MTBF Logic

The reason so many use MTBF is because so many use MTBF. ‘Our data sheet has to include MTBF since all the other data sheets have MTBF’. Which seems to be primary reason MTBF is so common. It’s because it is so common.

Against this logic is the desire I have to use a measure of reliability that actually is understood. Using reliability (probability of success over a specified duration) as a measure seems some how odd or novel. It is easy to understand and it doesn’t obscure the reliability.

Doing what everyone else is doing and what customers are asking for is compelling. Using MTBF is easy, it is only one number, and with a few assumptions we don’t have to make any duration claims. The idea that we have to use MTBF since customers are requesting it is valid. Equally valid is to provide enough information that customers can truly understand any reliability claims or expectatIons.

When someone asks for MTBF, they are often asking for how long will a product last (a duration question) coupled with the assumption that there is a relatively low chance of failure over that time period. Saying something is ‘reliable’ generally means there is a low chance of failure over some period of time. By stating the probability of success and the duration, say 95% chance of surviving 5 years, is a very direct way to answer the query.

Instead, saying 97.5 years MTBF to answer the same question. 0.95 = exp [ – 5 / 97.5 ] Of course, 97.5 years sounds like a long time and it conveys that product is very reliable. Although it means that there is a 95% chance of surviving 5 years.

So, instead of stating 97.5 years MTBF for a five year product, instead say the product is 95% reliable at 5 years (or there is a 95% chance of the product surviving for 5 years). Better would be to include the entire life distribution and likely failure mechanisms. Yet, no one does that and we certainly don’t want to be unlike everyone else.

How do you request or specify reliability information? If you’re using MTBF, why?

Author: Fred Schenkelberg

I am an experienced reliability engineering and management consultant with my firm FMS Reliability. My passion is working with teams to create cost-effective reliability programs that solve problems, create durable and reliable products, increase customer satisfaction, and reduce warranty costs.

5 thoughts on “MTBF Logic”

  1. Fred,
    I have been arguing against the use of MTBF for years but have just now come across this website. I wholeheartedly agree that MTBF is so over used and mis-used by owner/operators of equipment (manufacturers┬┤ maybe have reason to use MTBF). A simple explanation……..if one examines the “bath tub curve, the majority of equipment life is during the constant failure probability, flat section where failures occur randomly. Random, by definition, states that failure probability is NOT a function of time. A metric based upon time (i.e. MTBF) is therefore invalid for a variable that is not a function of time.

    1. Hi Barry,

      Glad you found and enjoy the site. Keep in mind that the flat part is a myth as most products or only very, very rare products really do have a random and constant failure rate. The exponential distribution parameter, theta or mtbf, described the failure rate per unit time (choose your favorite units). It is just the chance of failure per unit – for example, a 50k hour mtbf means there is a one in 50k of a failure in any hour – first, last, any hour.

      A simple conversion would allow per year, month or whatever unit you’d like, just changing the odds of failure per unit. It also assumes that the failures have no relationship to how long or under what conditions it has been operated. Which is rarely true for the vast majority of failure mechanisms.

      cheers,

      Fred

      1. O contrare Fred. In the process and refining industry, rotating equipment and other dynamic equipment (not static) do fail in a random fashion. Why is this so? Because most failures, 75% to 95%, are caused by events that are indeed random. Do not doubt the flat part of the bath tub. Hundred of studies performed on real equipment validates the shape.

  2. Hi Barry,

    No argument if you have the data and it fits an exponential – go for it. And, rotary equipment makes me think that wear out plays a role at some point, right. Impellers, bearings, valves, etc all wear out. If they are replaced well before the onset of any significant wear than you may well have a nice flat part of the curve.

    When specifying reliability for that equipment, MTBF is still not the way to go. Specify reliability (probability of success) and duration as a minimum. If you and team insist on using MTBF, include the duration with it, please. See what kind of discussion that generates and permits you to reduce the mis understanding of MTBF in general.

    cheers,

    Fred

  3. It may be surprising but rotating equipment, operating in oil and gas facilities; such as refineries, pipeline stations, production sites, transportation sites, and distribution sites; rarely fail due to wearout of components. Failures are random, failures follow events, and events overwhelmingly arise from two root causes: 1) equipment is operated outside the design limits, and 2) human interaction with the process or equipment is performed incorrectly. MTBF cannot be used to measure anything useful in improving the reliability of equipment in the oil and gas industry.

Leave a Reply

Your email address will not be published. Required fields are marked *