Mixture Physical Property Prediction With Two-Stage AI Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods struggle to efficiently and accurately predict the physical properties of complex mixtures with multiple components due to the increasing complexity of component ratios, making it difficult to develop and mass-produce new mixtures for various applications.
Innovation Solution
An AI-based device and method utilizing a first AI model to extract feature data from component materials and a second AI model to predict mixture physical properties, considering interactions between components, trained in an end-to-end manner, using models like attention-based and molecular contrastive learning-based models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional non-AI technologies are used to predict physical properties of mixtures, then the prediction process is simpler, but the accuracy and efficiency deteriorate as the number of components and complexity of component ratios increase
Solution Approach 1:
The prediction system is segmented into two distinct AI models: a first AI model that extracts feature data from individual component materials, and a second AI model that predicts mixture physical properties based on the extracted features. This segmentation allows each model to specialize in specific tasks, improving overall prediction accuracy while managing complexity through functional decomposition.
Solution Approach 2:
Feature data extracted by the first AI model serves as an intermediary representation that bridges the gap between raw material information and final mixture property predictions. This intermediary feature space enables the second AI model to focus solely on mixture-level predictions, improving accuracy without requiring the model to process all raw material details directly.
2Adaptability or versatility
If the number of component materials and range of component ratios increase, then the versatility and applicability improve, but the prediction difficulty and computational complexity increase
Solution Approach 1:
The system dynamically handles varying numbers of component materials and their ratios through the sequential AI model processing. The first AI model can process different numbers of component materials, and the second AI model adapts to different mixture compositions, enabling the system to maintain high prediction accuracy across a wide range of mixture types without requiring manual reconfiguration.
Solution Approach 2:
The system transforms the complex multi-dimensional problem of predicting mixture properties for varying compositions into a simplified feature extraction and prediction pipeline. By extracting features in an intermediate dimension and then performing predictions based on these features, the system reduces the computational difficulty of handling high-dimensional composition spaces.
3Productivity
If conventional prediction methods are used, then the computational resources required are lower, but the time required for prediction increases and productivity decreases
Solution Approach 1:
The first AI model performs preliminary extraction of feature data from component materials before the second AI model performs the final prediction. This preliminary action prepares the input data in advance, allowing the second model to perform faster predictions since it only needs to process the extracted features rather than raw material information, thereby improving overall prediction speed.
Solution Approach 2:
The sequential AI model architecture enables continuous and efficient prediction by maintaining the useful action of feature extraction and property prediction in a streamlined pipeline. The models are trained in an end-to-end manner, ensuring that the prediction process flows continuously from input to output without interruption, maximizing productivity while managing computational resources efficiently.
Data Source
AI summary
A device for predicting the physical properties of a mixture including a plurality of component materials is disclosed. The device may comprise a memory in which a first AI model trained to output first feature data of material information, and a second AI model trained to output physical property prediction information of the first feature data are stored, and a processor for executing the first AI model and the second AI model, wherein the processor may input, into the first AI model, material information of each of the plurality of materials to acquire first feature data of each of the plurality of materials, and input the first feature data into the second AI model to acquire physical property prediction information of the mixture.