Droplet Shape Estimation for Additive Manufacturing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Additive manufacturing (AM) processes face significant limitations in predicting the shape of printed parts due to manufacturing imperfections such as variability in droplet deposition location, frequency, and temperature, leading to uncertainties in the as-printed shape, which can result in porosity, surface roughness, and mechanical property degradation.
Innovation Solution
A computationally efficient reduced-order model using hybrid machine learning, specifically training an Artificial Neural Network (ANN) to predict the shape of solidified droplets on substrates of arbitrary geometry, significantly reducing computational time from hours to milliseconds, allowing for rapid estimation of part geometry and accounting for manufacturing uncertainties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional computational models are used to predict droplet shape and part geometry, then accuracy in capturing manufacturing imperfections is improved, but computational time increases from milliseconds to hours
Solution Approach 1:
The patent applies preliminary action by pre-training an Artificial Neural Network on comprehensive datasets that capture manufacturing imperfections such as droplet deformation, coalescence, and substrate interactions. This pre-training enables the model to make accurate predictions during actual manufacturing without requiring time-consuming computational resources at runtime, thus resolving the contradiction between prediction accuracy and computational time.
Solution Approach 2:
The patent replaces traditional mechanics-based computational models with a machine learning-based neural network model. This substitution allows the system to capture complex manufacturing imperfections and droplet behaviors through learned patterns rather than explicit physical simulations, achieving both high accuracy and fast computational speed simultaneously.
2Measurement precision
If detailed analysis of multiple scanned prints is performed to determine manufacturing errors, then measurement precision is improved, but device complexity and computational requirements increase
Solution Approach 1:
The patent introduces an intermediary neural network model that processes multiple scanned print data. This intermediary model learns to extract manufacturing error patterns from the scanned prints and provides simplified yet accurate predictions, reducing the complexity of direct detailed analysis while maintaining measurement precision.
3Reliability
If hybrid machine learning models are used to account for substrate geometry and manufacturing uncertainties, then reliability of prediction is improved, but model complexity increases
Solution Approach 1:
The patent applies parameter changes by training the neural network to adapt to different substrate geometries and manufacturing conditions through varied input parameters. The model learns to adjust its predictions based on substrate curvature, material properties, and process parameters, achieving high reliability across diverse scenarios without requiring separate complex models for each condition.
Data Source
AI summary
A plurality of scanned prints of a product part and a scan-path are received. A shape of a minimum printable feature of the product part is determined by analyzing the respective prints in a scan-path representation. A manufacturing error of the minimum printable feature is determined based on the analysis. A manufacturing error of a shape of the part is determined based on the determined manufacturing error of the minimum printable feature. An estimated manufactured shape of the part is produced based on the determined manufacturing error of the part.