We are a small manufacturing company that makes customer specific equipment for emergency systems for ships. The equipment, once commissioned is normally operational 24/7. We ship products to customers and often assist in starting the equipment up after installation. After warranty periods, we may or may not hear about a problem with our equipment. We often get requests from our customers to provide MTBF information. My question is how can I calculate MTBF when I’m not always sure when the equipment is put in service and when there is a problem with it. Also, we have approximately 50 systems in service at this time, each with various startup dates and each with different sub-components.
Great question. First, we almost never have all the data. If you have information on systems that have failed during warranty, then limit the time you use to the warranty period. So, if the warranty is 1 year and you have not heard about a failure, then that unit has run for one year and hasn’t failed.
Instead of MTBF directly, which isn’t very useful, instead use the data you have and fit a Weibull Distribution. If you’re interested in the expected number of failures over the warranty period this should work very well.
Beyond the warranty period, you could extrapolate the Weibull fit, yet that is risky as other failure mechanisms may reveal themselves after the warranty period and significantly change the results.
Instead, focus on the subsystems and components most likely to fail and use the appropriate model or testing to estimate the post warranty reliability. From there you can calculate the MTBF if necessary for a customer request.
We are a small manufacturing company that makes customer specific equipment for emergency systems for ships. The equipment, once commissioned is normally operational 24/7. We ship products to customers and often assist in starting the equipment up after installation. After warranty periods, we may or may not hear about a problem with our equipment. We often get requests from our customers to provide MTBF information. My question is how can I calculate MTBF when I’m not always sure when the equipment is put in service and when there is a problem with it. Also, we have approximately 50 systems in service at this time, each with various startup dates and each with different sub-components.
HI Scott,
Great question. First, we almost never have all the data. If you have information on systems that have failed during warranty, then limit the time you use to the warranty period. So, if the warranty is 1 year and you have not heard about a failure, then that unit has run for one year and hasn’t failed.
Instead of MTBF directly, which isn’t very useful, instead use the data you have and fit a Weibull Distribution. If you’re interested in the expected number of failures over the warranty period this should work very well.
Beyond the warranty period, you could extrapolate the Weibull fit, yet that is risky as other failure mechanisms may reveal themselves after the warranty period and significantly change the results.
Instead, focus on the subsystems and components most likely to fail and use the appropriate model or testing to estimate the post warranty reliability. From there you can calculate the MTBF if necessary for a customer request.
Cheers,
Fred