Where to Find Failure Data

Finding meaningful and accurate failure rate data is one of the key challenges of SIS engineering.  According to IEC 61511 2nd edition,

The lack of reliability data reflective of the operating environment is a recurrent shortcoming of probabilistic calculations” (11.9.3 note 2).

Ideally, everyone implementing SIS would have a large database of high quality, locally sourced, prior use data.  In reality, engineers often have to draw from a variety of sources of failure rate data to accurately model a system.

The list below is not exhaustive, but it should provide a good starting point toward finding the required reliability data for any SIS design where prior use data is unavailable.  It is a good idea to cross-reference multiple sources to understand the potential uncertainties in the data.  Some more sophisticated data sources will even provide uncertainty estimates with their data.

A Word of Caution

Common issues to look out for include failure rate data that:

  • Is based on manufacturer warranty return data.  These may under-report failures.
  • Is based on high demand or accelerated life testing (e.g. B10 testing). These may be based on different wear-out failure modes than encountered in process applications.
  • Is based on theoretical FMEDA studies.  These studies may assume ideal environmental or process conditions and neglect real-world failure modes (e.g. plugging, corrosion).
  • Is reported as PFD or PFDavg rather than failure frequency. The basis of the PFD calculation (e.g. test interval, test coverage, etc.) needs to be thoroughly understood.
  • Is significantly lower than data for similar devices, especially for SIL certificates. There is a wide range of quality in the certification world.  Strong claims should be backed up with strong data, preferably from real-world application.
  • Omits portions of the device or interface. For example does the valve include the actuator? the solenoid?  Does the transmitter include the impulse lines?
  • Has an unclear or inconsistent definition of failure. For example, I recently looked at Pressure Relief Valve (PRV) failure data, and I found that my four data sources had four different definitions of “failure”.
  • Does not specify or specifies a very short useful life.  A failure rate based on a 5 year useful life is meaningless if the equipment is expected to be in service for 15 years.

I would like to reiterate the last point about useful life.  Useful life has historically been been poorly understood and inconsistently observed in the SIS community.  Maybe I will cover it in detail in a future post.  For now, consider that IEC 61508 states that probabilistic calculations based on constant failure rates are “meaningless” when components are used beyond their useful life.

None of the above issues imply that any of the data sources are incorrect or invalid, but the basis of the failure rate data must be understood to determine if the data is applicable for a particular application.  Even when perfectly applicable data is unavailable, it is often possible to adjust the available data to better reflect the application.

Free Resources

SIL Certificate Databases

SIL certificates are a convenient source of data, but should be view skeptically and cross-referenced with other sources since they may sometimes omit certain components from the analysis or make unrealistic assumptions about operating conditions.

Manufacturer Websites

In some cases, it may be easier to find SIL certificates directly from the manufacturer. I am just listing a few popular vendors who happen to have centralized locations for many of their certificates. Keep in mind that in addition to SIL Certificates, the product Safety Manual will often have more detailed information.  Watch out for manufacturers that don’t have safety manuals; that’s a strong sign they do not understand IEC 61508 / 61511.

Other Web Resources

Miscellaneous other free resources

Nuclear Industry

The nuclear industry has a large volume of failure rate data that is generally applicable to the process industries

Coursera - Hundreds of Specializations and courses in business, computer science, data science, and more

PAID RESOURCES

Unfortunately, not all data is free.  However, some of these resources may be free to you via the Internet, depending on company subscriptions, copyright scruples, etc.

Books

Standards

Software

There are many commercially available SIS software packages that have built-in failure rate data. I am not covering these here, as many of them obtain their data from one or more of the sources listed above.  Software tools and built-in data are convenient, but all of the same caveats apply.  Data must be thoroughly understood to ensure it is used meaningfully.  Beyond data, it is also critical to understand how the software is using the data. Perfect failure data applied to the wrong model is still meaningless!

OTHER GUIDANCE

These resources do not provide failure rate data, but rather provide guidance on how to use and interpret data from other sources.


A common theme among much of the guidance literature is that failure rates are uncertain.  The goal of searching for failure rates should not be to cherry-pick the data until we find the lowest failure rate. We should evaluate all of the available data, estimate the uncertainty, and judge the applicability to our own specific applications and environments.  Later, the data should be updated based on actual experience in the application.  By the way, the Bayesian framework is an excellent way to accomplish this initial analysis and future updating.

I hope this post is helpful to you.  Please add any resources I may have missed in the comments.


3 thoughts on “Where to Find Failure Data

Leave a Reply

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