Selective Laser Melting Parameter Prediction Using Physics-Based ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Additive manufacturing processes face challenges in predicting sensor values accurately due to complex print parameters and evolving physics at different scales, making it costly to use physical simulations for quality metric prediction.
Innovation Solution
A method that involves identifying machine process parameters, generating physics-based features, and training a machine-learning software model using real-world sensor readings and virtual data to predict sensor behavior, allowing for accurate and efficient predictions without extensive calibration or simulation analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physical simulations are used to predict quality metrics, then prediction accuracy is improved, but computational cost and time increase
Solution Approach 1:
The patent creates a virtual copy of the sensor behavior through machine learning models trained on simulation data. Instead of running full physical simulations to predict sensor values, the system uses pre-trained ML models that replicate sensor responses based on process parameters, dramatically reducing computation time while maintaining prediction accuracy.
Solution Approach 2:
The system performs preliminary actions by pre-training machine learning models using extensive simulation data before actual manufacturing operations. This pre-computation allows the models to quickly predict sensor values during real-time manufacturing without requiring ongoing complex simulations, resolving the time-accuracy tradeoff.
2Manufacturing precision
If the number of print parameters is increased to capture process complexity, then manufacturing precision is improved, but device complexity increases
Solution Approach 1:
The patent introduces machine learning models as intermediary components that bridge the gap between process parameters and sensor outcomes. These models learn complex relationships from simulation data and provide simplified predictions, allowing the system to account for multiple parameters without directly managing their complexity in real-time operations.
Solution Approach 2:
The system transforms the complex multi-parameter manufacturing problem into a different parameter space by using ML model predictions. Instead of directly analyzing numerous interdependent print parameters, the system uses learned parameter transformations that capture essential relationships, simplifying the complexity while maintaining precision.
3Measurement precision
If extensive calibration and simulation analysis are performed, then sensor prediction accuracy is improved, but productivity decreases
Solution Approach 1:
The system performs extensive calibration and simulation analysis in advance to train machine learning models. This preliminary action transfers the computational burden to an offline phase, enabling rapid predictions during actual manufacturing operations without sacrificing accuracy, thus resolving the productivity-precision conflict.
Solution Approach 2:
The patent creates simplified copies of complex sensor behaviors through ML models. These models capture the essence of extensive calibration results in compact form, allowing fast predictions that replicate the accuracy of extensive analysis without requiring the full computational effort during production.
Data Source
AI summary
A method includes identifying machine process parameters for an additive manufacturing process to produce a part, providing a real-world sensor to sense a characteristic associated with a real-world version of the additive manufacturing process, receiving sensor readings from the real-world sensor while the machine is performing the real-world version of the additive manufacturing process, generating, with a computer-based processor, physics-based features associated with the additive manufacturing process, and training a machine-learning software model based at least in part on the machine process parameters, the sensor readings, and the physics-based features to predict a behavior of the real-world sensor.