Focused Field Studies and Advanced Modeling Improve Reliability

Getting Failure Data

Failure rate information for equipment can be gathered from industrial databases, company estimates, field return data from manufacturers, end-of-life analysis, end-user failure studies, or by means of a Failure Modes, Effects & Diagnostics Analysis (FMEDA).

Industry databases such as OREDA1 offer valuable information on failure rates and failure modes for various pieces of equipment. Depending on the intended use of this data however, care should be taken. This database includes all realistic failures and does not try to distinguish between failures that are result of the equipment or failures that result from site-specific operational issues.

Another common place to find data is in company-specific estimate reports. Many operating companies generate a list of equipment along with the expected failure rate of that equipment. This information is usually based on expert opinion as well as observed performance of the devices in the field. These estimates can provide useful guidance, but they often don’t have a lot of data to support the conclusions.

Manufacturer field return studies have the advantage that they are based on real data but they have one major drawback: It's impossible for the manufacturer to know what happens to the device after it leaves their facility. They can’t know when it's installed, how it's tested, or how many failures are actually detected or reported back. This data is useful, though, because it will give a good indication of the overall systematic strength of the product and identify any common or recurring problems.

Detailed end-user failure studies are an excellent source for this reliability data. They take into account the equipment, testing and actual environmental conditions. The only disadvantage is that they can be labor-intensive to perform and require a well-defined process to ensure meaningful data is collected.

A practical alternative to these techniques is the FMEDA technique, which involves building a detailed model of the device in question. Since it's hard to get device level failure information, the FMEDA technique relies on component failure information, which includes failure rate as well as failure modes and is more readily available. This information is used in the model to build up the final device and generate the highest level failure rates as well as failure modes.

Functional Failure Modes

Regardless of the technique used to get failure rate information, it is important to consider functional failure modes, which help map failures at the component level and failures at the device level or functional failures. Also important to know is if, for example, a spring fails, will it impact process availability or safety reliability. It is most important to know the effect of that spring failure or the functional failure mode of that spring. A good example of a functional failure mode would be minor leakage on a valve seat. Depending on the application, this could be a dangerous failure—or it could have no effect on the safety function. If the application simply required the valve closed and met an ANSI Class III leakage criteria, then that minor leak on the valve seat would have no effect. However, if the application required a tight shutoff valve, a minor leak would be a dangerous failure. The ability to trace the failure rates and the failure modes from the component level up to the device level and properly assign them to a functional failure mode is a critical step in developing a robust availability and safety reliability model.

Application stress is another key factor in overall reliability. With the exception of the end-user data study, all of the techniques described are going to provide average or homogenized failure rates of failure modes over all industries. At this point it is very difficult to factor in specific application stresses. However, when information is available or when engineering judgment dictates the adjustments be made to the base data, this should be documented. Similarly, different plants will have different capabilities when it comes to maintaining their equipment. So just like one application may induce more stress, plant procedures may be more or less likely to detect and correctly repair problems in the system. This leads to failure rates varying between sites due to different process capabilities and, in one industry study, similar devices in similar applications showed as much as a four times the difference in demonstrated failure rate.

A Hybrid Approach

These rates should also be categorized by functional failure modes, allowing the user of the device to categorize the failures along safety and availability criteria. In addition, if typical application stress and site capability factors are documented, then these base failure rates can be adjusted accordingly. This would have the benefit of providing some scaling to account for known issues.

However, at this point, it's still an estimate. An important final step will be to collect actual failure information from the plants in question and to go back and use this data to validate the original assumptions or make adjustments.

Conclusions

The source of the method used to generate data is necessary to make informed decisions on how to use that data. The intended application of the data should impact the method and rigor used to collect the data and increased adherence to standards such as IEC 61511 will make more data available. As larger sets of data become well documented and published, data sets from single plants can be compared against that data and checked for validity, which will speed up the rate at which application-specific data can be generated.

Chris O’Brien is the author of “Final Elements” and the IEC 61508 and IEC 61511 Functional Safety Standards. He is a partner at Exida Consulting and may be reached at cobrien@exida.com.

1. OREDA is a project organization sponsored by eight oil and gas companies. According to the organization’s website, its main purpose is to collect and exchange reliability data among the participating companies and act as the forum for co-ordination and management of reliability data collection within the oil and gas industry.