Machine Learning NDE Data Fusion for Additive Manufacturing Stress Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current non-destructive evaluation (NDE) methods, particularly for additive manufactured components, are computationally intensive and time-consuming due to the complexity of analyzing stress concentrations around defects using finite element analysis, limiting their efficiency and practicality for real-time inspection.
Innovation Solution
Employing a machine learning data fusion module to integrate and process image and modeling data from multiple sources, predicting stress concentrations and identifying critical defects, thereby enhancing the efficiency and speed of NDE by transforming raw data into tabular and visual representations for easier interpretation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If numerical simulations (finite element method) are used to evaluate stress concentrations around defects, then measurement precision and reliability are improved, but productivity and processing speed deteriorate due to computationally intensive calculations
Solution Approach 1:
The patent creates a simplified digital copy or surrogate model of the complex finite element simulation system. This surrogate model captures the essential stress concentration behavior around defects but computes results much faster, replacing the need for full numerical simulations while maintaining adequate measurement precision for defect evaluation
Solution Approach 2:
The patent transforms the problem from solving complex differential equations with fine mesh discretization to using a simplified model with coarser parameters. By changing the mathematical representation from high-fidelity numerical simulation to a reduced-order model, the computation time is dramatically reduced while preserving the key stress concentration information needed for defect detection
2Reliability
If numerical simulations are used to evaluate stress concentrations, then reliability of NDE results is improved, but loss of time increases due to demanding computational requirements
Solution Approach 1:
The patent performs preliminary actions by pre-computing or pre-calibrating the simplified surrogate model using finite element analysis data. Once the surrogate model is trained or calibrated in advance, it can rapidly evaluate stress concentrations without requiring full numerical simulations during actual inspection, thus reducing computation time while maintaining reliability
Solution Approach 2:
The patent creates a simplified digital copy or surrogate model of the complex finite element simulation system. This surrogate model captures the essential stress concentration behavior around defects but computes results much faster, replacing the need for full numerical simulations while maintaining adequate measurement precision for defect evaluation
3Measurement precision
If mesh is modified to resolve small and intricately shaped defects, then measurement precision is improved, but device complexity and computational resources increase
Solution Approach 1:
Instead of creating a uniformly fine mesh throughout the entire component, the patent applies the simplified surrogate model locally at defect locations identified from initial scanning. This approach achieves high measurement precision for defect stress evaluation without the computational complexity of a globally refined mesh, as the simplified model only needs to evaluate stress at specific defect sites rather than throughout the entire geometry
Data Source
AI summary
A method of performing non-destructive evaluation (NDE) of a part or assembly may include receiving image and modeling data for the part or assembly from multiple sources, employing a machine learning model to fuse the image and modeling data into tabular data and fused image data associating predicted stress values with respective locations of the part or assembly, and linking the tabular data and fused image data together for display.