An FMEA (Failure Modes and Effects Analysis) is a large effort.
Here are three (3) simple steps to help improve your FMEA. We add some hints that make this process easier.
With any FMEA, the language we choose and record can make or break the analysis. We all interpret words differently. Without the proper use of words, some failure modes can be misleading when read by others. All is not lost. There are a few simple rules of thumb to follow. This will allow for better FMEA results.
Step 1 - Develop a Consistent Failure Modes Description
Why is it important to be descriptive?
First, others who read your FMEA need to be able to understand it. Follow this simple three-part naming convention. It will help you to develop consistent and repeatable ways to describe each failure.
1. Part causing failure (object) / 2. Failure mode (adjective) / due to / 3. Failure cause (why)
Bearing (object) seized (adj.) due to lack of lubrication (why)
Impellor (object) worn (adj.) due to the ingress of particles (why)
Step 2 - Development of the Failure Characteristics
Not all tasks are the same. In order to select the correct logical decision, we must understand the failure characteristics. If prior failure data or history exists, we can use it. This is typically done using statistical distributions. The most commonly used is called the Weibull distribution.
Throughout many years of experience in developing scheduled maintenance programs, we have learned how more efficient programs can be developed. This is done through the use of logical decision processes. These logical decision processes are actually decision trees. We use these trees to select tasks. These tasks comprise the maintenance program. A software tool such as the FMEA module of Reliability Workbench™ makes this process much easier.
When creating a Weibull model, two technical parameters are required. The Eta or characteristic life and the Beta shape factor are needed. These two parameters tell us how the failure mode will behave. With an applicable maintenance task, we hope to manage to an acceptable level of risk.
Consider as a simple example compare the failure modes below.
- Bearing failure due to end of life/wear
- Bearing seized due to lack of lubrication (after initial lubrication at installation)
The estimate Eta (characteristic life) value of a well-maintained bearing should be 10 yrs. Often bearing L10 life is rated in revolutions of the bearing, at a specified load. The second failure mode the scenario is different. Here we will assume the bearing was correctly installed and lubricated. Subsequently, it was not lubricated as specified by the manufacturer. Hence, it fails in 1 yr due to the lack of lubrication. Review of the Beta (shape factor) value is different for each failure mode (end of life vs. lack of lubrication). The Beta value should near or equal 1 for a well lubricated, and incorrectly loaded bearing. Such a bearing will achieve its stated L10 life.
The second failure mode description is well detailed. This makes it easy to understand. The first adjective we find is ‘seized’. The final part of the statement is ‘lack of lubrication’. This then describes why the failure mode occurred. Because of a lubrication issue, the wear out of the bearing is faster than the L10 life. This is due to the lubrication reaching its end of useful life. So we use a Beta value of 4.
|Bearing failure||10 years||1|
|Bearing seized due to lack of lubrication||1 year||4|
Step 3 - Determining the Applicable Maintenance Task
A maintenance task is said to be applicable if, the task is capable of improving on the reliability that would exist if the task was not performed.
There are 4 basic scheduled maintenance task types we can choose from:
- Scheduled inspection of an item at regular intervals to find potential failures.
- Scheduled rework of an item at or before some specified age limit.
- Scheduled discard of an item(or one of its parts) at or before some specified life limit.
- Scheduled inspection of a hidden-function to find functional failures.
When determining maintenance activities, the beta parameter plays a very important role. Beta helps determine which types of tasks might be applicable. We use different Beta for different failure modes. We don’t have to do this the hard way. Learn more about the easy way (Availability Workbench™ software)
The beta value of 4 suggests that the end of life is predictable. In fact, it follows a typical wear out or end-of-life behavior. In this scenario, the end of life of the bearing lubricant. A bearing with old lubrication causes the bearing to seize. The applicable task would thus be a Type 2 task. You would select to rework the bearing lubrication at or before its specified life limit.
A beta parameter of 1 suggests that the failure is random. That means the component has the same probability of failure the first day it is installed as compared with; say the 1000th day. The failure is not predictable and a time-based PM is not applicable. In these cases, choose a Type 1 task. Inspect the bearing at regular intervals to find potential failures. Typically this is performed using a Vibration Analysis.
Maintenance decisions that do not affect safety should be based on a cost-benefit analysis. If however, safety is affected then we must do more. If an acceptable maintenance task cannot be found to reduce the risk to an acceptable level, then the only way to protect from failure would be to redesign the system. With this redesign, we seek to remove that specific failure mode.
In summary, the use of language in FMEA to describe failure modes is very important, without datasets to support the failure mode estimating parameters can alter the outcomes of FMEA. It is important to consider the component (what failed), the mode of failure (an adjective) and the cause of the failure (the why) to gain real benefits from an FMEA, otherwise outcomes can be compromised and typically FMEA will not improve the equipment performance as intended.