ML24075A190
| ML24075A190 | |
| Person / Time | |
|---|---|
| Issue date: | 03/31/2024 |
| From: | Carol Nove NRC/RES/DE |
| To: | |
| Carol Nove 3014152217 | |
| References | |
| Download: ML24075A190 (18) | |
Text
Nondestructive Examination Research at NRC Carol Nove Senior Materials Engineer Office of Nuclear Regulatory Research Division of Engineering Materials Engineering Branch 29th WGIAGE Metal Sub-Group Meeting March 2024 This presentation was prepared as an account of work conducted by an agency of the U.S. Government. Neither the U.S. Government nor any agency thereof, nor any of their employees, makes any warranty, expressed or implied, or assumes any legal liability or responsibility for any third partys use, or the results of such use, of any information, apparatus, product, or process disclosed in this report, or represents that its use by such third party would not infringe privately owned rights. The views expressed in this paper are not necessarily those of the U.S.
Nuclear Regulatory Commission.
Recently Completed Tasks 2
Ultrasound Modeling & Simulation Assessment of commercially available modeling and simulation tools used for ultrasonic testing, including CIVA, UltraVision, and OnScale using a multi-faceted approach that included:
Modeling challenging inspection scenarios that are the subject of relief requests Validating simulation results with empirical data Developing key takeaways and guidance for developing models, executing simulations, and evaluating results 3
PNNL-28362 Ultrasound Modeling and Simulation:
Status Update December 2018 Gerges Dib Richard E. Jacob Michael R. Larche Pradeep Ramuhalli Matthew S. Prowant Aaron A. Diaz ML17095A969 ML18331A254 ML20122A252 ML22087A127 ML22311A009 ML23005A295
Incomplete Coverage Assessment 4
- Develop methods for ascertaining the largest flaw that may be missed given a specified beam coverage
- Assess impact of materials such as austenitic and cast austenitic stainless steels in limited coverage examinations ML23284A269 ML20248H555 ML17318A118
Current NDE Program Focus 5
Dual approach:
Evaluating machine learning (ML) for Ultrasonic Examinations (UT) - Oak Ridge National Laboratory (ORNL)
Evaluate commercially available automated data analysis platforms including rule-based and ML-based systems - Pacific Northwest National Laboratory (PNNL)
Objectives:
Assess current capabilities of ADA for UT NDE applications Provide technical basis to support regulatory decisions and Code actions related to ADA for NDE 6
Automated Data Analysis (ADA)
Flaw inside the weldment Longitudinal 45 Shear 45 Variation in Data (probe/mode)
Rule-Based ADA: Empirical Evaluation of Commercial ADA Systems Data analysis with two different commercial ADA software packages compared to analysis by qualified Level III UT analyst Statistical analysis of results using established methodologies Rule-based ADA is likely not fit for nuclear pipe inspections on its own Rule-based ADA could potentially be used alongside learning-based methods depending on the use-case 7
Assessment of Commercially-available Algorithms/Systems
- Technical Letter Report entitled Evaluation of Commercial Rule-Based Assisted Data Analysis in the review cycle
- Confirmatory analysis of the commercial ML system being tested by industry in field trials has recently begun
- Focus on upper head examinations
- Mockups being designed and fabricated
- Assessment will include:
- Pre-trained algorithm tested with vendor collected UT data on NRC-owned mockups
Machine Learning (ML) for UT NDE ML, if used with care, can be used for NDE data classification Capable of high true positive (TPR), low false positive (FPR) and false negative metrics (FNR)
May be able to learn key signatures using data from simple flaws (e.g., saw cuts) and generalize well to other flaw types (e.g. TFC)
Transfer learning techniques (retraining) may be useful for improving accuracy with new data sets Training data should be representative of the types of data expected during testing High accuracy possible if test data is in distribution relative to training data 9
ML for UT NDE H. Sun, P. Ramuhalli, and R. Jacob, Machine Learning for Ultrasonic Nondestructive Examination of Welding Defects: A Systematic Review, Ultrasonics, Vol. 127 Issue 1, Jan 2023, Pages 106854 (ML22284A071)
H. Sun, R. Jacob, and P. Ramuhalli, Classification of Ultrasonic B-Scan Images from Welding Defects Using a Convolutional Neural Network, Proc. 13th NPIC&HMIT 2023, Pages 272 - 281. ISBN 978-0-89448-791-0 (ML23241A961) 10
- Upcoming activities:
- Comparison of ML results with manual analysis performed by a qualified analyst
- Detection and sizing of degradation that the ML system has not been trained on
- Qualification of ML
- Methods for establishing confidence in ML results ML24046A150
Objective:
Evaluate the capabilities and limitations of two advanced PAUT techniques on nuclear power plant piping welds.
- Full Matrix Capture (FMC)
- Plane Wave Imaging (PWI)
Total focusing method (TFM) processing Assessment via mini round-robin-study (RRS)
- 3 inspection vendors
- Five specimens with six challenging flaws Results from mini RRS compared to data collected with standard PAUT 11 DMW data acquired at 2.25 MHz from the CS side.
Evaluation of Advanced PA Techniques
Key Takeaways to date:
On WSS and DM welds, standard PA-UT outperformed both FMC and PWI (statistically significant result)
PWI consistently had lower SNR than the FMC and standard PAUT making flaw detection more challenging FMC data, where it detected flaws had lower noise floor than standard PAUT on same flaws FMC showed promise for overall improved detectability when the UT energy could penetrate the material microstructure Next Steps:
Focus on dramatically expanding the specimen set but reducing other variables to manage data volumes, acquisition time, and analysis time Improve statistical analysis and POD results by increasing flaw set Evaluate additional commercial advanced PAUT techniques 12 Evaluation of Advanced PA Techniques ML23216A009
Evaluation of Simulated and Digitally Modified Flaws RES assessing current capabilities of industry-developed software tool for modifying volumetric ultrasonic data focusing on the virtual and synthetic flaws being used to create mockups for training and testing purposes with the following objectives:
Validate that digitally modified flaws retain the response characteristics of the original flaw but appear as a separate, standalone flaw Validate the characteristics of simulated flaws compared to real flaw data Assess ASME Code acceptability of digitally modified and simulated flaws 13 Original File & Flaw Location Modified File & Flaw Location
Assessment of NDE for Carbon Fiber Reinforced Plastic (CFRP) Repairs 14 Objective: Evaluate the capabilities and limitations of NDE methods for examining CFRP repairs in NPPs Identify guidance and best practices for qualification mockup fabrication Assess various commercially available NDE methods to evaluate capabilities and limitations for detecting and characterizing flaws Assess the reliability of tap testing vs. other, more sophisticated techniques to provide a permanent record
NDE of Advanced Manufacturing Technologies (AMT) Components Objective: Perform confirmatory testing on relevant AMT materials to understand which NDE methods and techniques will be effective for inservice inspection.
Scope: Evaluation of NDE methods and techniques on a variety of relevant AMT samples and mockups.
- Evaluate inspectability (e.g., UT penetration, attenuation, scatter, and frequency response)
- Evaluate flaw detectability
- Determine capabilities and limitations of different NDE approaches Work informed by a 2020 literature search that identified 3 key questions related to pre-service and inservice NDE:
- 1. What flaws need to be found?
- 2. What sections of the component volume should be inspected?
- 3. How will the critical flaws be identified?
15 ML20349A012
NDE of AMT Components PNNL examined four EBW plates using UT and radiography
- One plate was manufactured with ideal weld parameters
- Three plates had intentionally poor weld parameters resulting in lack of fusion defects at different depths and locations Radiography was not effective at identifying the defects Multiple ultrasonics methods were tested: FMC, TOFD, Tamden and Pulse-echo TOFD was the most effective method for detection and characterization
- TOFD showed starkly different signal intensities from unbonded versus bonded regions, whereas there was virtually no difference between such regions using FMC
- Depth sizing and length sizing of the weld defects was possible with TOFD 16 180 mm 90 mm
NDE of AMT Components Next Steps:
Metallography and heat-treatment of a portion of one plate NDE of mockups made using wire DED (directed energy deposition),
also known as WAAM (wire arc additive manufacturing) 17 From ML23324A242
18