ML22348A127

PNNL-33733, Pionic NDE Modeling and Simulation Exercise

ML22348A127
Person / Time
Issue date:	12/31/2022
From:	Jacob R, Meyer R, Carol Nove NRC/RES/DE, Pacific Northwest National Laboratory
To:
Carol Nove 301-415-2217
Shared Package
ML22348A126	List: ML22348A127 ML22348A128
References
PNNL-33733
Download: ML22348A127 (106)

Text

Choose an item.

PNNL-33733 PIONIC NDE Modeling and Simulation Exercise December 2022 Ryan M. Meyer Richard E. Jacob Prepared for the U.S. Nuclear Regulatory Commission Office of Nuclear Regulatory Research Under Contract DE-AC05-76RL01830 Interagency Agreement: NRC-HQ-60-17-D-0010

Choose an item.

DISCLAIMER This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor Battelle Memorial Institute, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof, or Battelle Memorial Institute. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.

PACIFIC NORTHWEST NATIONAL LABORATORY operated by BATTELLE for the UNITED STATES DEPARTMENT OF ENERGY under Contract DE-AC05-76RL01830 Printed in the United States of America Available to DOE and DOE contractors from the Office of Scientific and Technical Information, P.O. Box 62, Oak Ridge, TN 37831-0062; ph: (865) 576-8401 fax: (865) 576-5728 email: reports@adonis.osti.gov Available to the public from the National Technical Information Service 5301 Shawnee Rd., Alexandria, VA 22312 ph: (800) 553-NTIS (6847) email: orders@ntis.gov <https://www.ntis.gov/about>

Online ordering: http://www.ntis.gov

PNNL-33733 PIONIC NDE Modeling and Simulation Exercise December 2022 Ryan M. Meyer Richard E. Jacob Prepared for the U.S. Nuclear Regulatory Commission Office of Nuclear Regulatory Research Under Contract DE-AC05-76RL01830 Interagency Agreement: NRC-HQ-60-17-D-0010 Carol A. Nove, Contracting Officer Representative Pacific Northwest National Laboratory Richland, Washington 99354

PNNL-33733 Acknowledgments ii Acknowledgments We wish to acknowledge the following organizations for providing CIVA and/or finite element modeling resources and expertise to this activity:

• Japan Power Engineering and Inspection Corporation

• Mitsubishi Heavy Industries

• Swiss Association for Technical Inspections, Nuclear Inspectorate

• VTT Technical Research Center of Finland.

PNNL-33733 Summary ii Summary This document summarizes collaborative nondestructive examination (NDE) modeling and simulation exercises performed by participants of the Program for Investigation of NDE by International Collaboration (PIONIC). PIONIC is an ongoing international research collaboration for topics related to NDE in commercial nuclear power plants with participants from the United States, Japan, South Korea, Finland, Sweden, and Switzerland. The main objective of this effort was to utilize the collective capabilities of PIONIC participants to evaluate consistency of NDE modeling and simulation tools and their sensitivity to user variation. In essence, the primary goals of this effort were to answer the following two questions:

• What degree of similarity can be expected for results generated by multiple independent teams implementing the same examination scenario with the same tool?

• How sensitive are results to different techniques for modeling anisotropy in welds?

The outcome of this effort can serve industry and regulators by providing information on the level of detail required for the creation of standard NDE modeling and simulation procedures and for effective regulatory reviews of results from NDE modeling and simulation. The primary NDE modeling and simulation tool used in these efforts was Extende CIVA; however, some teams also contributed finite element modeling (FEM) capabilities. Efforts initially focused on the implementation of a phased array ultrasonic testing (PAUT) procedure used in the open testing of the Program to Assess the Reliability of Emerging Nondestructive Technologies (PARENT)

(Meyer et al. 2017). PARENT, a predecessor to PIONIC, focused on generating empirical probability of detection data for dissimilar metal welds. Thus, the initial selection of the PAUT procedure and test block for these simulation exercises were motivated by the desire to leverage resources already developed under PARENT.

The activity with the PAUT procedure was primarily divided into two exercises: 1) a probe calibration exercise with a test block model containing no weld and two side-drilled hole (SDH) reflectors, and 2) a weld anisotropy exercise in which participants developed models of anisotropic weld regions and simulated responses from planar reflectors within those welds.

Following the activity with the PAUT procedure, the decision was made to conduct a second activity with a simplified conventional ultrasonic testing (UT) procedure. This activity was also split into the same two main exercises: 1) a probe calibration exercise, and 2) a weld anisotropy exercise.

Overall, these efforts highlighted how user variance can affect NDE modeling and simulation results. Potential user variance is an aspect that should be addressed when technical evaluations rely on the outputs of NDE modeling and simulation tools. More specifically, the following conclusions were drawn from these efforts:

1) All model input parameters should initially be considered significant, and an effort should be made to make sure they are all identified for the examination scenario. A systematic and comprehensive approach should then be applied to justify choices of all modeling parameters, including default parameters.

2) The increased difficulty for consistent implementation of a PAUT probe versus a conventional UT probe was notable in these efforts. Advanced techniques (e.g., PAUT) introduce additional input parameters that require additional scrutiny to justify the parameter selections and to make sure the technique is implemented as intended.

PNNL-33733 Summary iii

3) Many users are likely to limit themselves to options for weld anisotropy modeling that are pre-packaged in off-the-shelf tools. Although it is possible to develop custom techniques that can capture grain scattering effects in welds, many users may consider them impractical to implement.

4) Qualitative agreement in results was observed between 3D FEM and CIVA (3D) models for relatively simple test block scenarios involving an isotropic material structure and SDH reflectors; however, precise quantitative agreement was not achieved.

5) Simplified 2D FEM models produced results that were qualitatively different from 3D FEM and CIVA models, even for simple test block scenarios.

6) Although it is possible to create detailed anisotropic welds in 3D FEM tools, they are challenging to implement, and this approach may be considered impractical for many users.

7) Even if quantitative agreement with empirical results cannot be achieved, NDE modeling and simulation tools may still agree qualitatively, which allows them to be useful for such functions as optimizing examination methods.

PNNL-33733 Acronyms and Abbreviations ii Acronyms and Abbreviations CAD computer-aided design EBSD electron backscatter diffraction FB flat bar FEM finite element modeling ID inner diameter IPF inverse pole figure JAPEIC Japan Power Engineering and Inspection Corporation LW longitudinal waves MHI Mitsubishi Heavy Industries NDE nondestructive evaluation OD outer diameter PARENT Program for Assessing the Reliability of Emerging Nondestructive Technologies PAUT phased array ultrasonic testing PIONIC Program for Investigation of NDE by International Collaboration PNNL Pacific Northwest National Laboratory SDH side-drilled hole SVTI Swiss Association for Technical Inspections, Nuclear Inspectorate SW shear waves TRL transmit-receive longitudinal UT ultrasonic testing WSS wrought stainless steel

PNNL-33733 Contents iii Contents Acknowledgments......................................................................................................................... ii Summary....................................................................................................................................... ii Acronyms and Abbreviations........................................................................................................ ii Contents........................................................................................................................................ iii 1.0 Introduction....................................................................................................................... 1 2.0 Case Study Description..................................................................................................... 3 2.1 Procedure Case Study.......................................................................................... 3 2.2 Probe Specifications.............................................................................................. 3 2.3 Physical Test Block Specifications........................................................................ 4 3.0 Probe Calibration Exercise................................................................................................ 6 3.1 CIVA Interpreted Probe Model Response Summary............................................. 6 3.2 Consistent CIVA Probe and Flaw Models............................................................. 9 3.3 2D and 3D Finite Element Modeling Results (JAPEIC)....................................... 12 4.0 Anisotropic Weld Model Implementation......................................................................... 19 4.1 PNNL CIVA Implementation................................................................................ 19 4.2 Ogilvy Weld Model.............................................................................................. 24 4.3 MHI CIVA Implementation................................................................................... 25 4.4 Team 114 Response Data from Specimen P42 in PARENT............................... 27 4.4.1 Discussion of LW Responses............................................................... 27 4.4.2 Discussion of SW Responses............................................................... 29 5.0 Conventional Probe Implementation............................................................................... 32 5.1 Results of Conventional UT Probe Calibration.................................................... 32 5.2 Results of Conventional UT Probe with Complex Weld Models.......................... 38 5.2.1 PNNL CIVA Implementation................................................................. 38 5.2.2 SVTI CIVA Implementation................................................................... 44 5.2.3 MHI CIVA Implementation.................................................................... 50 6.0 Discussion and Conclusions........................................................................................... 55 6.1 Variation in Approaches...................................................................................... 55 6.2 Modeling Weld Anisotropy................................................................................... 56 6.3 Finite Element Models......................................................................................... 57 6.4 Comparisons with Empirical Data........................................................................ 58 7.0 References...................................................................................................................... 59 Appendix A - Team 114 TRL45 CIVAProbe and Array Settings........................................... A.1 Appendix B - CIVA User InterfacesAn Example.................................................................... B.1 Appendix C - Results from Interpreted Probe Specifications................................................... C.1 Appendix D - Summary of Input Settings for Interpreted Probe Model Activity........................ D.2 Appendix E - Conventional Probe Specifications...................................................................... E.1

PNNL-33733 Contents iv Appendix F - Example UT Modeling and Simulation Standard Template................................. F.1 Figures Figure 2.1 Coordinate System Definition, Dimensions, and Illustrations of FB Test Blocks P28, P29, P30, P31, P32, P42, and P46................................................... 5 Figure 3.1 PNNL Team Results for Interpreted Probe Model Specification........................... 7 Figure 3.2 PNNL Team Results for Consistent Probe Model with LW and for a Linear 45° Scan Presented in the Format of Figure 3-21............................................... 10 Figure 3.3 PNNL Team Results for Consistent Probe Model with LW and for a Linear 60° Scan Presented in the Format of Figure 3-21............................................... 10 Figure 3.4 MHI Team Results for Consistent Probe Model with LW and for a Linear 45° Scan Presented in the Format of Figure 3-21............................................... 11 Figure 3.5 MHI Team Results for Consistent Probe Model with LW and for a Linear 60° Scan Presented in the Format of Figure 3-21............................................... 11 Figure 3.6 Screenshot of Slide Summarizing Defining Parameters of JAPEIC 2D FEM Model for Probe Calibration (Courtesy of JAPEIC)..................................... 13 Figure 3.7 B-Scan Response Images Obtained by JAPEIC 2D FEM Model for A)

Upper SDH at 45°, B) Lower SDH at 45°, C) Upper SDH at 60°, D) Lower SDH at 60° (Courtesy of JAPEIC)....................................................................... 14 Figure 3.8 Depictions of the (Top) 2D and (Bottom) 3D Probe Geometries in the ComWAVE User Interface; the 3D Geometry Allows the Definition of a Dual Transmit-Receive Probe (Courtesy of JAPEIC).......................................... 15 Figure 3.9 Depiction of the Delay Calculation of Wave Paths Constrained to the Central Plane in the JAPEIC 3D FEM Model (Courtesy of JAPEIC)................... 16 Figure 3.10 B-Scan Response Images Obtained by JAPEIC 3D FEM Model for A)

Upper SDH at 45°, B) Lower SDH at 45°, C) Upper SDH at 60°, D) Lower SDH at 60° (Courtesy of JAPEIC)....................................................................... 18 Figure 4.1 (Top) Photograph of Etched and Polished Cross Section of WSS-WSS Austenitic Weld from Specimen 3C-022; (Bottom) IPF Displaying Y-Component of the Crystal Orientations Relative to the Surface of the Weld Cross Section............................................................................................. 20 Figure 4.2 Illustration of the Process of Converting a Downsampled EBSD Image of a Weld Cross Section to CAD Format that Can Be Imported into CIVA.............. 21 Figure 4.3 (Top) B-Scan Response Obtained for 45° LW Scan by PNNL Team with Anisotropic EBSD Weld Model and Notch Reflector and (Bottom) Ray Trace Diagram Depicting the Effect of the Modeled Weld Anisotropy on Beam Scatter and Redirection............................................................................ 22 Figure 4.4 PNNL Team Responses for 45° LW Scan with Anisotropic EBSD Weld Model and Notch Reflector.................................................................................. 23 Figure 4.5 True B-Scan Baseline Response Obtained for 45° LW Scan with Isotropic Weld Properties and Notch Reflector by PNNL Team......................................... 23

PNNL-33733 Contents v

Figure 4.6 Baseline Response Obtained for 45° LW Scan with Isotropic Weld Properties and Notch Reflector by PNNL Team, Including B-Scans and Echodynamic Response Images......................................................................... 24 Figure 4.7 Diagram Depicting the Various Parameters (T, D,,, ) that Determine Variation of Crystallographic Orientation with Position through Eq. 4-1.............. 25 Figure 4.8 Screenshot of CIVA Input Menu for Defining Ogilvy Model................................. 26 Figure 4.9 MHI Team CIVA Responses for Isotropic Weld Model and for Ogilvy Model, with T = 0.3, 0.5, and 0.7......................................................................... 26 Figure 4.10 B-scan Response Images Obtained by Team 114 in PARENT Open Testing from FB Test Block P42.......................................................................... 28 Figure 4.11 B-Scan Response Images Generated in CIVA from a Notch Reflector with Isotropic Weld Properties; Responses Including Only the Primary LW Interactions are Shown on the Left, and Responses Including Mode-Converted SW Signals are Shown on the Right.................................................. 29 Figure 4.12 B-Scan Response Images Obtained from SW Scans on FB Test Block P42 by Team 114 in PARENT Open Testing...................................................... 30 Figure 4.13 Simulated CIVA Responses for 45 SW and 60 LW Scans on Notch Reflector with Isotropic Weld Properties.............................................................. 31 Figure 5.1 Illustration of Conventional UT Probe (See Appendix D) Scan of Two SDH Reflectors in CIVA Model............................................................................ 32 Figure 5.2 PNNL Response for Conventional UT Probe Scan of Two SDH Reflectors Showing B-Scan Images (Top) and Echodynamic Amplitude vs Scan Position (Bottom Left) and Echodynamic Amplitude vs Time (Bottom Middle)................................................................................................................. 33 Figure 5.3 CIVA Simulation Settings Selected by MHI and SVTI in the Conventional UT Probe Calibration Exercise............................................................................ 34 Figure 5.4 B-Scan Response Image (Left) and Echodynamic Amplitude vs Scanning Position Response (Right) for the PNNL Simulation of the Conventional UT Probe Scan of Two SDH Reflectors.............................................................. 35 Figure 5.5 MHI Response for Conventional UT Probe Scan of Two SDH Reflectors Showing B-scan Images (Top Left) and Echodynamic Amplitude vs Scan Position (Top Right) and Echodynamic Amplitude vs Time (Bottom Right)........ 36 Figure 5.6 SVTI Response for Conventional UT Probe Scan of Two SDH Reflectors Showing the Echodynamic Amplitude vs Scan Position Response.................... 36 Figure 5.7 B-Scan Response Image for 3D FEM Simulation of Conventional Probe Response from Two SDH Reflectors by JAPEIC................................................ 37 Figure 5.8 Echodynamic Amplitude vs Time (Left) and Echodynamic Amplitude vs Scan Position (right) for 3D FEM Simulation of Conventional Probe Response from Two SDH Reflectors by JAPEIC................................................ 37 Figure 5.9 (Left) the Cross-Sectional Weld Sample That Was Scanned with a Red Line Outlining the Weld Region, (Middle) an IPF Representation of the EBSD Scan Data, and (Right) the Weld Model after Downsampling and Converting into a Matrix of Square Pixels........................................................... 38 Figure 5.10 CIVA 3D Depiction of the Conventional UT Probe with EBSD Generated Anisotropic Weld Model....................................................................................... 39

PNNL-33733 Contents vi Figure 5.11 PNNL Response for Conventional UT Probe Scan of Two Notch Reflectors Showing a True B-Scan Image (Top Left), a B-scan Image (Top Right), Echodynamic Amplitude vs Scan Position (Bottom Left) and Echodynamic Amplitude vs Time (Bottom Middle).............................................. 40 Figure 5.12 PNNL Response for Conventional UT Probe Scan of a Notch Reflector Located within the Near Side of the Anisotropic Weld Showing a True B-Scan Image (Top Left), a B-Scan Image (Top Right), Echodynamic Amplitude vs Scan Position (Bottom Left) and Echodynamic Amplitude vs Time (Bottom Middle).......................................................................................... 41 Figure 5.13 PNNL Response for Conventional UT Probe Scan of a Notch Reflector Located within the Far Side of the Anisotropic Weld Showing a True B-Scan Image (Top Left), a B-Scan Image (Top Right), Echodynamic Amplitude vs Scan Position (Bottom Left) and Echodynamic Amplitude vs Time (Bottom Middle).......................................................................................... 41 Figure 5.14 PNNL Response for Conventional UT Probe Scan of a Notch Reflector Located within the Far Side of the Anisotropic Weld Showing a True B-Scan Image (Top Left), a B-Scan Image (Top Right), Echodynamic Amplitude vs Scan Position (Bottom Left) and Echodynamic Amplitude vs Time (Bottom Middle).......................................................................................... 42 Figure 5.15 PNNL Skew Angle 90° Response for Conventional UT Probe Scan of a Notch Reflector Located Outside of the Anisotropic Weld, on the Far Side, Showing a True B-Scan Image (Top Left), a B-Scan Image (Top Right),

Echodynamic Amplitude vs Scan Position (Bottom Left) and Echodynamic Amplitude vs Time (Bottom Middle)..................................................................... 43 Figure 5.16 PNNL Skew Angle 270° Response for Conventional UT Probe Scan of a Notch Reflector Located Outside of the Anisotropic Weld, on the Far Side, Showing a True B-Scan Image (Top Left), a B-Scan Image (Top Right),

Echodynamic Amplitude vs Scan Position (Bottom Left) and Echodynamic Amplitude vs Time (Bottom Middle)..................................................................... 43 Figure 5.17 Stiffness Matrix Definition for the Definition of a Weld as a Single Orthotropic Region in CIVA by SVTI and MHI Teams......................................... 44 Figure 5.18 Top: The Examination of an Elliptical Notch Located in the Middle of the Weld Region for the Test Block is Depicted for SVTI; Middle: The True B-Scan Response Image for the Elliptical Notch within the Weld Defined as a Single Orthotropic Region; Bottom: The Echodynamic Amplitude vs Time Response for the Elliptical Notch within the Weld Defined as a Single Orthotropic Region................................................................................... 45 Figure 5.19 Results from SVTI Parametric Analysis of Effect of Values for Elastic Constants C11, C22, and C33 on Relative Tip-to-Corner Response Amplitude and Arrival Time of the Tip Response................................................ 46 Figure 5.20 Top: Depiction of the SVTI Model of an Elliptical Notch Located inside of an Ogilvy Model Defined Weld; Bottom: True B-Scan Response of the Elliptical Notch Located inside of an Ogilvy Model Defined Weld....................... 47 Figure 5.21 Stiffness Matrix Definition Used by SVTI in the Ogilvy Model Weld Definition in CIVA................................................................................................ 48 Figure 5.22 Top: Depiction of the SVTI Model of the Examination of Two Rectangular Notches with One Notch Located outside of the Weld Region and the

PNNL-33733 Contents vii Other Notch Located in the Middle of a Weld Defined as a Single Orthotropic Region; Bottom: True B-Scan Response for the Simulation of Two Rectangular Notches................................................................................... 49 Figure 5.23 Top: Depiction of the MHI Model of an Elliptical Notch Located inside of an Ogilvy Model Defined Weld; Bottom: True B-Scan Response of the Elliptical Notch Located inside of an Ogilvy Model Defined Weld....................... 51 Figure 5.24 Top: Depiction of the MHI Model of an Elliptical Notch Located inside of an Ogilvy Model Defined Weld; Bottom: True B-Scan Response of the Elliptical Notch Located inside of an Ogilvy model Defined Weld....................... 52 Figure 5.25 Top: Depiction of the MHI Model of the Examination of Two Rectangular Notches with One Notch Located outside of the Weld Region and the Other Notch Located in the Middle of a Weld Defined as a Single Orthotropic Region; Bottom: True B-Scan Response for the Simulation of Two Rectangular Notches................................................................................... 53 Tables Table 1.1 Summary of Participating Teams and Software Platforms.................................... 2 Table 2.1 Summary of Specifications for PAUT Probe Employed by Procedure 114-PA1........................................................................................................................ 3 Table 3.1 Summary of the Reflector Specifications for the Probe Calibration Step.............. 6 Table 3.2 Summary of Response Features from Echodynamic Scan and Time Plots for All Teams......................................................................................................... 8 Table 3.3 Summary of Calibration SDH Response Amplitudes for the Consistent Probe Model in CIVA and 2D and 3D FEM Simulations...................................... 12 Table 4.1 Ogilvy Parameters Utilized in MHI CIVA Weld Mode.......................................... 25 Table 5.1 Summary of Conventional UT Probe Calibration Results for CIVA Teams from PNNL, MHI, and SVTI, and for an FEM Team from JAPEIC...................... 33 Table 5.2 Summary of CIVA Simulation Settings Input Likely Contributing to Differences in PNNL Results in Comparison to MHI and SVTI Results.............. 34 Table 5.3 Summary of Notch Reflector Responses (Relative Amplitude of Notch Tip Response to Notch Corner) for SVTI and MHI Simulation Scenarios................. 54

PNNL-33733 Introduction 1

1.0 Introduction This report documents the outcomes of a collaborative nondestructive evaluation (NDE) modeling activity performed under the Program for Investigation of NDE by International Collaboration (PIONIC). PIONIC includes participants from several organizations in the United States, Finland, Sweden, Japan, South Korea, and Switzerland. The scope of PIONIC includes four topic areas defined by the participants as follows:

• Topic Area 1: NDE Modeling and Simulation

• Topic Area 2: Flaw Relevance Evaluation

• Topic Area 3: Materials Degradation Monitoring in Extended Periods of Operation

• Topic Area 4: Probability of Detection Analysis.

The activities described in this report were performed as part of Topic Area 1: NDE Modeling and Simulation.

The objective of the collaborative NDE modeling activity was to utilize the collective capabilities of PIONIC to evaluate the reliability of computer modeling and simulation tools for NDE as applied to a weld exam. This effort sought to address two main questions:

• What degree of similarity can be expected for results generated by multiple independent teams implementing the same examination scenario with the same tool?

• How sensitive are results to different techniques for modeling anisotropy in welds?

This effort did not attempt to address questions of model validation (i.e., how well the models can reproduce empirical phenomena), or attempt to make claims about the relative merits of different tools for simulating different scenarios. Rather, this effort was focused on the consistency of the tools and their sensitivity to user variation. The outcome of this effort can serve industry and regulators by providing information on the level of detail required for the creation of standard NDE modeling and simulation procedures and for effective regulatory reviews of results from NDE modeling and simulation.

An initial comparison is made between modeling results and empirical results obtained from the Program to Assess the Reliability of Emerging Nondestructive Technologies (PARENT) open testing (Meyer, Holmes, and Heasler 2017), in section 4.4. However, this does not represent an exhaustive validation effort. The dominant software package used was Extende CIVA; however, some teams also contributed finite element modeling (FEM) capabilities. A summary of the participating teams and software platforms is provided in table 1.1.

PNNL-33733 Introduction 2

Table 1.1 Summary of Participating Teams and Software Platforms Team Country Platform Pacific Northwest National Laboratory (PNNL)

USA CIVA Swiss Association for Technical Inspections, Nuclear Inspectorate (SVTI)

Switzerland CIVA Mitsubishi Heavy Industries (MHI)

Japan CIVA Japan Power Engineering and Inspection Corporation (JAPEIC)

Japan ComWAVE (FEM)

VTT Technical Research Center of Finland Finland CIVA The activity, leveraging efforts from PARENT, was initially performed based on a case study from PARENT open testing (Meyer, Holmes, and Heasler 2017). The selected test case included the exam of a flat bar (FB) test block by team 114, using procedure 114-PA1 (i.e., a phased array ultrasonic testing [PAUT] procedure). The test block and procedure are described in section 2.0. The FB test blocks provided a relatively simple geometry, and multiple FB test blocks were fabricated, which included a variety of simulated defects (i.e., notches, mechanical fatigue cracks, laboratory-grown stress corrosion cracks [SCC]). This is beneficial for enabling the comparison of notch responses to responses from more realistic flaws.

The first part of the activity emphasized variation in probe model definitions resulting from user input selections and minimized variation related to flaw and material definitions by considering simplified flaw and material scenarios (e.g., excluding a weld). In the next part of the exercise, participants attempted to model the response from a planar reflector in a weld with anisotropic properties. In the calibration exercise, participants were initially provided with written specifications for the probe and inspection procedure. Next, a common CIVA model of the probe was distributed to every team as an.xml file for implementation.

After completing the exercises based on the 114-PA1 procedure, a second activity was initiated based on a simpler conventional ultrasonic testing (UT) probe and procedure. The decision to perform the conventional UT simulation activity was made based on the experiences and outcomes of the activity with the more complex 114-PA1 procedure.

Section 2.0 of this report includes a description of the 114-PA1 procedure and test block that is modeled in these activities. Section 3.0 describes the 114-PA1 probe model calibration step and presents results obtained from written probe specifications and from using a common CIVA.xml probe model. Section 4.0 presents results from simulations of planar reflectors in a weld with anisotropic properties for the 114-PA1 procedure. Next, section 5.0 summarizes the conventional UT simulation activity, including calibration exercises and simulations of planar reflectors in anisotropic welds. Finally, section 6.0 provides a summary of observations and conclusions from these efforts.

PNNL-33733 Case Study Description 3

2.0 Case Study Description The collaborative PIONIC modeling activity was designed to leverage and extend the work conducted in PARENT by basing the activity on examinations performed as part of the PARENT open testing (Meyer, Holmes, and Heasler 2017). Data collected under PARENT can serve as empirical baselines for comparing the results from modeling attempts.

2.1 Procedure Case Study The procedure utilized for this case study was a PAUT procedure employed by team 114 in PARENT open testing (Meyer, Holmes, and Heasler 2017). The procedure is labeled 114-PA1 and was used in examinations on small bore dissimilar metal weld test blocks P1, P4, P41, and on FB test blocks P28, P29, P30, P31, P32, P38, P42, and P46. Procedure 114-PA1 is a commercial procedure that uses 1.5 MHz probes with a transmit-receive configuration and employs both shear waves (SWs) and longitudinal waves (LWs). The procedure is applied for detection and characterization of cracks originating from the inner diameter (ID) surfaces of pipes and is implemented by accessing the outer diameter (OD) surface. Procedure 114-PA1 uses manual encoded scanning from both sides of the weld in PARENT open testing. A summary of procedure 114-PA1 is provided in appendix C of NUREG-CR/7236 (Meyer, Holmes, and Heasler 2017). The procedure employs linear scanning at multiple angles for both SW and LW modes. For circumferential flaws, the specific examination angles for the SW modes are 45 degrees and 60 degrees; for the LW modes, they are 30 degrees, 45 degrees, 60 degrees, and 70 degrees.

2.2 Probe Specifications The specifications for the PAUT probe employed by procedure 114-PA1 were distributed to participants by team 114 to facilitate the creation of models of the probe in CIVA and FEM software packages. These specifications are summarized in table 2.1, which provides physical information related to the geometry of the elements in the probe and the wedge and includes the nominal operating frequency and 6 dB bandwidth.

Table 2.1 Summary of Specifications for PAUT Probe Employed by Procedure 114-PA1 Parameter Phased Array Team 114 P41 Phased Array Team 114 and 122 FB blocks Note PA probe Type 1-D Linear Array 1-D Linear Array Probe separation (mm) 25.56 25.50 Configuration Pitch and Catch Pitch and Catch Transmitter probe name 1.5M32x2E64-15 1.5M32x2E64-15 Receiver probe name 1.5M32x2E64-15 1.5M32x2E64-15 Probe skew angle (°)

90.0 90.0 Probe frequency (MHz) 1.5 1.5 Bandwidth (%)

55 55 Number of elements on primary axis 32 32 Primary axis pitch (mm) 2.00 2.00 Primary element size (mm) 2.00 2.00 Secondary element size (mm) 15.00 15.00

PNNL-33733 Case Study Description 4

Parameter Phased Array Team 114 P41 Phased Array Team 114 and 122 FB blocks Note Wedge Curvature along secondar axis (mm)

R = 162 Flat Wedge angle (°)

22.3 22.3 Roof angle (°)

7.0 7.0 Longitudinal sound velocity (m/s) 2330 2330 Height of first element (mm) 8.42 8.42 Primary offset of first element (mm) 8.33 8.33 Secondary offset of first element 12,22 12,22 Primary axis position on wedge reference 85.00 85.00 Secondary axis position on wedge reference 25.00 25.00 Wedge length (mm) 85.00 85.00 Wedge width (mm) 50.00 50.00 Front length L1 (mm) 42.5 42.8 Measured from PA Calculator image using cursors Back length L2 (mm) 42.5 42.2 Height L4 (mm) 18.4 20.5 Despite the detail provided in table 2.1, ambiguity remained with respect to the exact element sequencing and the focusing method. This information is prompted in CIVAs Array Settings input menus. Without precise direction, implementations were subject to the users individual preference and interpretation. Thus, more complete probe specifications were provided by team 114 through screen captures of their own CIVA input menu settings, and this is provided in appendix A.

It is notable that the 114-PA1 procedure only used a group of 10 adjacent elements within the 32-element array at a time and implemented an electronic scanning procedure to sweep the 10 elements sequentially over all the elements in the array.

2.3 Physical Test Block Specifications The case study is developed based on the application of procedure 114-PA1 to an FB test block from PARENT open testing. The nominal thickness of each FB test block is 30 mm (30.3 mm exact dimension). A total of eight FB test blocks were used in PARENT open testing. The coordinate system and dimensions of test blocks P28, P29, P30, P31, P32, P42, and P46 are provided in figure 2.1. In this case, the weld area was in the middle of the specimen. Also, the flaws (one flaw in each block) were located at the middle of the specimens and extended across the full width of the specimens, except for P46, which was blank.

PNNL-33733 Case Study Description 5

Figure 2.1 Coordinate System Definition, Dimensions, and Illustrations of FB Test Blocks P28, P29, P30, P31, P32, P42, and P46

PNNL-33733 Probe Calibration Exercise 6

3.0 Probe Calibration Exercise The first part of the collaborative modeling activity focused on calibrating team outputs for a simple reflector and no weld. The objective of this step was to minimize complexities associated with the reflector and material specifications so that differences in output caused by varying probe implementations could be more easily identified. This step was conducted by defining a test block with two side-drilled hole (SDH) reflectors and providing teams with a description of the probe used by team 114. The description of the probe included the information in table 2.1.

A summary of the reflector specifications is included in table 3.1 below. Initially, teams were instructed to recreate probes in their software packages using the information provided in table 2.1. Appendix B provides screen captures of several CIVA user interface menus and the settings input by the PNNL team, as an example, for readers desiring context on how models are defined in CIVA.

Table 3.1 Summary of the Reflector Specifications for the Probe Calibration Step Dimensions Reflectors Thickness: 30 mm Width: 80 mm Two 3 mm diameter holes separated by 30 mm. One SDH located 10 mm from top and bottom surface of test block to center of hole.

Based on the outcome of the initial activity based on defining the probe with table 2.1, an.xml file of the probe model was then distributed to CIVA teams to make sure a consistent model of the probe was implemented. The remainder of this section provides a summary of the results obtained from instructing teams to recreate probe models based on information in table 2.1 and describes important differences in input settings that contributed to variations in results. Next, the results obtained with the consistent CIVA probe model and 2D and 3D FEM models are provided, followed by discussion of the agreement in results among CIVA teams and agreement in results between CIVA and FEM models.

3.1 CIVA Interpreted Probe Model Response Summary As an example, the results obtained for the PNNL team are displayed in figure 3.1. The results generated by SVTI, VTT, and MHI teams are provided in appendix C. Responses from the two SDHs were evident in the B-scan and True B-scan response images in figure 3.1. The B-scan response provides a 2D map of intensity as a function of signal receive time and scan position.

The True B-scan response provides a corrected image by mapping the signal intensity to the true location within the test specimen. Both B-scan images showed a higher amplitude response from the SDH near the bottom surface. These were consistent with the echodynamic responses of amplitude versus position and amplitude versus time. A larger peak was followed by a smaller peak in the amplitude versus position response, indicating that the SDH located near the bottom surface, which was the first flaw encountered in the scan, produced a more intense response. The order of the large and small peaks was flipped in the amplitude versus time response, also indicating the SDH near the bottom surface generated the larger response.

PNNL-33733 Probe Calibration Exercise 7

Figure 3.1 PNNL Team Results for Interpreted Probe Model Specification Response characteristics for all the CIVA modeling teams are summarized in table 3.3 below.

This table summarizes features obtained primarily from the echodynamic amplitude versus scan position and amplitude versus time plots.

• In the first column, 1st Hole refers to the SDH that is first encountered as the probe is scanned in the positive X direction along the surface of the test specimen, and an entry of lower describes that SDH as being the hole located at greater depth, or farther from the scanning surface relative to the other SDH. An entry of upper describes the SDH as being located at a shallower depth, or closer to the scanning surface relative to the other SDH.

• In the second column, 1st Peak refers to the first peak that is encountered in the echodynamic scanning plot as one scrolls in the direction of increasing scan position. An entry of larger means the first peak encountered has greater amplitude of two response peaks generated by the two SDHs, and an entry of smaller means the first peak encountered has smaller amplitude of the two peaks.

• In the third column, Largest Response refers to the SDH producing the most intense signal response. An entry of lower means that the SDH located at greater depth produces the most intense signal response, and an entry of upper means the SDH located at shallower depth produces the most intense signal response.

• The fourth column displays the Peak Amplitude of the largest SDH response in absolute units (points).

• The fifth column displays the Relative Amplitude of the smaller SDH response as a percentage of the amplitude of the peak of the larger SDH response.

• Finally, the sixth column displays Peak Separation, which refers to the distance between the peak responses generated by the two SDH reflectors, as measured in the amplitude

PNNL-33733 Probe Calibration Exercise 8

versus scan position echodynamic response. It is measured as the difference between the locations of maximum amplitude in the peaks generated by both SDHs.

From table 3.3, it is evident that the lower SDH produced the largest response, except for the SVTI team results and the VTT team results for a linear scan with LW at 70 degrees. Both the SVTI and VTT teams implemented the SDHs such that the SDH located nearer the front wall is encountered first in the scan. It is possible that this orientation of SDHs resulted in a shadowing effect that reduced the response from the lower SDH. It can further be observed that the upper SDH response was more prominent for the high angle (70 degrees) scan by the VTT team. In general, the relative intensity of the upper SDH response, with respect to the lower SDH response, was seen to increase with steering angle.

In reviewing the Peak Separation column in table 3.3, there appears to be two approximate groups of peak separations. One group was near 40 mm, and another group was near 20-25 mm. These groupings appeared to correlate with the order in which the upper and lower SDHs are positioned in the test block. The positioning of the upper SDH before the lower SDH resulted in a smaller peak separation and supported the hypothesis of interference of the response from the lower SDH by shadowing effect.

Finally, in reviewing the Peak Amplitude column in table 3.3, it was evident that approximately three groupings were present based on the absolute maximum intensity observed. Both the PNNL and SVTI team results indicated an intensity that is much greater than that observed by the VTT and MHI teams. The significant distinction between these two groups of teams was that the PNNL and SVTI teams implemented array models in which all 32 elements of the probe were fired in one shot. In the VTT and MHI array models, only 10 elements of the array were used per shot. The intensity of the MHI results were also significantly greater than the VTT results. In this case, the sectorial scan implementation by MHI likely resulted in the increased intensity relative to the linear scan implementations by VTT.

Table 3.2 Summary of Response Features from Echodynamic Scan and Time Plots for All Teams 1st Hole 1st Peak Largest

Response

Peak Amplitude Relative Amplitude Peak Separation PNNL lower larger lower 64.0 pts 22%

40 mm SVTI upper larger upper 61.0 pts 33%

25 mm VTT-L45 upper smaller lower 2.7 pts 62%

25 mm VTT-L60 upper smaller lower 1.4 pts 95%

25 mm VTT-L70 upper larger upper 1.1 pts 83%

20 mm VTT-S45 upper smaller lower 1.4 pts 43%

20 mm VTT-S60 upper smaller lower 0.9 pts 53%

20 mm MHI-L30-70 lower larger lower 9.5 pts 54%

40 mm MHI-S45-60 lower larger lower 5.6 pts 13%

40 mm In addition to the differences in specifying linear versus sectorial scanning, or firing 10 elements versus all 32 elements, input simulation settings also varied among the teams. The most significant variation appeared to be with tracking the incident beam echoes from the flaw and specimen surfaces.

PNNL-33733 Probe Calibration Exercise 9

The PNNL and VTT teams selected the Direct option, from the Control prompt, which restricted the computation to accounting only for interactions with the flaws and direct skips for user-selected specimen surfaces. The SVTI team selected the Half Skip option which restricted computations to accounting for contributions from three successive skips from the specimen surfaces and the flaw (half V). Finally, the MHI team selected the Full Skip option which accounted for a maximum of five successive skips from the specimen surfaces and the flaw (full V). Although these input settings varied among the teams, the effect of this variation on results appeared to be overwhelmed by the effects of variation in the phased array parameter definitions and the positioning of the SDHs. An overview of the input settings for all the modeling teams and discussion of key differences is provided in appendix E.

3.2 Consistent CIVA Probe and Flaw Models Following the exercise with interpreted probe specifications, a CIVA probe file was distributed to participants to make sure a consistent probe model was input by teams. Also, the MHI team changed the Control option setting under the Initialization tab in the Simulation Settings menu from Full Skip to Direct. During this portion of the exercise, the SVTI team was impeded by compatibility issues with the CIVA probe file that was distributed to them, and VTT was unable to provide results with the placements of the SDHs adjusted to be consistent with placements in PNNL and MHI models.

Results from PNNL calibration simulations used for L45 degrees and L60 degrees simulations are provided in figure 3.2 and figure 3.3, and the results from the MHI calibration simulations are provided in figure 3.4 and figure 3.5. A summary of the peak amplitude responses from the SDHs for the consistent probe model are provided in table 3.3 for CIVA model results by VTT, MHI, and PNNL. Despite the difference in SDH placement between VTT models and models by MHI and PNNL, the peak amplitude responses were consistent. The VTT results did deviate from the MHI and PNNL results for predicting which SDH produces the largest response for L60-degree scans. However, the relative amplitudes for the smaller peak responses were all in the upper 90 percent range, indicating that the peaks were nearly equal in amplitude for all teams for L60-degree scans.

As expected, supplying teams with a consistent.xml probe model file to implement resulted in much greater agreement in the results. However, although the results are quantitatively similar, exact numerical agreement was not achieved for either peak absolute intensities or for the relative peak amplitudes. The reasons for this were not fully explored, but it is postulated that variations in input simulation settings are the likely reason for the remaining quantitative differences in the results.

PNNL-33733 Probe Calibration Exercise 10 Figure 3.2 PNNL Team Results for Consistent Probe Model with LW and for a Linear 45° Scan Figure 3.3 PNNL Team Results for Consistent Probe Model with LW and for a Linear 60° Scan

PNNL-33733 Probe Calibration Exercise 11 Figure 3.4 MHI Team Results for Consistent Probe Model with LW and for a Linear 45° Scan Figure 3.5 MHI Team Results for Consistent Probe Model with LW and for a Linear 60° Scan

PNNL-33733 Probe Calibration Exercise 12 Table 3.3 Summary of Calibration SDH Response Amplitudes for the Consistent Probe Model in CIVA and 2D and 3D FEM Simulations Peak Amplitude Relative Amplitude of Smaller Peak VTT - L45 (CIVA) 2.7 pts (lower) 62% (upper)

MHI - L45 (CIVA) 2.2 pts (lower) 59% (upper)

PNNL - L45 (CIVA) 2.8 pts (lower) 67% (upper)

JAPEIC - L45 (FEM - 2D)

(upper) 86% (lower)

JAPEIC - L45 (FEM - 3D)

(lower) 42% (upper)

VTT - L60 (CIVA) 1.4 (lower) 95% (upper)

MHI - L60 (CIVA) 1.3 (upper) 99% (lower)

PNNL - L60 (CIVA) 1.7 (upper) 97% (lower)

JAPEIC - L60 (FEM - 2D)

(upper) 59% (lower)

JAPEIC - L60 (FEM - 3D)

(upper) 93% (lower) 3.3 2D and 3D Finite Element Modeling Results (JAPEIC)

JAPEIC implemented the probe calibration model in the ComWAVE FEM software tool. The FEM implementation of the problem in 2D and 3D is presented here, along with the results from both implementations, which are compared to results obtained by CIVA. The implementation of the 2D FEM model is summarized in figure 3.6, which specifies the properties of the test specimen and the wedge material. Data from CIVA input menus were adapted to define the probe and test specimen in ComWAVE. Simulations of LW modes were performed using the 10-element electronic scan at 45 degrees and 60 degrees. The method of sequencing resulted in 23 element combinations (shots) for each point of data collection; also, SDH reflectors are incorporated, as defined in table 3.1. Unlike the CIVA implementation, positional scanning of the probe was not performed. Rather, the probe remained in a fixed position and was positioned so that the SDH was on the center beam line. In the ComWAVE implementation, the SDHs were implemented separately and the responses from each SDH were computed in individual simulations.

Four B-scan response images from the 2D model are displayed in figure 3.7 for A) the upper SDH at 45 degrees, B) the lower SDH at 45 degrees, C) the upper SDH at 60 degrees, and D) the lower SDH at 60 degrees. Comparison of the upper and lower SDH responses at 45 degrees indicated they were similar in amplitude, while a review of the upper and lower SDH responses at 60 degrees clearly indicated that the upper SDH generated a response with higher peak amplitude. This is quantitatively summarized in table 3.3, which reports the relative peak amplitudes of the lower and upper SDHs for the 2D FEM model and compares them to results obtained with CIVA. This summary shows that the 2D FEM results predicted that the upper SDH provides the largest response at both 45 degrees and 60 degrees. These contradicted results obtained in CIVA modeling by all teams, which showed that the lower SDH produces the largest response at 45 degrees.

PNNL-33733 Probe Calibration Exercise 13 Figure 3.6 Screenshot of Slide Summarizing Defining Parameters of JAPEIC 2D FEM Model for Probe Calibration (Courtesy of JAPEIC)

PNNL-33733 Probe Calibration Exercise 14 Figure 3.7 B-Scan Response Images Obtained by JAPEIC 2D FEM Model for A) Upper SDH at 45°, B) Lower SDH at 45°,

C) Upper SDH at 60°, D) Lower SDH at 60° (Courtesy of JAPEIC)

PNNL-33733 Probe Calibration Exercise 15 Following the 2D FEM modeling effort, JAPEIC implemented a 3D FEM model for probe calibration using ComWAVE. A comparison of the 2D and 3D probe models as input to ComWAVE can be made by viewing the top and bottom screen captures in figure 3.8. The 3D geometry allowed for defining a dual probe with separate element arrays for transmitting and receiving. The probe was maintained in a fixed position, and the response from each SDH was simulated individually. ComWAVE does not support delay law calculations of the wave paths in 3D space, so the calculation was only performed in the central plane, which is depicted in figure 3.9.

Figure 3.8 Depictions of the (Top) 2D and (Bottom) 3D Probe Geometries in the ComWAVE User Interface; the 3D Geometry Allows the Definition of a Dual Transmit-Receive Probe (Courtesy of JAPEIC)

PNNL-33733 Probe Calibration Exercise 16 Figure 3.9 Depiction of the Delay Calculation of Wave Paths Constrained to the Central Plane in the JAPEIC 3D FEM Model (Courtesy of JAPEIC)

The results are displayed for the 3D model as four B-scan response images in figure 3.1 for A) the upper SDH at 45 degrees, B) the lower SDH at 45 degrees, C) the upper SDH at 60 degrees, and D) the lower SDH at 60 degrees. Comparison of the upper and lower SDH responses at 60 degrees indicated they are similar in amplitude, while a review of the upper and lower SDH responses at 45 degrees clearly indicated that the lower SDH generated a response with higher peak amplitude. This is quantitatively summarized in table 3.3, which reports the relative peak amplitudes of the lower and upper SDHs for the 3D FEM model and compares them to results obtained with CIVA. This summary showed that the 3D FEM results predicted that the lower SDH provided the largest response at 45 degrees and the upper SDH provided the largest response at 60 degrees. However, the responses from both SDHs were nearly equal for the 60 degrees scan. This was consistent with the results obtained in CIVA modeling by all teams. The 2D FEM results, which showed that the upper SDH produces the largest response at both 45 degrees and 60 degrees, deviated from both the 3D FEM results and CIVA results.

PNNL-33733 Probe Calibration Exercise 17 The 3D FEM results were both quantitatively and qualitatively similar to the CIVA results reported in section 3.2. However, the results produced by the 2D FEM models did not achieve qualitative similarity with the 3D FEM and CIVA models.

These exercises were based on the implementation of a transmit-receive longitudinal (TRL) probe, which effectively isolates the transmit and receive matrix arrays from cross-talk, transmitting LWs from one side and receiving them on the other side. Because the beam path of a TRL probe cannot be fully captured by a 2D cross section, the 2D implementation ultimately resulted in a distorted representation of the wave propagation.

PNNL-33733 Probe Calibration Exercise 18 Figure 3.10 B-Scan Response Images Obtained by JAPEIC 3D FEM Model for A) Upper SDH at 45°, B) Lower SDH at 45°,

C) Upper SDH at 60°, D) Lower SDH at 60° (Courtesy of JAPEIC)

PNNL-33733 Anisotropic Weld Model Implementation 19 4.0 Anisotropic Weld Model Implementation Following the probe calibration activity, participants attempted to model the response from a notch reflector in an anisotropic weld. This activity was also based on the same procedure, probe, and test specimen specifications described in section 2.0. Participants were instructed to include a 10 mm deep notch reflector with 0.1 mm width located in the center of the weld. The reflector was defined in this way to mimic FB test block P42, which contained a notch reflector fabricated by electrical discharge machining with nominal depth of 10 mm.

4.1 PNNL CIVA Implementation For this step, PNNL implemented an anisotropic weld model based on an electron backscatter diffraction (EBSD) technique for obtaining individual, grain-level detail on crystallographic orientation. EBSD can be used to determine properties such as grain size, shape, and orientation. EBSD data were collected for an actual austenitic weld cross section to create more realistic weld models in CIVA and other potential modeling tools. These data were collected from the cross section of a wrought stainless steel (WSS)-WSS austenitic weld from specimen 3C-022 and was performed as part of an ongoing effort to define best practices required for using computational models (in this case CIVA) to simulate UT scenarios being conducted on nuclear power plant components (Jacob et al. 2020).

For this PIONIC task, the same weld model was applied with small adaptations (dimensional scaling). For EBSD scanning, a cross-sectional slice of the weld was cut. A photograph of the slice with a chemically etched and polished surface is provided in the top of figure 4.1, which partially reveals grain sizes and shapes in the weld. The bottom of figure 4.1 displays an example inverse pole figure (IPF) of EBSD data, representing the y-component of the crystal orientation relative to the sample surface. The EBSD data were collected with a resolution of 4 µm, so importing the image into CIVA required a process of downsampling and conversion to computer-aided design (CAD) format (see figure 4.2). The original weld cross section was 53 mm thick, which was scaled down to adapt to the 30 mm thickness of the specimen defined by table 3.1. The downsampled model contains 882 individual grain regions, which are 0.469 mm x 0.469 mm squares after scaling.

After production of the anisotropic grain structure in CAD format and import into CIVA, user input of Euler angles for each of the individual grain regions could be performed. Because the inputting of Euler angles from the EBSD data for each of the individual 882 regions could be quite tedious, the model was further simplified to facilitate faster implementation and initial evaluation of the approach. Therefore, the EBSD data were used to bin regions of the weld into one of seven distinct Euler angle sets that were obtained from literature. The assignment of the Euler angle set to individual grain regions was correlated with the EBSD data, as similar regions defined by the EBSD data were assigned the same Euler angle set. More details of the EBSD weld model implementation for the WSS-WSS austenitic weld sample can be found in Jacob et al. (2020).

PNNL-33733 Anisotropic Weld Model Implementation 20 Figure 4.1 (Top) Photograph of Etched and Polished Cross Section of WSS-WSS Austenitic Weld from Specimen 3C-022; (Bottom) IPF Displaying Y-Component of the Crystal Orientations Relative to the Surface of the Weld Cross Section

PNNL-33733 Anisotropic Weld Model Implementation 21 Figure 4.2 Illustration of the Process of Converting a Downsampled EBSD Image of a Weld Cross Section to CAD Format that Can Be Imported into CIVA.

Next, PNNL performed simulations of a 45-degree LW linear scan. Results from the simulation are displayed as an electronic scan response overlayed on an outline of the test specimen and weld in the top of figure 4.3, and a ray trace diagram in the bottom of figure 4.3 depicts the resulting beam scattering and redirection by the modeled weld. Electronic scan results are displayed because the position of the probe was not scanned in this case. Rather, the images in figure 4.3 were obtained from a single probe position. In figure 4.4, the PNNL team response is presented, including the echodynamic responses of amplitude versus position and amplitude versus time. The echodynamic scans displayed two peaks. The largest peak was associated with the flaw response and had an amplitude of 0.824 pts, while the second, smaller peak was associated with the response from the weld root and had an amplitude of 0.669 pts.

To compare the results in figure 4.3 and figure 4.4 against a baseline condition, the individual grain regions in the weld model were all converted to isotropic regions with properties matching that of the base material. The simulation was run under this condition, generating the output displayed in figure 4.5 and figure 4.6. Figure 4.5 and figure 4.6 show distinct responses from the flaw tip and flaw corner. This contrasts with the response displayed in figure 4.3 and figure 4.4 for the anisotropic model in which a distinct tip response is not evident. For the baseline condition, the amplitude of the flaw tip response (0.291 pts) was larger than the amplitude of the corner response (0.238 pts). These results are unexpected, but the difference is scanning for the images in figure 4.3 and figure 4.4 may be a factor. Through further discussions with

PNNL-33733 Anisotropic Weld Model Implementation 22 PIONIC participants, it is postulated that the effect of a larger tip response relative to corner response in figure 4.4 could be caused by simulation settings that fail to adequately capture the reflections from the specimen surfaces. The Direct option at the Control prompt under the Initialization tab of the Simulation Settings menu (appendix B.6) was selected in these simulations, which only accounts for interactions with the flaw and selected specimen surfaces.

In contrast, the Half Skip and Full Skip options account for more surface interactions. Thus, this example highlights the potential significance these setting options have on generated flaw responses.

Figure 4.3 (Top) B-Scan Response Obtained for 45° LW Scan by PNNL Team with Anisotropic EBSD Weld Model and Notch Reflector and (Bottom) Ray Trace Diagram Depicting the Effect of the Modeled Weld Anisotropy on Beam Scatter and Redirection

PNNL-33733 Anisotropic Weld Model Implementation 23 Figure 4.4 PNNL Team Responses for 45° LW Scan with Anisotropic EBSD Weld Model and Notch Reflector Figure 4.5 True B-Scan Baseline Response Obtained for 45° LW Scan with Isotropic Weld Properties and Notch Reflector by PNNL Team

PNNL-33733 Anisotropic Weld Model Implementation 24 Figure 4.6 Baseline Response Obtained for 45° LW Scan with Isotropic Weld Properties and Notch Reflector by PNNL Team, Including B-Scans and Echodynamic Response Images 4.2 Ogilvy Weld Model The Ogilvy model (Ogilvy 1985) refers to a simplified representation of complex anisotropic weld structures through approximate functions that define crystallographic orientation as a continuously varying function of spatial position. Equation 4-1 relates parameters defining crystallographic orientation, as depicted in figure 4.7, to position. Here, is the local grain orientation relative to this reference, D is half the weld root width, is the angle that the weld boundary makes with the normal to the specimen back wall, T is proportional to the tangent of the grain axes at the welds boundary with the base material, and (where 0 1) is a measure of how fast the grain orientation changes along the x-axis.

tan = T(D + z tan )

x 4-1 As a result, an anisotropic weld microstructure can be defined with a macro-model by selecting just a few key parameters, which avoids the arduous task of providing grain-level detail. CIVA accommodates defining weld structures through an Ogilvy model, and it has been investigated in previous PNNL modeling and simulation reports (Dib et al. 2018; Jacob et al. 2020).

PNNL-33733 Anisotropic Weld Model Implementation 25 Figure 4.7 Diagram Depicting the Various Parameters (T, D,,, ) that Determine Variation of Crystallographic Orientation with Position through Eq. 4-1.

4.3 MHI CIVA Implementation Following the probe calibration activity, the MHI team implemented a weld model into the test specimen based on the weld geometry depicted for the FB blocks in figure 2.1 and the implementation of an Ogilvy model through the CIVA provided menus. To facilitate simpler comparison with FEM results from JAPEIC, MHI chose to model a conventional probe for a 45-degree LW inspection scenario.

The Ogilvy model was defined through a CIVA user input menu as shown in figure 4.8. The model was defined under the Specimen menu and the Material tab. The material defining the weld region was selected, and the option Parametric anisotropic orientation (Welding) was selected at the Type prompt. Under the Orientations subtab, Ogilvy was selected for the Law Type, which provided prompts to enter values for the Ogilvy parameters. The values selected for this instance are summarized in table 4.1. Three different cases were simulated to observe the influence of the T parameter. These results are presented in the form of B-scans and ray tracing images in figure 4.9. For comparison, the results from simulation with an isotropic weld condition are also provided in figure 4.9. For all cases, distance tip and corner responses were evident and there appeared to be minimal distortion of these signals. The main effect of the Ogilvy model appeared to be beam redirection.

Table 4.1 Ogilvy Parameters Utilized in MHI CIVA Weld Mode.

Parameter Value(s)

D 9 mm T

0.3, 0.5, and 0.7 (three cases) 15° 0.85

PNNL-33733 Anisotropic Weld Model Implementation 26 Figure 4.8 Screenshot of CIVA Input Menu for Defining Ogilvy Model Figure 4.9 MHI Team CIVA Responses for Isotropic Weld Model and for Ogilvy Model, with T = 0.3, 0.5, and 0.7

PNNL-33733 Anisotropic Weld Model Implementation 27 4.4 Team 114 Response Data from Specimen P42 in PARENT Response images collected from the test case scenario in PARENT open testing (Meyer, Holmes, and Heasler 2017) are included in this section. In this case, the response images were obtained from the scanning of test block P42 by team 114. Details regarding the dimensions of the P42 test specimen, weld region, and the materials are provided in figure 2.1. Test specimen P42 had a 10 mm deep electrical discharge machining notch machined into the middle of the weld, and the notch reflector used in simulations described in section 4.1 was based on this specimen.

4.4.1 Discussion of LW Responses B-scan response images for 45 LW, 60 LW, and 70 LW scans are provided in figure 4.1. Images obtained from scans approaching from both sides of the weld are shown and labeled as skew =

90 degrees and skew = 270 degrees scan images, respectively. The figure also includes traces of the maximum amplitude versus depth to the left of each B-scan image. An outline of the test block boundaries and the boundaries of the weld are superimposed on the B-scan images of figure 4.1.

CIVA response simulations for an isotropic weld model were generated for the 45 LW and 60 LW examination scenarios, including responses in which mode conversion was not simulated and responses that include mode conversions (see figure 4.1). These were generated to aid in the interpretation of the results in figure 4.1. In figure 4.1, the 45 LW and 60 LW B-scan images show strong indications outside and below the test block region. These are mode-converted SW signals from the notch reflector, and the CIVA responses including mode conversion on the right in figure 4.1 are consistent with this. There are also discrepancies between the PARENT results in figure 4.1 and the CIVA results in figure 4.1. For instance, the locations of the mode-converted SW results generated by CIVA appeared to be rotated counterclockwise relative to the PARENT skew = 90 results in figure 4.1. In fact, the CIVA output for the 60 LW examinations placed all the indications within the upper and bottom surface boundaries of the test block. In figure 4.1, it appears that the mode-converted indications are still located below the test block boundary or at the bottom surface boundary.

In reviewing the 45 LW and 60 LW responses in figure 4.1, it is possible to identify notch tip and corner signals that appear to be associated with the primary LW beam. There are indications within the skew = 90 and 270 images for the 45 LW scan and the skew = 270 image for the 60 LW scan that correspond to expected locations for tip and corner signals, respectively.

However, the primary LW signals are weaker than the mode-converted SW signals, which is not consistent with the CIVA outputs in figure 4.1 that indicate that the primary LW signals are still stronger than the mode-converted SW signals. The other B-scan responses in figure 4.1 (skew

= 90 image for the 60 LW exam, skew = 90 and 270 for the 70 LW exam) do not show indications that correspond with the expected locations of LW tip and corner signals.

PNNL-33733 Anisotropic Weld Model Implementation 28 Figure 4.10 B-scan Response Images Obtained by Team 114 in PARENT Open Testing from FB Test Block P42

PNNL-33733 Anisotropic Weld Model Implementation 29 Figure 4.11 B-Scan Response Images Generated in CIVA from a Notch Reflector with Isotropic Weld Properties; Responses Including Only the Primary LW Interactions are Shown on the Left, and Responses Including Mode-Converted SW Signals are Shown on the Right 4.4.2 Discussion of SW Responses Results from 45 SW and 60 SW scans by team 114 on FB test block P42 in PARENT open testing are shown in figure 4.1 for both skew = 90 and skew = 270. These images appeared simpler than the images produced by LW scans shown in figure 4.1 due to the absence of mode conversion of the primary beam. Although strong corner responses are evident in each of the scan images, it is difficult to see a distinguishable tip signal for most of the response images in figure 4.1, with the exception of the B-scan image for the 45 SW scan at skew = 270.

The B-scan images in figure 4.1 can be compared with CIVA-generated response images for SW scans from a notch reflector in figure 4.1. Here, the orientations of the corner responses in figure 4.1 and figure 4.1 are consistent and correlate with the steering angle. The tip signals generated in figure 4.1 are much weaker than the corner signals. For both the 45 SW and 60 SW scans, the peak amplitude of the tip signal was 37 percent of the peak amplitude for the corner signal.

PNNL-33733 Anisotropic Weld Model Implementation 30 Figure 4.12 B-Scan Response Images Obtained from SW Scans on FB Test Block P42 by Team 114 in PARENT Open Testing

PNNL-33733 Anisotropic Weld Model Implementation 31 Figure 4.13 Simulated CIVA Responses for 45 SW and 60 LW Scans on Notch Reflector with Isotropic Weld Properties

PNNL-33733 Conventional Probe Implementation 32 5.0 Conventional Probe Implementation The outcomes of the probe calibration exercise in section 3.0 and weld anisotropy modeling exercise in section 4.0 showed that the PAUT procedure (114-PA1) was challenging for teams to implement consistently and that implementing in FEM tools was a significant burden.

Therefore, teams were instructed to implement a conventional UT probe model in a subsequent exercise. The exercise was performed based on the same test block described in section 2.3.

Specifications for the conventional UT probe and procedure model are included in appendix D.

5.1 Results of Conventional UT Probe Calibration Initially, simulations with the conventional UT probe applied to the calibration block with two SDH reflectors were performed by the teams. The size and locations of the SDHs were specified the same as in table 3.1 with the additional instruction that the SDH nearest the lower surface (ID surface) should be scanned before the upper SDH to avoid the shadowing effect of the upper SDH on the lower SDH response. A depiction of this scenario is illustrated in figure 5.1, based on PNNLs implementation. Results from the PNNL team are provided in figure 5.2.

Responses from the two SDHs can be seen in the B-scans and in the echodynamic scan of amplitude versus scan position. A summary of the results for teams from PNNL, MHI, and SVTI is provided in table 5.1 by recording the peak absolute amplitude of the lower SDH response and the relative peak amplitude of the upper SDH response. Table 5.1 also includes results from a 3D FEM implementation by JAPEIC.

Figure 5.1 Illustration of Conventional UT Probe (See Appendix D) Scan of Two SDH Reflectors in CIVA Model

PNNL-33733 Conventional Probe Implementation 33 Figure 5.2 PNNL Response for Conventional UT Probe Scan of Two SDH Reflectors Showing B-Scan Images (Top) and Echodynamic Amplitude vs Scan Position (Bottom Left) and Echodynamic Amplitude vs Time (Bottom Middle)

Table 5.1 Summary of Conventional UT Probe Calibration Results for CIVA Teams from PNNL, MHI, and SVTI, and for an FEM Team from JAPEIC Peak Amplitude Relative Amplitude of Smaller Peak SVTI - L45 (CIVA)

Not Reported (lower) 26.3% (upper)

MHI - L45 (CIVA) 9.2 pts (lower) 26.5% (upper)

PNNL - L45 (CIVA) 7.8 pts (lower) 25.7% (upper)

JAPEIC - L45 (FEM-3D)

(lower) 15.0% (upper)

Overall, the results for all the CIVA models were nearly identical, though not exact. PNNL, MHI, and SVTI each computed a relative peak amplitude from the upper SDH response of approximately 26 percent with respect to the peak amplitude of the lower SDH response. It is still surprising that relative peak amplitudes from the upper SDH computed by PNNL and MHI deviated by nearly 1 percent and that a 1.4 pt difference existed between the absolute peak amplitude responses for the lower SDH. Input parameter settings by all teams were reviewed to identify the reason for the difference. In this case, it was found that the SVTI and MHI teams chose different simulation settings from PNNL. PNNL maintained the settings in figure B.16, figure B.17, and figure B.20, which defined simulation settings using the Direct option for the Control prompt. This limited the simulation to tracking reflections directly from defined flaws and/or directly from selected surfaces of the specimen. Mode tracking could be explicitly defined as shown in figure B.17, and only incident and reflected LW waves from flaws were tracked in the PNNL simulation. Finally, the option to approximate the incident beam as a plane wave was

PNNL-33733 Conventional Probe Implementation 34 selected under the Options tab in Figure B.20. In contrast, the SVTI and MHI teams selected simulation settings as shown in figure 5.3. The Half Skip option was selected for the Control prompt under the Initialization tab. This allowed the simulation to track up to three (3) successive reflections from the surfaces of the specimen and the flaw. Finally, the Full Incident Beam option is selected under Options. A summary of the discrepancy in input simulation parameters is provided in table 5.2.

Figure 5.3 CIVA Simulation Settings Selected by MHI and SVTI in the Conventional UT Probe Calibration Exercise Table 5.2 Summary of CIVA Simulation Settings Input Likely Contributing to Differences in PNNL Results in Comparison to MHI and SVTI Results Initialization - Control Field/reflector interaction SVTI Half skip Full incident beam MHI Half skip Full incident beam PNNL Direct Plane wave approximation A trough was observed in the echodynamic amplitude versus scan position response for the upper SDH in the PNNL response, as shown in figure 5.2, and featured more prominently in figure 5.4. This trough split the upper SDH response into two smaller peaks. A similar feature

PNNL-33733 Conventional Probe Implementation 35 can be observed in the responses by MHI and SVTI in figure 5.5 and figure 5.6, respectively.

For all these responses, the relative peak amplitude of the upper SDH response in table 5.1 is based on the larger of the two smaller peaks observed in the upper SDH response. Finally, the results of 3D FEM simulation by JAPEIC are shown in a B-scan in figure 5.7 and by echodynamic scans in figure 5.8. As shown in table 5.1, the 3D FEM results deviated from the CIVA results, although the larger response of the lower SDH was still predicted. The response from the upper SDH was weaker and the split peak feature observed in the CIVA results is not apparent in figure 5.8. The lack of the split peak feature could be caused, in part, by the FEM simulation capturing more complex interactions of the incident beam with the specimen surfaces and flaw, which resulted in a lower signal-to-noise ratio for the upper SDH and less distinct response. However, this does not explain why the overall peak response of the upper SDH for the FEM simulations was weak in comparison to the response of the upper SDH responses in CIVA.

Figure 5.4 B-Scan Response Image (Left) and Echodynamic Amplitude vs Scanning Position Response (Right) for the PNNL Simulation of the Conventional UT Probe Scan of Two SDH Reflectors

PNNL-33733 Conventional Probe Implementation 36 Figure 5.5 MHI Response for Conventional UT Probe Scan of Two SDH Reflectors Showing B-scan Images (Top Left) and Echodynamic Amplitude vs Scan Position (Top Right) and Echodynamic Amplitude vs Time (Bottom Right)

Figure 5.6 SVTI Response for Conventional UT Probe Scan of Two SDH Reflectors Showing the Echodynamic Amplitude vs Scan Position Response

PNNL-33733 Conventional Probe Implementation 37 Figure 5.7 B-Scan Response Image for 3D FEM Simulation of Conventional Probe Response from Two SDH Reflectors by JAPEIC Figure 5.8 Echodynamic Amplitude vs Time (Left) and Echodynamic Amplitude vs Scan Position (right) for 3D FEM Simulation of Conventional Probe Response from Two SDH Reflectors by JAPEIC

PNNL-33733 Conventional Probe Implementation 38 5.2 Results of Conventional UT Probe with Complex Weld Models Following the calibration exercise for the conventional UT probe, teams from PNNL, SVTI, and MHI input anisotropic weld models into the test block models and replaced the SDH reflectors with notch reflectors. This was the same progression of exercises as was followed for the implementation of the team 114 PAUT probe model discussed in sections 3.0 and 4.0. Notches with 10 mm depth spanning the full width of the test block were specified in this exercise.

5.2.1 PNNL CIVA Implementation PNNL implemented an anisotropic weld model based on an EBSD technique for obtaining individual grain-level detail on crystallographic orientation. The technique described in section 4.1 was used, except that new EBSD data were generated from a cross-sectional sample of a weld obtained for a test block described in section 2.3. A process described in Jacob et al.

(2020) was used for downsampling the EBSD data (acquired with 10 m resolution) into square pixels. The downsampled model contained a matrix of 50 x 39 square pixels (individual grain regions), which were each 0.640 mm x 0.640 mm in size. Figure 5.9 shows the cross-sectional weld sample that was scanned with a red line outlining the weld region. The middle image of figure 5.9 displays an IPF representation of the EBSD scan data, and the right-side image of figure 5.9 shows the weld model after downsampling and converting into a matrix of square pixels.

Figure 5.9 (Left) the Cross-Sectional Weld Sample That Was Scanned with a Red Line Outlining the Weld Region, (Middle) an IPF Representation of the EBSD Scan Data, and (Right) the Weld Model after Downsampling and Converting into a Matrix of Square Pixels After conversion to the matrix of square pixels, the weld image was further converted into a CAD file to allow import into CIVA. An automated technique (using MATLAB) for assigning Euler angle values to each square pixel was applied and was based on the collected EBSD data. This process was different from the process described in section 4.1, in which case, assigned EBSD

PNNL-33733 Conventional Probe Implementation 39 values were obtained from literature. Overall, the EBSD weld model that was implemented as part of this exercise had more direct relevance to the test block described in section 2.3 than the EBSD model described in section 4.1. A 3D depiction of the conventional UT probe with the EBSD generated anisotropic weld model is displayed in figure 5.10.

Figure 5.10 CIVA 3D Depiction of the Conventional UT Probe with EBSD Generated Anisotropic Weld Model 5.2.1.1 Simulations with Two Rectangular Notches Simulations were performed with two rectangular notch reflectors of equal depth (10 mm) and spanning the entire width of the test block. Both notches were surface breaking on the ID surface of the test block and were positioned so that they were separated by 30 mm in the X direction (i.e., the axial direction, or direction perpendicular to the orientation of the weld). One notch was positioned within the anisotropic weld, while the second notch was positioned outside of the weld region. The result of the simulation is shown in figure 5.1. The response from the notch outside of the weld region can be compared to the response from the notch within the anisotropic weld. As expected, the notch response within the anisotropic weld was distorted when compared to the response from the notch outside of the weld, which displayed distinct tip and corner responses in the True B-scan and B-scan images, respectively, and showed a stronger corner response relative to tip response. For the notch within the anisotropic weld region, the responses from the corner and tip were less distinct and the peak of the response was no longer located at the corner of the notch but closer to its tip. However, the phenomenon of the notch corner response losing relative intensity with respect to the notch tip was not observed when exploring the effect of notch location, described in the next section. Thus, it appeared that the second notch located outside of the weld region might have interfered with the notch response within the weld region.

PNNL-33733 Conventional Probe Implementation 40 Figure 5.11 PNNL Response for Conventional UT Probe Scan of Two Notch Reflectors Showing a True B-Scan Image (Top Left), a B-scan Image (Top Right),

Echodynamic Amplitude vs Scan Position (Bottom Left) and Echodynamic Amplitude vs Time (Bottom Middle) 5.2.1.2 Effect of Notch Location within Weld Region The location of the notch within the weld region was varied to observe how increased path length through the anisotropic weld would impact the response from the notch reflector. This was performed assuming that an increased path length in the anisotropic weld region would lead to more scattering of the incident beam and greater distortion of the notch response.

Simulations were performed locating the notch at the near side of the weld, the far side of the weld, and outside of the weld region on the far side of the weld. The responses for these three notch locations are provided in figures 5.12, 5.13, and 5.14, respectively. Distortion effects included the loss of distinction between corner and tip responses with increasing beam path length through the weld. Separate notch and tip responses can be identified in all figures; however, the responses smeared into each other. For instance, in figure 5.12, the notch and tip responses are separated by an almost complete drop in signal when viewing the B-scan. The echodynamic amplitude vs. time scan also displayed a deep valley between the peaks generated by the tip and the corner. In figure 5.13, the smearing is evident as the valley between the peaks generated by the tip and corner in the echodynamic amplitude vs. time is shallower. Also, the B-scan showed that the signal does not drop out as much between the tip and corner responses.

Finally, the response for the notch located outside of the weld region, on the far side of the weld (figure 5.14), also showed some smearing. However, the responses from the tip and notch were more distinct, which is most evident in the echodynamic amplitude vs. time scan, than for the notch located within the far side of the weld. It is also notable that the tip response is equal in amplitude (even slightly greater) to the corner response. The equilibration of the tip and corner responses is consistent with the idea that the tip and notch responses should lose distinction because of increased scattering of the sound beam through the anisotropic weld material.

However, it is not obvious why the tip and notch responses are more distinct in figure 5.14 than

PNNL-33733 Conventional Probe Implementation 41 figure 5.13. These results indicate that distortion of flaw responses does not only depend on distance traveled by the beam through anisotropic media, but that actual location of a flaw within highly scattering weld material or within isotropic base material is also important.

Figure 5.12 PNNL Response for Conventional UT Probe Scan of a Notch Reflector Located within the Near Side of the Anisotropic Weld Showing a True B-Scan Image (Top Left), a B-Scan Image (Top Right), Echodynamic Amplitude vs Scan Position (Bottom Left) and Echodynamic Amplitude vs Time (Bottom Middle)

Figure 5.13 PNNL Response for Conventional UT Probe Scan of a Notch Reflector Located within the Far Side of the Anisotropic Weld Showing a True B-Scan Image (Top Left), a B-Scan Image (Top Right), Echodynamic Amplitude vs Scan Position (Bottom Left) and Echodynamic Amplitude vs Time (Bottom Middle)

PNNL-33733 Conventional Probe Implementation 42 Figure 5.14 PNNL Response for Conventional UT Probe Scan of a Notch Reflector Located within the Far Side of the Anisotropic Weld Showing a True B-Scan Image (Top Left), a B-Scan Image (Top Right), Echodynamic Amplitude vs Scan Position (Bottom Left) and Echodynamic Amplitude vs Time (Bottom Middle) 5.2.1.3 Effect of Weld Geometric Asymmetry The effect of the weld geometric asymmetry was explored by simulating notch reflector responses for skew angles of 90 degrees and 270 degrees (i.e., approaching the weld from opposite directions). Because the anisotropic weld model was developed based on an actual weld cross section (see figure 5.9), the model incorporates realistic geometry asymmetry. Thus, notch reflector responses were not expected to be identical for skew angles of 90 degrees and 270 degrees. Simulations were performed with a notch located within the far side of the weld for skew angles of 90 degrees and 270 degrees in figure 5.15 and figure 5.16, respectively. The responses appear to be qualitatively similar, although subtle differences are apparent. It is easiest to determine that the responses are not identical by reviewing the echodynamic amplitude vs. time plots in figure 5.15 and figure 5.16. Distinct tip and corner responses can be observed in the echodynamic amplitude vs. time plots in figure 5.15 and figure 5.16; however, the peak amplitude of the corner response relative to the tip response appears greater for the skew 270-degree scan than for the skew 90-degree scan. Overall, the results shown in figure 5.15 and figure 5.16 were consistent with expectations because examinations for both 90-degree and 270-degree skews resulted in similar but not identical responses.

PNNL-33733 Conventional Probe Implementation 43 Figure 5.15 PNNL Skew Angle 90° Response for Conventional UT Probe Scan of a Notch Reflector Located Outside of the Anisotropic Weld, on the Far Side, Showing a True B-Scan Image (Top Left), a B-Scan Image (Top Right), Echodynamic Amplitude vs Scan Position (Bottom Left) and Echodynamic Amplitude vs Time (Bottom Middle)

Figure 5.16 PNNL Skew Angle 270° Response for Conventional UT Probe Scan of a Notch Reflector Located Outside of the Anisotropic Weld, on the Far Side, Showing a True B-Scan Image (Top Left), a B-Scan Image (Top Right), Echodynamic Amplitude vs Scan Position (Bottom Left) and Echodynamic Amplitude vs Time (Bottom Middle)

PNNL-33733 Conventional Probe Implementation 44 5.2.2 SVTI CIVA Implementation SVTI implemented an anisotropic weld model by defining the weld region as a single region with orthotropic properties. Orthotropy is a type of anisotropy in which material properties vary along three mutually orthogonal axes. Figure 5.17 shows a CIVA input menu for defining the elastic constants of an orthotropic material. In this case, a set of representative elastic constants were defined by SVTI. A simulation was performed with an elliptical shaped notch reflector located in the middle of the defined weld region, as shown in the top of figure 5.18. The True B-scan image response and the echodynamic amplitude vs. time scan are shown in the middle and bottom of figure 5.18, respectively. In the echodynamic amplitude vs. time scan, it is noted that the amplitude of the tip response for the notch is 7.3 dB less than the amplitude of the corner response.

A parametric study of matrix coefficients C11, C22, and C33 was performed by assigning them the same values of 150 GPa, 200 GPa, and 250 GPa. The results for these simulations are provided in figure 5.19 and show that the amplitude of the tip response decreases relative to the amplitude of the corner response for increasing values of the C11, C22, and C33 constants. The time associated with the tip response also changes, indicating a slight change in velocity.

Figure 5.17 Stiffness Matrix Definition for the Definition of a Weld as a Single Orthotropic Region in CIVA by SVTI and MHI Teams

PNNL-33733 Conventional Probe Implementation 45 Figure 5.18 Top: The Examination of an Elliptical Notch Located in the Middle of the Weld Region for the Test Block is Depicted for SVTI; Middle: The True B-Scan Response Image for the Elliptical Notch within the Weld Defined as a Single Orthotropic Region; Bottom: The Echodynamic Amplitude vs Time Response for the Elliptical Notch within the Weld Defined as a Single Orthotropic Region

PNNL-33733 Conventional Probe Implementation 46 Figure 5.19 Results from SVTI Parametric Analysis of Effect of Values for Elastic Constants C11, C22, and C33 on Relative Tip-to-Corner Response Amplitude and Arrival Time of the Tip Response Next, the weld region was defined using the built-in CIVA Ogilvy model for defining an anisotropic weld region with Ogilvy parameters. The Ogilvy model is briefly described in section 4.2, which defines crystallographic orientation as a continuously varying function of spatial position. The Ogilvy parameters were defined by SVTI based on EBSD images discussed in section 5.2.1 and as depicted in figure 5.20. The stiffness matrix for the material within the weld region is provided in figure 5.21. Here, the constants C11, C22, and C33 were assigned the value of 272 GPa. The response from the simulation of the Ogilvy weld is provided in the bottom of figure 5.20, including a label indicating the amplitude of the tip response relative to the corner response. The relative amplitude of the tip response to the corner response was much less than observed for the simulations of single orthotropic regions in figure 5.19. The amount of reduction in the relative amplitude of the tip response appeared greater than what can be explained by the slightly greater stiffness values. Therefore, most of the reduction in the tip response is attributed to the varying grain orientations defined by the Ogilvy model, resulting in less of the tip response being directed back to the probe receiver.

PNNL-33733 Conventional Probe Implementation 47 Figure 5.20 Top: Depiction of the SVTI Model of an Elliptical Notch Located inside of an Ogilvy Model Defined Weld; Bottom: True B-Scan Response of the Elliptical Notch Located inside of an Ogilvy Model Defined Weld

PNNL-33733 Conventional Probe Implementation 48 Figure 5.21 Stiffness Matrix Definition Used by SVTI in the Ogilvy Model Weld Definition in CIVA SVTI also performed simulations with two rectangular notch reflectors of equal depth (10 mm) and spanning the entire width of the test block, like the PNNL simulations described in section 5.2.1.1. Both notches were surface breaking on the ID surface of the test block and were positioned so that they were separated by 30 mm in the X direction (i.e., the axial direction).

One notch was positioned in the middle of the weld region, while the second notch was positioned outside of the weld region. In this case, the weld region is defined as a single orthotropic region using the stiffness matrix from figure 5.17.

Figure 5.22 depicts the setup and displays the results for the simulation with a True B-scan image. The image shows that the relative amplitude of the tip response to the corner response is similar to the results in figure 5.20. They contrast with the results of simulations of an elliptical notch in a single orthotropic region weld, which are shown in figure 5.18 and figure 5.19. This is unexpected since these simulations also incorporated a single orthotropic region weld definition.

These results imply that both simple variations in the geometry of the notch (elliptical vs.

rectangular) and material properties of the weld region (Ogilvy variation vs. orthotropic definition) have significant influence on the intensity of the tip response.

PNNL-33733 Conventional Probe Implementation 49 It is also interesting to observe the relative intensities of the corner responses for the two notches in figure 5.22. In this case, the corner response for the notch outside of the weld region had a greater amplitude than the corner response for the notch in the middle of the weld region.

Although one might expect that anisotropy in the weld region will cause some attenuation in the signal response, it is also possible that the notch outside of the weld region is interfering with the response from the notch inside of the weld region. This effect was also suspected in the PNNL response shown in figure 5.11.

Figure 5.22 Top: Depiction of the SVTI Model of the Examination of Two Rectangular Notches with One Notch Located outside of the Weld Region and the Other Notch Located in the Middle of a Weld Defined as a Single Orthotropic Region; Bottom: True B-Scan Response for the Simulation of Two Rectangular Notches

PNNL-33733 Conventional Probe Implementation 50 5.2.3 MHI CIVA Implementation MHI proceeded with a similar approach to SVTI for modeling weld anisotropy in CIVA. Initially, MHI implemented a weld as a single orthotropic region using the same stiffness values defined in figure 5.17. MHI then performed simulations using an Ogilvy definition of the weld region using the same stiffness values from figure 5.17 and basing the Ogilvy parameters on EBSD images for the weld cross section. Finally, MHI performed simulations of two rectangular notches with one notch positioned in the middle of a weld defined as a single orthotropic region (using stiffness values from figure 5.17) and the other notch positioned outside of the weld region. Depictions of the three simulation scenarios are shown in figure 5.23, figure 5.24, and figure 5.25, respectively, along with True B-scan response images and labels indicating the relative amplitude of tip-to-corner responses for flaws.

A summary of SVTI and MHI results is provided in table 5.3. A significant difference can be observed for the notch response for the scenario of the elliptical notch within a weld defined as a single orthotropic region. The relative tip response was much greater for SVTI in comparison to MHI. This outcome is surprising when comparing the responses for the Ogilvy weld model and the rectangular notches. In these cases, the SVTI and MHI results were similar, although they were not exactly the same. Notably, results for a rectangular notch outside of the weld were very consistent for SVTI and MHI. Thus, differences observed for notches placed inside the weld region are most likely caused by unexplained differences in the weld region definition.

PNNL-33733 Conventional Probe Implementation 51 Figure 5.23 Top: Depiction of the MHI Model of an Elliptical Notch Located inside of an Ogilvy Model Defined Weld; Bottom: True B-Scan Response of the Elliptical Notch Located inside of an Ogilvy Model Defined Weld

PNNL-33733 Conventional Probe Implementation 52 Figure 5.24 Top: Depiction of the MHI Model of an Elliptical Notch Located inside of an Ogilvy Model Defined Weld; Bottom: True B-Scan Response of the Elliptical Notch Located inside of an Ogilvy model Defined Weld

PNNL-33733 Conventional Probe Implementation 53 Figure 5.25 Top: Depiction of the MHI Model of the Examination of Two Rectangular Notches with One Notch Located outside of the Weld Region and the Other Notch Located in the Middle of a Weld Defined as a Single Orthotropic Region; Bottom:

True B-Scan Response for the Simulation of Two Rectangular Notches

PNNL-33733 Conventional Probe Implementation 54 Table 5.3 Summary of Notch Reflector Responses (Relative Amplitude of Notch Tip Response to Notch Corner) for SVTI and MHI Simulation Scenarios SVTI MHI Elliptical Notch, Single Orthotropic Region 7.3 dB 16.7 dB Elliptical Notch, Ogilvy Model 13.7 dB 11.8 dB Rectangular Notch inside Weld, Single Orthotropic Region 16.2 dB 19.1 dB Rectangular Notch outside Weld 18.0 dB 17.6 dB

PNNL-33733 Discussion and Conclusions 55 6.0 Discussion and Conclusions This section provides discussion and conclusions for the collaborative NDE modeling exercise undertaken under PIONIC and described in the earlier sections of this report. Significant themes that emerged from this effort are: 1) it is difficult to anticipate the variation in approaches to model a common scenario by multiple independent teams, and 2) implementing models of weld anisotropy present significant challenges to users of CIVA and FEM tools. The discussion focuses on these two themes in relation to questions this effort aimed to address and the implications with respect to the use of NDE modeling and simulation tools in industry. This effort sought to address two main questions:

• What degree of similarity can be expected for results generated by multiple independent teams implementing the same examination scenario with the same tool?

• How sensitive are results to different techniques for modeling anisotropy in welds?

Discussion is also provided regarding the comparison of modeling results with empirical results that were obtained in PARENT open testing.

6.1 Variation in Approaches The collaborative exercise was performed, in part, to test the assumption that if multiple teams used the same modeling tool to simulate the same scenario, they will produce identical results.

In general, the outcome of this activity demonstrated practical challenges to that assumption, mostly with respect to multiple independent teams being able to implement the same scenario.

Inconsistencies were encountered with defining scenarios because teams were either provided incomplete guidance or the guidance that was provided was not interpreted the same way by all teams. The probe calibration exercise was initially approached by distributing written specifications to teams. Although the specifications were meant to cover relevant parameters, it was understood that they were incomplete. Incomplete specifications were initially distributed to see how much variation would occur among teams if some parameter choices were left to their individual discretion. A large variation in approaches and results was observed in this case.

Somewhat unexpectedly, the variation was not caused only by the unspecified parameters, but also because several teams had different interpretations of the specifications that were provided. Following this outcome, an.xml file providing the explicit probe definition was distributed to all teams. Teams generated results that agreed qualitatively, in this case, but precise quantitative agreement was still not achieved between any two teams.

The differences observed between the teams approaches can provide insight into how seemingly minor variations in approach or interpretation of the problem can affect results. This insight serves both industry and regulators by providing information on the level of detail required for the creation of standard NDE modeling and simulation procedures and for effective regulatory review of NDE modeling and simulation results. For example, in the first case, teams were given a description of a specimen that had two SDHs. The location of the holes relative to one another and to the ID and OD surfaces was given, but the direction of scanning was not.

Some teams scanned over the hole near the OD surface first, whereas other teams did the opposite. The intention was to have teams start from the hole near the ID surface and move toward the other hole so that the hole echo signals would not interfere with one another.

However, without this explicit instruction, the scan direction was left to individual team discretion. To illustrate how standard NDE modeling and simulation procedures could be developed, JAPEIC prepared an example UT modeling and simulation standard template based

PNNL-33733 Discussion and Conclusions 56 on ISO 17640:2018 Non-destructive testing of welds Ultrasonic testing Techniques, testing levels, and assessment. This is a UT standard that describes general techniques of ultrasonic weld testing. The example UT modeling and simulation standard template is provided in appendix F.

When simulating a scenario for a relief request, a vendor will be constrained by parameters in the scan procedure. Potential variability in the probe geometry, focal laws, focal depth, scan type, probe frequency, and probe translation should be limited. However, the results from this effort show how difficult it can be to anticipate all the relevant parameters that should be defined explicitly. Even with limiting variations in the probe model, significant variations in the results among teams were observed.

The second part of the collaborative exercises focused on the implementation of a conventional UT probe and procedure (appendix D). This contrasts with the first part of the collaborative exercises, which focused on implementing a PAUT probe and procedure (appendix A). For the conventional UT probe, consistent results were more easily obtained for the calibration scenario with two SDH reflectors. This was expected because a conventional UT probe is simpler than a PAUT probe and requires fewer inputs for defining in CIVA. However, precise quantitative agreement between teams was not achieved even in this case because teams selected different simulation settings.

Conclusions

1) All model input parameters should initially be considered significant, and an effort should be made to make sure they are all identified for the examination scenario. A systematic and comprehensive approach should then be applied to justify choices of all modeling parameters, including default parameters.

2) The increased difficulty for consistent implementation of a PAUT probe versus a conventional UT probe was notable in these efforts. Advanced techniques (e.g., PAUT) introduce additional input parameters that require additional scrutiny to justify the parameter selections and to make sure the technique is implemented as intended.

6.2 Modeling Weld Anisotropy Models of weld anisotropy were also explored in conjunction with implementation of PAUT and conventional UT probes. The primary approaches to modeling weld anisotropy included Ogilvy models and weld models based on EBSD data from requisite weld samples. In addition, welds were also defined as single orthotropic regions by SVTI and MHI teams. This part of the activity was motivated to answer the two following questions:

• How much variation in results is caused by different types of weld models?

• If multiple independent teams implement the same type of weld model in CIVA, will they get consistent results?

This effort did not attempt to address the question of what type of model is best for a given type of weld material microstructure. It became apparent by other efforts (Jacob et al. 2020) that the answer to the first question is a lot. Generally, results from EBSD weld models and results from Ogilvy models or the single orthotropic region models were not comparable. The latter two models do not account for randomly varying grain orientations or assume that the orientation

PNNL-33733 Discussion and Conclusions 57 changes continuously. As a result, the Ogilvy models and single orthotropic region models capture beam redirection effects but do not capture scattering effects.

The MHI and SVTI teams implemented Ogilvy models and single orthotropic region models with the conventional UT probe. Overall, their results were consistent, but there were some unexpected differences that indicate some relevant definition parameters were not explained.

This outcome was similar to the outcome of calibration exercises for PAUT and conventional UT probe implementations because precise quantitative agreement was elusive, despite expectations. Thus, this activity reinforced conclusion 1), above.

Another observation gleaned from the anisotropic weld modeling activity was the tendency of teams to choose to implement anisotropy models built into CIVA, despite the limited capability of these models (i.e., Ogilvy model and single orthotropic region models) to capture scattering effects. The EBSD model implemented by PNNL was more cumbersome to create. Beyond the collection of the EBSD data, this approach required a process to downsample the data, create a CAD model of the weld with discretized grains, import the CAD model into CIVA, and assign Euler angles to each of the grains. This represented a significantly greater level of effort compared to implementing the already available tools in CIVA.

Conclusion

3) Many users are likely to limit themselves to options for weld anisotropy modeling that are pre-packaged in off-the-shelf tools. Although it is possible to develop custom techniques that can capture grain scattering effects in welds, many users may find them impractical to implement.

6.3 Finite Element Models JAPEIC contributed FEM capability to the collaborative exercises using ComWAVE. This allowed some comparison of CIVA with FEM. FEM was contributed to probe calibration exercises for the PAUT and conventional UT probes. For the PAUT probe implementation, 2D and 3D models were created for the calibration exercise with two SDH reflectors. While the 3D FEM simulations agreed qualitatively with the CIVA simulations (also 3D), the 2D FEM simulations did not agree qualitatively with either the 3D FEM simulations or with CIVA simulations. These exercises were based on the implementation of TRL probes, which transmitted LWs from one side and received them on the other side. Because the beam path of a TRL probe cannot be fully captured by a 2D cross section, the 2D implementation ultimately results in a distorted representation of the wave propagation. Qualitative agreement was also observed for 3D FEM simulations and CIVA simulations of the calibration exercise with two SDH reflectors. However, quantitative discrepancies existed between 3D FEM and CIVA results for both the PAUT and conventional UT probe implementations. Likely explanations for this include the inability of ComWAVE to support delay law calculations of the wave paths in 3D space and, more generally, the fact that information is lost or distorted when converting a probe model from CIVA to ComWAVE.

Similar to CIVA, ComWAVE contains built-in functions and capabilities to support modeling of anisotropic materials and standard weld geometries. If an appropriate mesh is built and appropriate material properties are allocated to each element, FEM simulation should provide realistic results. However, establishing an elaborate FEM representation can require significant effort.

PNNL-33733 Discussion and Conclusions 58 Conclusions

4) Qualitative agreement in results was observed between 3D FEM and CIVA (3D) models for relatively simple test block scenarios involving an isotropic material structure and SDH reflectors; however, precise qualitative agreement was not achieved.

5) Simplified 2D FEM models produced results that were qualitatively different from 3D FEM and CIVA models, even for simple test block scenarios.

6) Although it is possible to create detailed anisotropic welds in 3D FEM tools, they are challenging to implement, and this approach may be considered impractical for many users.

6.4 Comparisons with Empirical Data The PAUT probe implementation exercise was based on the team 114 PAUT procedure from PARENT open testing (Meyer, Holmes, and Heasler 2017). Thus, the simulation results could be compared to empirical data from that effort. Relevant empirical results are summarized in section 4.4 for LW and SW responses. Some observations of the empirical responses included the following:

• Empirical results indicated the mode-converted SW response was stronger than the direct LW response.

• The corner response was much stronger than the tip response for the direct LW examinations.

• The tip response was difficult to identify in many of the direct SW response images and appeared to either be overwhelmed by the corner response amplitude or smeared into the corner response.

• Weld scattering effects appeared to be significant in the direct LW and direct SW examination responses.

Some CIVA simulations were performed based on an isotropic weld region in section 4.4 to analyze the empirical results summarized there. Several notable differences were observed between the empirical response data and the CIVA-generated responses. These are as follows:

Mode-converted SW responses were not dominant for LW examinations in the CIVA responses. Rather, the mode-converted SW and direct LW responses were similar in amplitude.

Tip responses were of similar amplitude or greater than corner responses in the CIVA responses, especially for LW examinations.

However, the CIVA simulations were consistent with empirical results in predicting weaker tip responses for SW examinations in comparison to LW examinations.

Conclusion

7) Even if quantitative agreement with empirical results cannot be achieved, NDE modeling and simulation tools may still agree qualitatively, which allows them to be useful for such functions as optimization of examination methods.

PNNL-33733 References 59 7.0 References Dib, G., R.E. Jacob, M.R. Larche, P. Ramuhalli, M.S. Prowant, and A.A Diaz. 2018. Ultrasound Modeling and Simulation: Status Update. PNNL-29899. Pacific Northwest National Laboratory (Richland, WA).

Jacob, R.E., M.S. Prowant, C.A. Hutchinson, N. Deshmukh, and A.A Diaz. 2020. Modeling and Simulation of Austenitic Welds and Coarse-grained Specimens. PNNL-29899. Pacific Northwest National Laboratory.

Meyer, R.M., A.E. Holmes, and P.G. Heasler. 2017. Results of Open Testing for the Program to Assess the Reliability of Emerging Nondestructive Techniques. NUREG/CR-7236, PNNL-24708. (Washington, DC: U.S. Nuclear Regulatory Commission).

Ogilvy, J.A. 1985. "Computerized Ultrasonic Ray Tracing in Austenitic Steel." NDT International 18 (2): 67-77. https://doi.org/10.1016/0308-9126(85)90100-2.

Appendix A A.1 Appendix A - Team 114 TRL45 CIVAProbe and Array Settings

Appendix A A.2

Appendix A A.3

Appendix A A.4

Appendix A A.5

Appendix A A.6

Appendix B B.1 Appendix B - CIVA User InterfacesAn Example This section walks through the CIVA input menus using screenshots for data input by the Pacific Northwest National Laboratory (PNNL) team for performing the calibration exercise and interpreting the specifications provided in table 2.1. This does not provide an exhaustive capture of all the CIVA input menu options but provides the reader with an illustration of the CIVA user interface structure and context for the discussions of CIVA inputs within the report.

B.1 Specimen Settings The two main tabs under specimen settings include Geometry and Material. The shape and dimensions of the specimens are defined in the Geometry tab, as shown in figure B.1. A planar geometry is selected, which results in prompts to specify the specimen dimensions based on a Cartesian coordinate reference. Options exist to specify the specimen in 2-D and to also import drawing files of the specimen.

The Material tab includes prompts to define several material inputs and includes a library of several common predefined materials (see figure B.2). In addition to the material density, the longitudinal wave (LW) and shear wave (SW) velocities can be defined. This tab also includes the prompts to define a material as isotropic or anisotropic and to define how the material properties are distributed throughout the specimen. In this test case, the specimen is defined as isotropic. In the PNNL implementation, the predefined material 302 stainless steel was selected to define the specimen properties.

Figure B.1 Screenshot Showing the Geometry Tab Options under the Specimen Menu

Appendix B B.2 Figure B.2 Screenshot Showing the Material Tab Options under the Specimen Menu B.2 Flaw Settings Under Flaw Settings, tabs are provided to define the geometry and positioning of the flaws in the specimen. Figure B.3 shows a screenshot of the menu for the Geometry tab and two flaws with SDH geometries are created. The diameter and length of both holes are specified. The Positioning tab includes prompts for defining the position and orientation of each flaw. The position is defined by providing absolute coordinates of the center of the defect (see figure B.4).

The option also exists to define the flaw position with respect to one of the specimen surfaces.

The orientation of the flaw can also be defined as depicted in figure B.4. For SDH flaws, the orientation was specified by defining the Tilt, Skew, and Rotation parameters.

Figure B.3 Screenshot Showing the Geometry Tab Options under the Flaws Menu

Appendix B B.3 Figure B.4 Screenshot Showing the Positioning Tab Options under the Flaws Menu B.3 Probe Settings The Probe menu brings up several tabs for defining the UT probe used in the simulation. In this case, specifications for the team 114 PAUT probe are input based on the information provided in table 2.1. Figure B.5 shows the definition for a dual element linear phased array transmit-receive probe. Under the Crystal shape tab, parameters to define the configuration and dimensions of the sensing elements were input. Based on table 2.1, individual elements had a 2 mm pitch and were 15 mm long, and there was a total of 32 elements in each linear array. It is noted in table 2.1 that the individual element widths were 2 mm, which is the same as the pitch.

It was later clarified that individual element widths were 1.8 mm with a 0.2 mm gap between adjacent elements. Under the Focusing tab in figure B.6, Flat surface type was selected.

Appendix B B.4 Figure B.5 Screenshot Showing the Crystal Shape Tab Options under the Probe Menu Figure B.6 Screenshot Showing the Focusing Tab Options under the Probe Menu

Appendix B B.5 Clicking on the Wedge tab under the Probe menu brings up two more tabs labeled Geometry and Material for defining the wedge. In the Geometry tab, shown in figure B.7, several dimensions for the wedge were input. Most of this information is provided explicitly in table 2.1, and parameters not included in table 2.1 (L5 and L6) were auto-calculated after values for the other parameters were entered. Under the Material tab in figure B.8, prompts are provided for defining the wedge material. Rexolite was selected for the wedge material, and the associated properties, such as density and wave velocities, were autofilled based on the selection.

Finally, the Signal tab under the Probe menu provides prompts for defining temporal characteristics of the excitation pulses, as shown in figure B.9. A signal of 1.5 MHz and bandwidth of 55 percent was input based on the information provided in table 2.1. A Hanning signal was selected, although this was not specified in table 2.1.

Figure B.7 Screenshot Showing the Geometry Tab Options under the Wedge Tab Located in the Probe Menu

Appendix B B.6 Figure B.8. Screenshot Showing the Material Tab Options under the Wedge Tab Located in the Probe Menu

Appendix B B.7 Figure B.9 Screenshot Showing the Material Tab Options under the Wedge Tab Located in the Probe Menu B.4 Array Settings The Array menu accommodates defining the electronic firing of the elements in the PAUT probe, whereas the Probe menu mostly accommodates the definition of the physical aspects of the probe. The Array menu has several tabs, as shown in figure B.10. Under Initialization, Function was set to unisequential firing. The firing sequence was defined by selections made under the Transmission tab, which is displayed in figure B.11. In PNNLs initial implementation, focusing was swept over angles from 30 degrees to 70 degrees by selecting Multi-points focusing with five steps.

Appendix B B.8 Figure B.10 Screenshot of the Initialization Tab under the Array Settings Menu Figure B.11 Screenshot of the Transmission Tab under the Array Settings Menu B.5 Inspection Settings The Inspection Settings menu has several tabs to define the movement of a probe over the surface of the test specimen. The Configuration tab, shown in figure B.12, includes prompts to

Appendix B B.9 define several parameters of the inspection. For the simulation, PNNL selected Along X direction for the inspection plane and positive for the scanning direction. The Adapted probe option was checked, which automatically conforms the geometry of the probe wedge to match the surface contour of the test specimen so that there is not a gap. For a simple test specimen with a flat surface, this option is less important than scenarios in which the test specimen has curvature.

The Coupling Medium tab is shown in figure B.13. In this example, the properties of water were entered to define water as the coupling medium. A screenshot of the Bottom Medium tab is included to illustrate defining the medium beyond the boundaries of the test specimen. In this case, the properties of air were entered, as shown in figure B.14.

Finally, a screenshot of the Scanning tab is included to illustrate defining the method of probe scanning on the test block, as shown in figure B.15. Prompts were included to define the scanning increments and the number of scanning steps in two dimensions on the test specimen surface. In addition, users can specify if reverse scanning should be part of the scan pattern and can select the increment skip. In the PNNL simulation, a single line scan was performed in the positive X direction, so the selections regarding reverse scanning and increment skip were inconsequential.

Figure B.12 Screenshot of the Configuration Tab Settings Selected for the PNNL Simulation in the Inspections Settings Menu

Appendix B B.10 Figure B.13 Screenshot of the Coupling Medium Tab Settings Selected for the PNNL Simulation in the Inspections Settings Menu Figure B.14 Screenshot of the Bottom Medium Tab Settings Selected for the PNNL Simulation in the Inspections Settings Menu

Appendix B B.11 Figure B.15 Screenshot of the Scanning Tab Settings Selected for the PNNL Simulation in the Inspections Settings Menu B.6 Simulation Settings The Simulation Settings menu includes tabs to define the types of wave interactions that are tracked in a simulation, to define the region of the test specimen in which the modes are tracked, and to define parameters that affect the accuracy and speed of computations. Under the Initialization tab in figure B.16, users can select an option for the Control prompt. The Direct option is selected in this example, which limits computations to only including reflections contributed by interactions of the incident beam with flaws and direct skips from user-selected walls of the specimen.

Under the Interactions tab, several subtabs are available to further define the types of interactions tracked in the simulation. The Modes subtab in figure B.17 allows users to specify the types of modes to track involving interactions with a flaw. In this example, only reflected LW mode signals from the flaw were tracked. Under the Flaws subtab in figure B.18, users can define the type of scattering model for each flaw. In this example, the scattering model defined

Appendix B B.12 for each SDH was the separation of variables model. In the Sensitivity Zone subtab in figure B.19, a prompt is provided to allow users to enable a sensitivity zone. This is useful for computationally intensive models, allowing users to define limited regions over which highly accurate solutions are computed. This is helpful for complex material structures in which significant scattering occurs, such as an anisotropic weld region. For the PNNL example, a sensitivity zone was not enabled. Finally, under the Options tab, as shown in figure B.20, users can make several selections that affect the speed and accuracy of computations. For instance, a prompt for 3D simulations was provided, as well as an option for approximating the incident beam as a planar wave.

Figure B.16 Screenshot of the Initialization Tab Settings Selected for the PNNL Simulation in the Simulation Settings Menu Figure B.17 Screenshot of the Interactions Tab and Modes Subtab Settings Selected for the PNNL Simulation in the Simulation Settings Menu

Appendix B B.13 Figure B.18 Screenshot of the Interactions Tab and Flaws Subtab Settings Selected for the PNNL Simulation in the Simulation Settings Menu Figure B.19 Screenshot of the Interactions Tab and Sensitivity Zone Subtab Settings Selected for the PNNL Simulation in the Simulation Settings Menu Figure B.20 Screenshot of the Options Tab Settings Selected for the PNNL Simulation in the Simulation Settings Menu

Appendix C C.1 Appendix C - Results from Interpreted Probe Specifications Initially, teams created probe models for procedure 114-PA1 using information in table 2.1 and appendix A. The results obtained from this implementation are presented in this appendix. To help standardize the presentation of results, the results from each team are presented as a collection of response images, along with the CIVA depiction of the physical model of the test block and probe. These images, and their arrangement, are organized as depicted in figure C.1.

B-scan (amplitude as a function of scan position [mm] and time [us])

True B-scan (amplitude as a function of scan position [mm] and depth [mm])

Echodynamic response (amplitude vs. scan position [mm])

Echodynamic response (amplitude vs. time

[us])

CIVA depiction of the physical model of test block and probe (3D view)

Figure C.1 Depiction of Standard Arrangement of Response Images to Facilitate Comparison of Results C.1 PNNL Team Results for Interpreted Probe Specifications (LW Mode 30 Degrees-70 Degrees)

The results obtained for the PNNL team are displayed in figure C.2. Responses from the two SDHs were evident in the B-scan and True B-scan response images. The B-scan images also showed a more intense response from the SDH near the bottom surface. These were consistent with the echodynamic responses of amplitude versus position and amplitude versus time. A larger peak was followed by a smaller peak in the amplitude versus position response, indicating that the SDH located near the bottom surface, which was the first flaw encountered in the scan, produced a more intense response. The order of the large and small peaks was flipped in the amplitude versus time response, also indicating the SDH near the bottom surface generated the larger response.

Appendix C C.2 Figure C.2.

PNNL Team Results for Interpreted Probe Specifications C.2 SVTI Team Results for Interpreted Probe Specifications (LW Mode 45 Degrees-70 Degrees)

The results obtained for the SVTI team are displayed in figure C.3. The responses from the SDHs in the B-scan and True B-scan images were apparent but not obvious because of reflections from the back wall. In the SVTI implementation, the Half Skip option was selected for the Control prompt under the Simulation Settings menu and Initialization tab. The half skips were restricted to the back wall. Conversions from longitudinal to transverse waves were also enabled and tracked in the simulation. The Half Skip selection resulted in the spatially elongated response observed to the right of the SDH responses in the B-scan and True B-scans. The response from the back wall was also evident in the amplitude versus position echodynamic response and appears as an additional peak in the amplitude versus time echodynamic response.

The top SDH was positioned such that it was encountered first in these scans and produced the most intense response relative to the other SDH. The peaks from the SDHs were more closely spaced in the amplitude versus position response in comparison to PNNL results in figure 3.1.

Appendix C C.3 Figure C.3 SVTI Team Results for Interpreted Probe Specifications Presented in the Format of Figure C.1 C.3 MHI Team Results for Interpreted Probe Specifications (LW Mode 30 Degrees-70 Degrees)

The results obtained for the MHI team for LW simulations are displayed in figure C.4.

Responses from the two SDHs were evident in the B-scan and True B-scan response images.

The B-scan images also showed a more intense response from the SDH near the bottom surface. These were consistent with the echodynamic responses of amplitude versus position and amplitude versus time. Overall, the responses are similar to the responses observed by the PNNL team in figure C.2.

Appendix C C.4 Figure C.4 MHI Team Results for Interpreted Probe Specifications with LW Presented in the Format of Figure C.1 C.4 MHI Team Results for Interpreted Probe Specifications (SW Mode 45 Degrees-60 Degrees)

The results obtained for the MHI team for SW simulations are displayed in figure C.5.

Responses from the two SDHs were evident in the B-scan and True B-scan response images.

The B-scan images also showed a more intense response from the SDH near the bottom surface. The profiles in the echodynamic amplitude versus time profiles appeared complicated in comparison to the profile observed for the MHI LW mode simulations in figure C.4 and the PNNL LW mode simulations in figure C.2. The complication appeared to be caused by the Full Skip option selected for the Control prompt under the Initialization tab in the Simulation Settings menu. As a result, contributions following a skip off the front wall appeared in the results. The SW mode could skip off the back and front walls without losses caused by mode conversion.

Appendix C C.5 Figure C.5 MHI Team Results for Interpreted Probe Specifications with SW Presented in the Format of Figure C.1 C.5 VTT Team Results for Interpreted Probe Specifications (LW mode 45 Degrees)

The results obtained for the VTT team are displayed in figure C.6 for a linear scan performed at 45 degrees with LW modes. Responses from the two SDHs were evident in the B-scan and True B-scan response images. The B-scan images also showed a more intense response from the SDH near the bottom surface. These were consistent with the echodynamic responses of amplitude versus position and amplitude versus time. A smaller peak was followed by a larger peak in the amplitude versus position response, indicating that the SDH located near the bottom surface, which was the second flaw encountered in the scan, produced a more intense response. The sequential order of the large and small peaks was the same in the amplitude versus time response, also indicating the SDH near the bottom surface generated the larger response.

The VTT team simulated multiple other linear scan scenarios, including LW modes at 60 degrees and 70 degrees and SW modes at 45 degrees and 60 degrees. For brevity, the response images for these simulations are not provided here or discussed in detail, but the results are included in the summary provided in table 3.2.

Appendix C C.6 Figure C.6 VTT Team Results for Interpreted Probe Specifications with LW and for a Linear 45° Scan Presented in the Format of Figure C.1

Appendix D D.2 Appendix D - Summary of Input Settings for Interpreted Probe Model Activity This section summarizes the input settings for participating teams based on interpretations of the information in table 2.1. This section specifically emphasizes input settings for which there was significant variation among teams. Key distinctions among the various implementations included 1) the utilization of all the elements in the array, versus a subset of the elements, and

2) implementation as a swept sectorial scan versus multiple linear scans.

D.1 Phased Array Settings Many array settings were not explicitly defined by the information provided in table 2.1, and teams presented varied implementations of array settings to simulate linear scanning with shear wave (SW) and longitudinal wave (LW) modes at the specified angles. This section details array settings implementations for participating teams and is summarized in table D.1.

The array settings of the Pacific Northwest Laboratory (PNNL) team were covered in the CIVA user interface example in appendix B.4. Delay laws were calculated for LW modes, and Multi-points focusing was selected for focusing type with aligned focusing points. Under the Initialization tab, a unisequential function was specified.

The SVTI team selected delay law calculations for LW modes. Sectorial scanning was selected for focusing type from 45 degrees to 70 degrees over 10 steps. Under the Initialization tab, a unisequential function was specified.

The MHI team implemented a model with SW modes selected for delay law calculations.

Sectorial scanning was selected for focusing type from 45 degrees to 60 degrees in steps of 15 degrees. Under the Initialization tab, a simple electronic scanning function was selected, with a single sweep scanning type and 10 elements in each sequence.

The MHI team also implemented a model with LW modes selected for delay law calculations.

Sectorial scanning was selected for focusing type from 30 degrees to 70 degrees in steps of 10 degrees. Under the Initialization tab, a simple electronic scanning function was selected, with a single sweep scanning type and 10 elements in each sequence.

The VTT team simulated multiple linear scans for both SW and LW modes selected for delay law calculations. Under the Initialization tab, a simple electronic scanning function was selected, with a single sweep scanning type and 10 elements in each sequence. A focusing type of Depth and Direction was selected, with Geometric Point focusing at a depth of 100 mm for angles at 45 degrees, 60 degrees, and 70 degrees.

Appendix D D.3 Table D.1. Summary of Array Settings Input by Teams in Response to Specifications Provided in Table 2.1.

Team Mode and Angle Focusing and Sequencing CIVA Depiction PNNL Longitudinal (30°-70°)

Focusing Type =

Multi-points Focusing Alignment =

Aligned Points Function =

Unisequential SVTI Longitudinal (45°-70°)

Focusing Type =

Sectorial Scanning

1. of steps = 10 Initial Angle = 45° Final Angle = 70° Function =

Unisequential

Appendix D D.4 Team Mode and Angle Focusing and Sequencing CIVA Depiction MHI Shear (45°-60°)

Focusing Type =

Sectorial Scanning Initial Angle = 45° Final Angle = 60° Step size = 15° Function = Simple Electronic Scanning Type of Scanning =

Single Sweep

1. of Elements in Sequence = 10 MHI Longitudinal (30°-70°)

Focusing Type =

Sectorial Scanning 4 Steps at 10° Initial Angle = 30° Final Angle = 70° Function = Simple Electronic Scanning Type of Scanning =

Single Sweep

1. of Elements in Sequence = 10

Appendix D D.5 Team Mode and Angle Focusing and Sequencing CIVA Depiction VTT Shear (45°)

Longitudinal (45°)

Shear (60°)

Longitudinal (60°)

Longitudinal (70°)

Focusing Type =

Direction and Depth Positioning Method =

Geometrical Point Angle of Inspection Plane = 45°, 60°, 70° Depth = 100 mm Function = Simple Electronic Scanning Type of Scanning =

Single Sweep

1. of Elements in Sequence = 10 D.2 Flaw Specifications The teams were provided written flaw specifications, as shown in table 3.1. This included two SDHs located 30 mm apart along the length of the specimen, one flaw located 10 mm from the top surface, and the other flaw located 10 mm from the bottom surface. Teams from PNNL and MHI implemented the flaw specifications with the flaw closer to the bottom surface positioned such that it was encountered first by the scanning probe. The teams from SVTI and VTT implemented the specifications such that the flaw located closer to the top surface was encountered first by the scanning probe. This is depicted in the CIVA illustrations in table D.1.

D.3 Simulation Settings Simulation settings were not provided to teams as part of the specifications. Therefore, each team made their own judgments regarding selection of simulation settings. All teams selected a 3D simulation with no attenuation and an accuracy parameter set to 1. The SVTI team was the only team to enable a sensitivity zone.

The PNNL and VTT teams selected the Direct option at the Control prompt under the Initialization tab of the Simulation Settings menu. This option restricted the computation to accounting only for interactions with the flaws and direct skips for user-selected specimen surfaces.

The SVTI team selected the Half Skip option at the Control prompt under the Initialization tab of the Simulation Settings menu. This option restricted computations to accounting for contributions from three successive skips from the specimen walls and the flaw. Under the Specimens subtab under the Interactions tab in the Simulation Settings menu, computational restrictions were further defined such that only skips from the back wall were computed.

The MHI team selected the Full Skip option at the Control prompt under the Initialization tab of the Simulation Settings menu.

Appendix E E.1 Appendix E - Conventional Probe Specifications Elements Dual element Rectangular, 15 mm wide x 25 mm long Flat surface (no focusing)

Wedge Plexiglass wedge (Perspex)

Flat wedge (not contoured)

Density 1.18 g/cc L-wave 2,736 m/s T-wave 1,320 m/s No attenuation Front length (L1) = 15.53 mm Back length (L2) = 21.8 mm Width (L3) = 35.08 mm Height (L4) = 13.95 mm Roof angle = 4.5 degrees Incidence angle = 19.5 degrees Focal depth (L5) = 37.447 mm Focal distance (L6) = 5.12 mm Refraction angle (R) = 45.366 degrees (L-waves)

Signal 2 MHz center frequency 50 percent BW at 6 dB100 0-degree phase Hanning window 70 MHz sampling frequency Scan Parameters Positive scan direction Starting at 50 mm from left edge, centered on width of block Water coupling, water on bottom (1 g/cc, 1,483 m/s, no attenuation) 130 steps, 1 mm/step (in CIVA, 130 steps generates 131 A-scans)

Single B-scan only (no rastering)

Half skip L-waves No Front or Bottom specimen echoes 3D computation mode Full incident beam No mode conversions or attenuation Field and defect accuracy = 1

Appendix F F.1 Appendix F - Example UT Modeling and Simulation Standard Template Contributed by JAPEIC This document provides general ideas to perform ultrasonic testing (UT) modeling and simulation based on actual UT measurement conditions. For any ultrasonic examination, an examination procedure shall be established to describe the conditions of application of ultrasonic examination. These conditions are determined by documents such as standards and specifications. To conduct an appropriate ultrasonic examination, UT standards stipulate:

1) prerequisites that confine a concrete procedure

2) parameters and conditions that should be clarified by a procedure

3) ways to represent and evaluate examination results.

When UT modeling and simulation are performed, an examination procedure is interpreted to prepare a shape model and set modeling and simulation parameters. In this process, because some parameters are not usually considered in an examination procedure, more detailed information in addition to an examination procedure is required to complete parameter settings for UT modeling and simulation.

ISO 17640:2018 Non-destructive testing of welds Ultrasonic testing Techniques, testing levels, and assessment is a UT standard that describes general techniques of ultrasonic weld testing, using standard criteria, for the most commonly used welded joints at object temperatures in the range 0 degrees C to 60 degrees C. The specific requirements of this standard cover the test equipment, preparation, performance of testing, and reporting. The parameters specified, in particular those for the probes, are compatible with the requirements of ISO 11666 and ISO 23279. Clause 14 in ISO 17640 presents items that should be included in the test report of an ultrasonic examination. Table 1 shows how to interpret and use these items in UT modeling and simulation, and gives additional items that are not usually considered in an ultrasonic examination but required for UT modeling and simulation.

Appendix F F.2 Table 1 Treatment of UT parameters in UT modeling and simulation Item Treatment in UT modeling and simulation An identification of the object under test The material and product form This information is used to determine the material properties.

A Material properties The material properties (usually, the velocities of longitudinal and transverse waves and the density) of the tested object are required for UT simulation.

The dimensions The dimensions are required to make a shape model of the tested object.

The location of tested weld/welded joint, sketch showing geometrical configuration (if necessary)

This information is necessary to the extent required to make an appropriate shape model.

H A reference to the welding procedure, specification and heat treatment This information may be utilized if a weld model is precisely established.

N The state of manufacture This item is not necessarily required.

H The surface conditions This item may be necessary if the contact transfer loss between the surface and transducer is considered.

H The temperature of the test object This item is necessary when a material property change due to the temperature is considered.

N The contract requirements This item is not necessarily required.

N The place and date of testing This item is not necessarily required.

N An identification of test organizations and identification and certification of operator This item is not necessarily required.

The maker and type of the ultrasonic instrument with identification number, if required The simulation conditions are limited to those that can be realized by an ultrasonic instrument if it is specified.

The maker, type, nominal frequency, size of transducer and actual angle of incidence of probes used with identification number, if required This item is used to make a shape model of the transducer and determine an input signal to the transducer. The nominal frequency is usually used as the center frequency of an input signal to the transducer.

A Waveform of input signal In simulation, an input signal to the transducer should be given in the form of an exact waveform. Parameters other than the center frequency should be determined to specify the waveform of the input signal.

A Propagation distance inside transducer wedge Although this item is less cared in actual examinations, it should be clarified to make a shape model of the transducer for UT simulation.

A Material properties of transducer wedge The material properties (usually, the velocities of longitudinal and transverse waves and the

Appendix F F.3 density) of the wedge are required for UT simulation.

An identification of reference blocks used with a sketch, if necessary This information may be helpful to make an appropriate shape model of reference blocks when simulation for calibration is performed.

H The coupling medium In simulation, the coupling medium is often ignored, and the transducer is supposed to have direct contact with the scanning surface. If the effect of the coupling medium needs to be included in UT simulation, the coupling medium is presented by its thickness and material properties.

N The testing level(s) and reference to written procedure when used This item is not necessarily required. Exact UT measurement conditions should be presented in other items.

N The extent of testing The extent of testing should be indicated by the scanning area.

The location of the scanning areas The scanning areas should be explicitly presented by the exact movement of probe positions.

The reference points and details of the coordinate system used The coordinate systems are not necessarily the same between an examination and simulation, but at least the correspondence between them should be presented.

An identification of probe positions This information is used to define probe positions.

The range setting This information corresponds to the time range for simulation results to be obtained.

The method and values used for sensitivity setting Sensitivity setting can be handled in the post processing in UT simulation.

The reference levels This information is necessary for sensitivity setting when simulation for calibration is performed.

N The result of the parent material test This item is not necessarily required.

The standards for acceptance levels This information is about ways to evaluate examination results.

N The deviations from this document or contract requirements This item is not necessarily required.

The coordinates of the discontinuities with details of associated probes and corresponding probe positions This information is about ways to represent examination results.

N The maximum echo amplitudes and information, if required, on the type and size of discontinuities This information is about examination results.

Appendix F F.4 N

The lengths of discontinuities This information is about examination results.

N The results of evaluation according to specified acceptance levels This information is about examination results.

A represents the item is not included in ISO 17640 and added because of its necessity in setting parameters for UT modeling and simulation.

N represents the item is not necessarily required for UT modeling and simulation.

H represents the item is not necessarily required for basic UT modeling and simulation but may be required for higher-level UT modeling and simulation.

Pacific Northwest National Laboratory 902 Battelle Boulevard P.O. Box 999 Richland, WA 99354 1-888-375-PNNL (7665) www.pnnl.gov l www.nrc.gov