Small Business Innovation Research/Small Business Tech Transfer
Model Inversion
Project Description
Forward and inverse modeling of nondestructive evaluation (NDE) are key needs for optimized, quantitative NDE. Some forward modeling tools exist commercially, but inverse modeling remains a topic mostly in low TRL research. The ill-posed nature of the problem in general requires data-driven methods that are computationally intensive and highly problem specific. We propose two innovations to provide significant improvement to inversion: First, modern classifier-based data reduction, to prepare data for the second innovation, Kriging methods for a generalized NDE inversion approach. Experimental data and/or modeled data can be used to define known points in a multi-dimensional solution space, and Kriging methods can provide efficient interpolation in this space to invert new NDE data. TRI/Austin, AeroMatter, and Computational Tools are teaming to develop and demonstrate inversion of ultrasonic NDE on composite structures for quantitative damage assessment. The proposed innovations to be provided are: 1. Combined use of experiment and model data for developing the known solution points. 2. Dimensional reduction of the data for efficient inversion using state of the art classifier techniques. 3. Kriging methods for interpolation of new NDE data in the solution space. 4. High performance computing (HPC) technologies to speed data reduction and Kriging results. The significance of the innovations are that this approach offers an ability to invert NDE data using known or truth data from experiment and/or models, and is readily adapted to high performance computing technologies for practical use.The NDEInverter will work with the rest of the tools in TRI/Austin's NDEToolbox. NDEToolbox serves as a foundational, evolving platform for the management and analysis of NDE data, interaction with NDE models, and risk / reliability prediction.
More »
Anticipated Benefits
The two main categories of NASA applications are in support of in house R&D, and in the inspection of structures for NASA applications by NASA or its suppliers. The NDEInverter software will provide a set of tools or workbox enabling the power user to setup an inversion solver for a problem of interest. One can envision using development specimens, calibration specimens, and forward models to populate a "truth" table for the specific problem of interest. This would be followed by a defined process for data reduction using classifier and GLM techniques contained in the NDEInverter software using the available open source libraries. Finally, new data would be input to the software, and the Kriging algorithm from AeroMatter would be executed seamlessly. For low rate or one-off production, or R&D applications typical of NASA use this will be a less formal process, and many options or even source code in Python may be exposed. In a higher rate production environment this may be sold as a software as a service model, with an initial investment in the setup of the specifics for the problem, followed by a license or even a pay as you go model using cloud services to take full advantage of HPC. We do anticipate the NDEInverter would be one of the family of tools in the TRI/Austin NDEToolbox software.
The inversion software has direct applicability to commercial and military aerospace manufacturers who build and inspect large composite structures with stringent defect requirements. Even sub-rejectable defects are of interest in production as they are often indicative of a process variation issue, and better characterization of defects allows better response by the production team. After parts are manufactured, defective parts that are expensive are referred to a materials or manufacturing review board (MRB) for disposition. These decisions could be made better with improved understanding of the characteristics of a defect. Finally, components are inspected inservice. Similar to the MRB problem, better inspection and repair/or replacement decisions can be made by improved characterization of defects. TRI/Austin has an excellent record of SBIR transition success and works to begin the transition to market early on in the SBIR effort. Recently, TRI/Austin was awarded a Tibbetts award, which honors a small number of SBIR/STTR program participants and supports that have created a significant economic or social impact through the use of SBIR/STTR funding. More »
The inversion software has direct applicability to commercial and military aerospace manufacturers who build and inspect large composite structures with stringent defect requirements. Even sub-rejectable defects are of interest in production as they are often indicative of a process variation issue, and better characterization of defects allows better response by the production team. After parts are manufactured, defective parts that are expensive are referred to a materials or manufacturing review board (MRB) for disposition. These decisions could be made better with improved understanding of the characteristics of a defect. Finally, components are inspected inservice. Similar to the MRB problem, better inspection and repair/or replacement decisions can be made by improved characterization of defects. TRI/Austin has an excellent record of SBIR transition success and works to begin the transition to market early on in the SBIR effort. Recently, TRI/Austin was awarded a Tibbetts award, which honors a small number of SBIR/STTR program participants and supports that have created a significant economic or social impact through the use of SBIR/STTR funding. More »
Project Library
Primary U.S. Work Locations and Key Partners
| Organizations Performing Work | Role | Type | Location |
|---|---|---|---|
| Texas Research Institute Austin, Inc. | Lead Organization | Industry | Austin, Texas |
Langley Research Center
(LaRC)
|
Supporting Organization | NASA Center | Hampton, Virginia |
Primary U.S. Work Locations
-
Texas
-
Virginia
Suggest an Edit
Recommend changes and additions to this project record.