Do I Really Need to Wait for Failure Data
Many reliability engineers, including me, constantly gripe that they don't have good failure data to work with. It just may be that we are barking up the wrong tree in some cases. It is true that good failure data makes our job much easier, but having that data means we have to wait for failures to occur. Try going to your maintenance manager and telling him/her that you need a few more failures before you can nail down a good maintenance interval for changeout or overhaul. Just make sure to update your resume before you head to the office.
When you think about it, there are relatively few items of equipment that wear out and fail. The Nowlan and Heap team found only about 11% of items in an aircraft had some wear out characteristic. A good portion of the items that do wear out could possibly provide information very early in their lives that would tell us when they might reach the point where they are no longer capable of performing their function. Maybe they just need to be subjected to some degradation analysis. All that is required is that they give some "signal" that is representative of the amount of degradation that is occurring over time.
Measuring things like thickness, tire tread depth, crack length, change in electrical output or required electrical input for a certain output, light output, or any of several bits of information may give us the information to predict potential times to failure. Fit the data to the right type of distribution, set a "soft failure limit", and generate pseudo times to failure for use in statistical distributions and voila, we have an estimated time to failure that can be used for generating optimized maintenance intervals to deal with these items.
Now the conversation becomes "Hey boss, I need some time during our next scheduled down to take some measurements on ?????" or "Hey boss, we found out that we can measure the lumens on this set of lights in order to determine when we will need to change them out to meet Reg XYZ-SAFE."
Tom Jacyszyn, All good comments. A better title for the article might have been do I have to wait for failure data. There are some failures for which failure data may not be available, but degradation data is available. I did not mean to imply that you shouldn't use failure data if you have it. Degradation analysis is another tool in the box just like Weibull, RCM and condition monitoring.
Another helpful tool is a Weibull Analysis and Reliability Centered Maintenance (RCM)to look at data, trends and failures. Use that to anticipate failures or periods where failures most likely won't occur or when to do preventative maintenance, inspections, and scheduled repairs.
Anybody think this could be a bad idea. Barry, not so sure I'd place a bet here or the longevity of the Asset Manager. Lets see, betting, no data (history) needed ...then you are not really managing things, just flying by the seat of your pants. I prefer a baseline, data points & trending, a FMEA/FMECA, a process, and a plan. Fahad is right, Physics of Failure is nice, if you can get the information and you have the resources and time. In the mean time ...maybe in your toolkit a Fault tree Analysis (FTA) for the likelihood for certain failures. Use all the tools you can to anticipate and do maintenance on your timeline vs sudden failures maintenance. Put your knowledge, experience, and planning to good use to control costs and lost production.
Hi Bill , Nice Article. I would just disagree to the point that physics of failure based approach work when we are sure that failure induced or degradation is not impacted by the external factors like maintenance , operations regimes etc. In general indistrial environment , we hardly find such privilage. For such case hard failure data is best to idnetify the root casues. However, I agree that for machines based purely upon degradation analysis are best for such analysis. I like SAS JMP for such analysis. Nice software.