Additive Manufacturing Thermal Mapping for Voxel-Level Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing additive manufacturing techniques face challenges in predicting transient thermal behavior due to the lack of quantitative knowledge about material behavior, leading to inadequate thermal imaging resolution and inaccurate voxel-level thermal prediction, which affects print quality and control.
Innovation Solution
A deep neural network-based approach that utilizes low-resolution thermal sensing data and contone maps to enhance thermal image resolution up to voxel-level precision, enabling online in-situ thermal prediction and closed-loop feedback control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If low-resolution thermal sensing is used, then device complexity and cost are reduced, but thermal image resolution and measurement precision are insufficient
Solution Approach 1:
A neural network model acts as an intermediary between low-resolution thermal sensing data and high-resolution thermal predictions. The model takes low-resolution thermal images and contone maps as input and generates high-resolution thermal predictions, effectively bridging the resolution gap without requiring expensive high-resolution thermal cameras.
Solution Approach 2:
The system creates a high-resolution copy of thermal data from low-resolution measurements. By using the neural network to predict and generate detailed thermal information at voxel-level resolution, the system produces a virtual high-resolution thermal image that mirrors what a physical high-resolution sensor would capture.
2Adaptability or versatility
If quantitative material behavior knowledge is lacking, then material selection flexibility is maintained, but thermal prediction accuracy deteriorates
Solution Approach 1:
The system incorporates feedback mechanisms where thermal predictions are continuously refined. The neural network learns from the relationship between contone maps (process parameters) and thermal outcomes, using this feedback to improve prediction accuracy for different materials without requiring explicit quantitative material models.
Solution Approach 2:
The approach changes from relying on fixed material properties to dynamically adjusting thermal predictions based on process parameters captured in contone maps. By focusing on controllable process parameters rather than inherent material properties, the system maintains versatility across materials while achieving accurate predictions.
3Measurement precision
If high-resolution thermal sensing is implemented, then voxel-level thermal prediction precision is improved, but device cost and complexity increase significantly
Solution Approach 1:
The system replaces the mechanical/optical approach of using high-resolution thermal sensors with a computational approach. Instead of physically resolving fine thermal details through high-resolution sensing, the system uses neural networks to computationally infer and generate high-resolution thermal data from low-resolution measurements.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Examples of methods for thermal mapping by an electronic device are described herein. In some examples, a map is obtained. In some examples, a first thermal image at a first resolution is obtained. In some examples, a neural network is used to determine a second thermal image at a second resolution based on the map and the first thermal image. The second resolution is greater than the first resolution in some examples.