Synthetic CAD-Trained CT Reconstruction for Artifact-Free Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing CT reconstruction algorithms struggle with beam hardening artifacts, particularly in metal additive manufacturing, leading to inaccurate defect detection and increased scan times, and current deep learning methods require costly, labor-intensive labeled data sets.
Innovation Solution
A system and method using CAD models and synthetic data to simulate artifacts, training a deep learning model to reduce beam hardening and detector noise, enabling artifact-free CT reconstructions without relying on real data sets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If physical filters are used to filter out low energies of the x-ray spectrum, then beam hardening artifacts are reduced, but the overall flux of the source is lowered and scan times increase
Solution Approach 1:
The patent replaces the physical filter mechanism with a computational algorithm that simulates and corrects beam hardening effects digitally. The system uses a neural network trained on synthetic data to predict and remove artifacts without physical intervention, thereby maintaining scan speed while reducing artifacts.
Solution Approach 2:
The patent creates synthetic CT data copies from CAD models to train the neural network. These synthetic datasets replicate beam hardening artifacts and ground truth reconstructions, allowing the model to learn artifact removal patterns without requiring physical filtering or real labeled data.
2Measurement precision
If real labeled data sets are used to train deep learning models, then artifact reduction accuracy is improved, but the process becomes costly and labor-intensive
Solution Approach 1:
The patent generates synthetic CT data copies from CAD models, creating unlimited training data without physical scanning or manual labeling. The synthetic datasets include realistic beam hardening artifacts and corresponding ground truth reconstructions, eliminating the need for costly real labeled data collection.
Solution Approach 2:
The patent performs preliminary simulation of CT scans and artifact generation before training the neural network. By pre-computing synthetic datasets with known ground truth from CAD models, the system prepares training data in advance, avoiding the need for time-consuming real data collection and manual annotation.
3Productivity
If standard CT reconstruction algorithms are used, then processing speed is maintained, but beam hardening artifacts cause inaccurate defect detection
Solution Approach 1:
The patent replaces standard reconstruction algorithms with a neural network-based system that processes CT data. The neural network learns non-linear mappings from artifact-ridden projections to artifact-free reconstructions, achieving both speed and accuracy by leveraging synthetic training data to capture complex artifact patterns.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Provides high-quality CT reconstructions with reduced artifacts and improved defect detection, lowering scan times and costs by leveraging synthetic training data and CAD models.
Implementation Method 1
the lower-energy photons in a polychromatic beam are absorbed more easily than the higher-energy photons, which harden the x-ray spectrum as it passes through the object. This effect, called 'beam hardening', breaks the fundamental assumption of linearity implicit in common CT reconstruction algorithms
Data Source
AI summary
Nondestructive evaluation (NDE) of objects can elucidate impacts of various process parameters and qualification of the object. Computed tomography (CT) enables rapid NDE and characterization of objects. However, CT presents challenges because of artifacts produced by standard reconstruction algorithms. Beam-hardening artifacts especially complicate and adversely impact the process of detecting defects. By leveraging computer-aided design (CAD) models, CT simulations, and a deep-neutral network high-quality CT reconstructions that are affected by noise and beam-hardening can be simulated and used to improve reconstructions. The systems and methods of the present disclosure can significantly improve the reconstruction quality, thereby enabling better detection of defects compared with the state of the art.