Defect detection system using finite element optimization and mesh analysis
Inventors
Hoole, S. Ratnajeevan H. • Sivasuthan, Sivamayam • Karthik, Victor Uthayakumar • Jayakumar, Paramsothy • Thyagarajan, Ravi S. • Udpa, Lalita • Rahunanthan, Arunasalam
Assignees
Michigan State University MSU • University of Toledo • United States Department of the Army
Publication Number
US-10949584-B2
Publication Date
2021-03-16
Expiration Date
2036-05-20
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
A defect detection system uses dedicated, simultaneously operating finite element optimization and mesh generation. Using an Eddy-current based probe, the system can detect and model surface and sub-surface defects.
Core Innovation
The invention describes a defect detection system that uses eddy current measurements on the surface of a metallic plate to detect and model surface and subsurface defects. This system employs parallel processing to perform finite element analysis and shape optimization continuously and iteratively. The finite element model iteratively receives self-correcting input values generated by a mesh generator module, allowing for nonstop run-time optimization that adjusts the shape of the defect model for each iteration.
The problem addressed is that existing approaches to evaluating defects in military vehicle armor or hull plates are inefficient and wasteful. Often, minor defects lead to unnecessary withdrawal from service, whereas hidden deep defects go unaddressed because they are invisible to the naked eye. Conventional mesh generators require pauses for parameter updates and manual intervention, which limits the speed and effectiveness of continuous shape optimization needed for accurate defect characterization. Thus, there is a need for a portable system capable of quickly characterizing corrosion and damage, particularly in deployed applications.
This invention solves these issues by integrating an eddy current probe with a mesh analysis system that includes main processors and dedicated GPU threads. The system generates multiple finite element mesh models in parallel, providing each to separate GPU threads which optimize the models via finite element analysis using genetic algorithms. The optimizations are continuously communicated back to the mesh generator to produce new meshes with updated model connectors, enabling continuous, parallel, and dynamic defect detection and shape characterization without the limitations of conventional interrupted mesh generation processes.
Claims Coverage
The independent claims cover an eddy current defect characterization system, a computer-implemented method for identifying defects, and a non-transitory computer-readable storage medium with instructions for characterizing defects.
Parallel finite element mesh generation and optimization using GPU threads
The system generates multiple finite element mesh models of a defect and assigns each to different GPU threads to perform finite element optimizations in parallel.
Continuous communication between GPU threads and main processors for mesh generation
The optimized finite element data from GPU threads are continuously communicated to main processors to serve as input for generating new full meshes with new model connectors, enabling nonstop iterative optimization.
Genetic algorithm optimization for objective function evaluation
Each GPU thread evaluates an objective function for the finite element mesh models using genetic algorithms to identify optimized meshes that minimize differences between measured and computed data.
Use of element-by-element Gaussian iterations for finite element optimization
The GPU threads perform matrix computations using element-by-element Gaussian iterations to optimize the finite element mesh models.
Parametric description of the defect based on optimizing difference between measured and computed profiles
The system determines a parametric description of the defect by minimizing the difference between measured electric field/voltage profiles from the eddy current probe and computed profiles from finite element optimizations.
The independent claims collectively define a system and method that perform continuous, parallel finite element mesh generation and optimization using dedicated GPU threads and genetic algorithms, with real-time communication to generate updated meshes for accurate defect characterization based on eddy current measurements.
Stated Advantages
The system provides continuous, nonstop optimization capability, overcoming limitations of conventional mesh generators that require interruption for parameter updates.
Parallel processing with dedicated GPUs accelerates the finite element optimization process, yielding practical times for field testing and defect characterization.
The approach enables detection and characterization of both surface and subsurface defects, including thin cracks, volume corrosion, and spall.
Use of open-source and commercial mesh generators adapted for continuous optimization provides flexibility and cost-effectiveness.
The use of genetic algorithms facilitates optimization without requiring gradient computations, suitable for complex defect shapes.
Documented Applications
Non-destructive evaluation (NDE) for detecting and characterizing defects such as corrosion damage and battle damage in military ground vehicle armor and hull plates.
Inspection of metal plates subjected to improvised explosive device (IED) attacks to determine if withdrawal from service is warranted.
Structural mapping and shape optimization in electromagnetic devices such as alternator rotors to achieve specific flux distributions.
General use in machine or device design requiring shape optimization through finite element analysis.
Interested in licensing this patent?