monitored to determine failure? (Soft failures, hard failures, etc.)
• What environment was used or controlled for the MTBF estimate?
• If field data, does this include no trouble found returns, if not, why not?

And, my new favorite question:

• What are the failure dates? (Better than asking only for failure rates)

Failure dates permit you to estimate the changing nature of the failure rates over time and/or stress. A grand average is not nearly as informative as the time to failure information. Does the item wear out, have early life failures, or really have a constant chance of failure per hour? Without the details about how the testing is done and the detailed results it is impossible to determine from just MTBF alone.

If you are looking for reliability related information, ask for reliability information, not the surrogate MTBF. You will be happier with the information and the results you achieve.