Non-dimensionalization for Additive Manufacturing Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Additive manufacturing processes face challenges in accurately modeling complex physical systems due to the need for extensive data and difficulty in capturing all physical mechanisms, which limits the application of machine learning concepts.
Innovation Solution
The use of non-dimensionalization transformations to input variables in machine learning algorithms reduces the amount of training data and variables required for accurate predictions, enabling more efficient modeling of physical systems by representing data in a non-dimensional form.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If dimensional variables are used in machine learning algorithms for additive manufacturing processes, then the model can capture detailed physical information, but the amount of training data and computational resources required increases significantly
Solution Approach 1:
The patent applies non-dimensionalization transformations to convert dimensional process variables (power, velocity, thermal diffusivity, etc.) into dimensionless parameters. This parameter transformation reduces the complexity of the input space, allowing machine learning models to achieve accurate predictions with fewer training data points while maintaining the essential physical relationships governing the additive manufacturing process
Solution Approach 2:
The patent identifies and applies non-dimensionalization specifically to the input variables of the machine learning model, while keeping the output predictions in dimensional form. This selective application of parameter transformation optimizes the training process without compromising the interpretability and practical utility of the final predictions
2Measurement precision
If dimensional variables are used in machine learning algorithms for additive manufacturing processes, then the model can represent physical systems directly, but the computational resources and time required for training increases
Solution Approach 1:
By transforming dimensional variables into dimensionless parameters, the patent reduces the effective dimensionality and complexity of the input space. This parameter transformation accelerates the machine learning training process while preserving the underlying physics, enabling faster model development and deployment in additive manufacturing applications
3Productivity
If non-dimensionalized variables are used in machine learning algorithms, then the amount of training data required is reduced, but the physical interpretability of input variables decreases
Solution Approach 1:
The patent applies non-dimensionalization only to the input variables used for training the machine learning model, while maintaining dimensional outputs for practical applications. This localized transformation improves modeling efficiency without completely sacrificing physical interpretability, as the dimensionless parameters can still be traced back to their physical origins through the transformation relationships
Solution Approach 2:
The dimensionless parameters serve as intermediaries between the physical process variables and the machine learning model. These intermediate representations capture the essential physics in a simplified form that is more efficient for training, while the model outputs remain in dimensional form for direct application to manufacturing processes
Data Source
AI summary
A method for training a machine learning engine for modeling of a physical system includes receiving process data representing measurements of a physical system. The method includes applying a transform to values of the at least two variables of the process data to generate a dimensionless parameter having a parameter value corresponding to each measurement of the physical system for the at least two variables. The method includes training the machine learning engine using a set of generated training data including the non-dimensionalized parameter, to output a prediction of a value of a physical effect of the physical system for values of the variables that are not included in the process data. The method includes controlling an additive manufacturing process for the material by setting the at least one physical property to the value of the at least one process variable during fabrication of a part.