What Makes A System Ideal For MTBF Prediction?

Knowing the average time between failures allows organizations to address numerous maintenance, repair, replacement, and upgrade needs. MTBF prediction processes, however, are suited to certain types of products, processes, and systems. How do you know whether you should use MTBF prediction and calculation in a particular application? Here are 5 traits that tend to make a system an ideal candidate.

Repairability

Generally, an MTBF calculation is aimed at something repairable. An IT business might have a server farm that includes many blades, or an oil and gas transport company could have a pipeline. The IT firm likely has processes in place for removing server blades and fixing them, just as the oil company can shut down the pipeline, remove sections, and replace them before restoring flow. Some processes and systems can also work within this model as long as you can halt them and address any issues.

Logging

Every MTBF calculation depends on the logging of data. If a component fails, you need a way to log the incident, date and time, and nature. Depending on how many different kinds of potential failures can happen, you may need to be able to code the incidents so you can quickly sort them out using computerized search methods.

Predictability

Failures can't be purely random in this scenario. Likewise, non-random errors have to occur within predictable patterns. These don't need to be easily predictable since you probably wouldn't need MTBF prediction and calculation tools at that point. However, there should be at least a sense that you could figure it out if you had sufficient data and computing power.

Frequency

Notably, failures also need to be common enough that you can produce a statistically significant data sample. If it takes years between failures, you could probably use simulation systems to make an educated guess. However, that sort of analysis generally isn't within the domain of MTBF prediction. A company that sees a couple of hundred incidents across multiple machines will have much better data than a company that only logs a few incidents a year on a single system.

Consistency

The analysis of any product or system should only address things that are identical or extremely close in consistency. To this extent, there are differences between data points and those differences need to fall on a regular curve. Age, for example, is usually an acceptable difference because most systems have consistent aging curves. Size differences, on the other hand, could introduce inconsistencies in performance and wear that make it impossible to compare data between two systems.

To learn more about MRBF calculation, reach out to a local service, such as Rel Teck.