Layer Image Distortion Prediction for Additive Manufacturing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Additive manufacturing experiences distortion issues during the build process, leading to dimensional inaccuracies and re-coater interference, which conventional simulations fail to prevent effectively.
Innovation Solution
A method utilizing image analysis and machine learning models, specifically recurrent neural networks, to predict distortions in real-time by comparing simulated distortions with actual layer images, allowing for predictive adjustments in the fabrication process, such as altering laser power or skipping layers, to minimize distortion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional simulations are used to predict distortion, then computational load is reduced, but prediction accuracy is insufficient leading to dimensional inaccuracies
Solution Approach 1:
The system segments the distortion prediction task into multiple components: image analysis module extracts surface features from camera images, build simulation module generates theoretical distortion data, and machine learning module integrates both inputs. This segmentation allows each module to specialize in specific functions, improving overall prediction accuracy while managing computational complexity through modular architecture.
Solution Approach 2:
The machine learning model serves as an intermediary that bridges image analysis data and build simulation data. It learns the complex relationships between observed surface features, simulated distortions, and actual dimensional outcomes, enabling accurate predictions without requiring the entire simulation pipeline to run at full computational cost for every prediction.
2Manufacturing precision
If real-time image analysis and machine learning are implemented, then distortion prediction accuracy improves, but computational resources and processing time increase
Solution Approach 1:
The system performs preliminary actions by capturing images of each layer during the additive manufacturing process and running image analysis immediately. The machine learning model has been pre-trained on historical data, allowing it to quickly process new inputs. Build simulations are also performed in advance to generate reference distortion data, enabling real-time predictions without excessive computational delay.
Solution Approach 2:
The system replaces traditional mechanical measurement methods with optical image analysis and virtual build simulations. Instead of physically measuring each layer to detect distortion, the system uses camera images and computational models, significantly reducing measurement time and enabling real-time monitoring and prediction.
3Reliability
If comprehensive build data is collected and analyzed, then prediction reliability improves, but data processing complexity increases
Solution Approach 1:
The machine learning model serves multiple functions: it processes image analysis data, integrates build simulation results, learns from historical build data, and generates distortion predictions. This multi-functionality consolidates complex data processing into a single unified system, improving reliability through comprehensive data utilization while managing complexity through a unified computational framework.
Solution Approach 2:
The system implements feedback by continuously comparing predicted distortion with actual measured distortion from image analysis. The machine learning model uses historical build data and outcomes to refine its predictions, learning from past performance. This feedback mechanism improves prediction reliability over time while the systematic approach to data collection and analysis manages processing complexity through structured methodologies.
Data Source
AI summary
Examples described herein provide a method that includes performing an image analysis on an image of a layer of an object being manufactured by an additive manufacturing system to identify an exposed surface in the image of the layer. The method further includes performing a build simulation to generate a simulated distortion for the layer. The method further includes evaluating build data to determining a value of an influencing factor for the layer. The method further includes predicting at least one of a predicted distortion or a predicted re-coater interference for a next layer, using a machine learning model, based at least in part on the image analysis, the build simulation, and the build data. The method further includes implementing an action, based at least in part on the at least one of the predicted distortion or the predicted re-coater interference, to alter fabrication of the next layer.