Your Reliability Engineering Professional Development Site
One role reliability engineers enjoy is that of analyzing data. We have available to us a wide range of statistical methods, yet each does require some form of data. In some cases, we are data rich with almost too much data. In other situations, we are lacking data or have insufficient data.
[Read more…]
This is the fourth in our webinar series, ‘Reliability Analysis … now what?’ We have examined how to turn data points (like the different failure points of devices under test) into something simple, useful, and sometimes beautiful! This is something that commercial software can’t do.
[Read more…]
This is the third in our webinar series ‘Reliability Analysis … now what?’ The first webinar showed what it means to turn reliability data into something a decision-maker can use. Not a chart. Not a graph. Not a ‘p’ value. Not anything else that you find in a textbook but isn’t useful. Last month’s webinar talked about how the best way to do this is to use computers … but in an easy way! And we talked about creating a ‘posse’ of possible ‘ways’ to explain the data we see. And this webinar is about how we use this thing called Markov Chain Monte Carlo Simulation (MCMC) to create this ‘posse.’ And the ‘posse’ isn’t random … it is based on the ‘likelihood’ of each one being the correct one, which is how MCMC works.
[Read more…]
Dick Moss, retired HP Corporate Quality and Reliability Manager, once mentioned there are only five equations necessary to be a good reliability engineer. He quickly listed four equations and, with a smile, said the fifth one is the one you need to solve the current problem.
[Read more…]
Some of you may have heard of ‘Bayesian analysis.’ You may think this is something fancy that only universities do. Or something we never do because it is all about ‘best guesses.’ Well … neither of these things are true.
Why would you be interested in Bayesian analysis? Simple. It allows you to use EVERY PIECE OF INFORMATION you have. This can be expert judgment. Or it can be test data on a similar ‘thing.’ Or it could be combining test data for one part of your ‘thing’ combined with test data for ‘all of your thing.’
[Read more…]
In 1942, the US Air Force needed to figure out how to get more bombers home. They were losing them at an atrocious rate. They made a study of where the returning bombers were being hit. It was a great data set. But they should have done better in interpreting that data. Abraham Wald, a mathematician, corrected their assumptions and created the “Survivorship Bias principle.” This principle has changed the way we look at data sets and trends. It is a concept that I apply to field failure data sets regularly—more than once. It has caught evidence of core issues that would have otherwise been missed.
[Read more…]
Many companies haven’t the resources to purchase much more expensive software specifically for reliability, but many have the quality software Minitab. Minitab has many reliability functions available; this presentation covers the basics, beginning with “individual distribution identification” to ensure your data fits one of the reliability models. Data is then analyzed with parametric distributions (right censoring vs arbitrary censoring). Additionally, the accelerated life testing functions are explored, focusing on interpreting the results. This is presented with practical examples for clarity.
[Read more…]
If you are a reliability engineer – chances are you have done a Weibull plot. You take something like failure data, put it into a piece of software, and presto!
[Read more…]
Once a product is launched, it’s time to understand its performance. The design is done, and the assembly process is working. Now we can focus on answering the question: is the product hitting reliability targets or not?
Thus, we do field data analysis.
[Read more…]
The Weibull distribution is a versatile tool to analyze time-to-failure data. Like any tool, it could be wielded well or not so well. In this short webinar, let’s discuss when and why you would use Weibull analysis, the basic process, and its limitations.
[Read more…]We have data. Often, an abundance of data concerning equipment failures. Failures per month or MTBF-type measures do not reveal sufficient insights to understand the pattern of failures.
[Read more…]
We often have more data than we can use. For time to failure, warranty claims, and related field data, we should take the time to learn what we can. Let’s explore organizing your field data and use basic techniques let the data speak.
[Read more…]