Bayesian Methods for NDE Data Fusion Become a Professional – Officially! Strategies and Advice for Applying for, Preparing for, and Passing the Principles and Practice of Engineering (PE) Exam
Download as PPTX, PDF0 likes424 views
This document discusses using Bayesian methods for fusing data from non-destructive evaluation (NDE) techniques. Bayesian methods allow incorporating measurement uncertainty into predictive digital twin models by applying Bayes' theorem to update predictions based on new measurements over an asset's lifetime. This helps enable more accurate predictions of when and how failures may occur, informing better engineering designs.
Bayesian Methods for NDE Data Fusion Become a Professional – Officially! Strategies and Advice for Applying for, Preparing for, and Passing the Principles and Practice of Engineering (PE) Exam
- 2. What is NDE? NonDestructive Evaluation Don’t Break It!
- 7. How do you fuse such diverse data? Provide Spatial Awareness
- 8. What is a digital twin? Image from The Matrix 2004 DVD extras
- 9. Why do we need to model Uncertainty?
- 10. Bayes’ Theorem P(A | B) = P(B | A)P(A) P(B) P(state | measurment) = P(measurement | state)P(state) P(measurement) Prior Measuremen t Accuracy Normalizing Term 1. Predict 2. Measure 3. Update
- 11. Where do the Bayesian Methods come in? Time CAD Model As Built In Service Indication Back to Service New/Change in Indications
- 12. How does this help anything? Prediction, Prediction, Prediction When will it break? How will it break? Better Design Allow Engineers to know how their designs actually fair throughout its lifetime
Editor's Notes
- #3: The Answer is: A whole bunch of very different things. X-Ray, Ultra Sonic Transmission, Vibro Thermography, Fluorescent Penetrant, Eddie Current, Computed Tomography (CT)
- #8: The Data has to having some element that makes it possible to integrate it. That Element is the the actual physical specimen. Part of what we are doing is adding spatial awareness to all the. The great thing about this diverse data is that we can know a lot about an object. And what we really want to do is make a Digital Twin!
- #9: Nope, not quite. A Digital would be a model of an actual part, or machine. Right now we have a lot of models of things, CAD Models, FEA models but those are usually models of an idealize generic part. Anyone who actually makes things know there is a difference between what you planned to make and what you actually make. The Digital twin uses that original model of what you intended to make and NDE data to give a better representation of an actual part in use. Example: I have an wind turbine in service…. So what is the problem…. Uncertainty!!
- #10: Because it exists! We get things wrong, we perceive things incorrectly, and measurements are faulty
- #12: We Use Bayesian Methods to track the best estimate and the uncertainty as time goes by and new inspection data becomes available
- #13: The next step