Sheet Dynamics, Ltd. Joe Kesler Tom Sharp Richard Roth Uriah - - PowerPoint PPT Presentation

Sheet Dynamics, Ltd.

Joe Kesler Tom Sharp Richard Roth Uriah Liggett 513-631-0579 jkesler@sdltd.com This presentation has been cleared for public release by the U.S. Air Force and U.S. Navy under public release numbers 88ABW-2009-2904 and YY-09-702

 Introduction  Data Organization

– Core NDT Image Management Technology – Technical Overview

 Data Mining

– Damage Trending and Reporting – NDT Coverage – Process Control – Manufacturing Process Control – Integration with Damage Analysis Packages

 Summary

 Review work being performed for Air Force and Navy

– Wanted to get more out of their inspection data

• Coverage
• Trending
• Comparison
• Improved communication with maintainers and engineers

 Modalities

– Initially looked primarily at computed radiography – Now working more with C-scan ultrasound and digital photographs as well

 What do we mean by Inspection Data

Management?

Organize Archive Mine

 Introduction  Data Organization

– Core NDT Image Management Technology – Relationship to Data Mining – Technical Overview

 Data Mining

– Damage Trending and Reporting – NDT Coverage – Process Control – Manufacturing Process Control – Integration with Damage Analysis Packages

 Summary

 “Google Maps” for NDT image data - We are

developing software to help organize, store and mine inspection data

 Core Principal: All inspection data should be

aligned to a CAD model of the inspected structure

 Core Capability: Automatically align inspection

image data to CAD models

• Acquire image (Radiograph, UT C-

scan, etc.)

• Automatically align to CAD model
• Store in database
• Repeat for entire structure

Acquired Image

 By retaining the spatial data associated with the

inspection images and indications, additional information can be mined

 Derived trends can have better than “part”

resolution

 Missing coverage is immediately apparent  Additional information available for process

control integration or other analysis tools

 Goal was to develop robust framework to organize

inspection data

 Initially reviewed a wide range of applications

– Data was not consistently aligned – Single image/Multiple images – Overlap/No overlap – Clear features/No clear features – Typically there was additional information – Wide range of “distortions” in the image data

 One approach will not handle all inspections

Incoming Data Tag Translator Alignment Method Database Data/Image Alignment Alignment Database Mining & Visualization Tools Annotation Translator

 Alignment algorithms can most broadly be

partitioned into area based and feature based algorithms

 Feature based methods extract salient features

and proceed to match those features to those associated with the model

 Area based methods operate on the image as a

whole

Apply Reference Alignment Image To Be Aligned Reference Image Aligned Wing Reference Image Alignment Correlation Based Image Registration

Extract Rivet Paterns Align Detected Rivet patterns to Rivet Pattern Model

Virtual
Camera: 
 Known
and
variable 
 pose
within
CAD 
 coordinate
system 
 CAD
Coordinate
 System
 World
Coordinate
 System
 Real
Camera: 
 Unknown
pose
in 
 either
coordinate 
 system 
 Assume
asset
and
 CAD
model
aligned
 Algorithm finds the real camera’s pose in the CAD coordinate system With knowledge of the real camera’s pose in the CAD coordinate system, it is relatively straight forward to map the image onto the model

 Introduction  Data Organization

– Core NDT Image Management Technology – Technical Overview

 Data Mining

– Damage Trending and Reporting – NDT Coverage – Process Control – Manufacturing Process Control – Integration with Damage Analysis Packages

 Summary

• 122 Tail numbers
• All annotations on

images marked on diagram

• Includes data from
• ver 6,000 CR

images

 Damage / Defects

– Location – Associated data

 Trends

– Spatial – Across Fleet – Across Time – Across Service Location – Etc.

Positive Material

• Align Data
• Extract Annotations
• Visualize Trends
• Extract Quantitative Data
• Drill Down to Original Data

Gaps in Data

• Provide real time insight into

area coverage

• Highlight areas of missing

data

• Streamline production and

maintenance practices



Portion of 4 Ultrasonic C-scans of a larger inspection



Small areas of missed coverage are much more apparent when the data is aligned

Original scan Scan after maintenance Difference between scans highlight new damage

• Comparing before and after maintenance scans can be useful for highlighting

new damage caused by process control issues

 Focus is on development of tools to improve

maintenance of composite structures

 Align digital photographs to 3-D CAD models  Export to analysis package

Take a Photograph Find damage

• n CAD model

Automatically Map Damage back to FEA model FEA Model of part

 Data organized by alignment to CAD

– Alignment is automated – Alignment to CAD enables multiple types of analysis

 Benefits

– Coverage – Trending – Comparison – Improved communication with maintainers and engineers – Export to analysis packages

 This work was supported by the US Air Force and

Navy under the following contracts:

– SBIR AF061-79 Phase II

• Contract FA8650-07-C-5210
• CTOR Gary Steffes (AFRL/RXLP)
• AF Public Release Case Number: 88ABW-2009-2904

– SBIR N07-116 Phase II

• Contract N68335-09-C-0001
• CTOR Andrew Guy (NAVAIR 4.3.3.5)
• Navy Public Release Case Number: YY-09-702