CO2 Separation Membranes Using ML Inverse Materials Design
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computational methods for materials discovery, such as High-Throughput Computational Materials Screening and Design (HCMSD), are computationally costly and lack efficient automated training and validation processes for optimizing molecules for carbon dioxide separation.
Innovation Solution
A computer-implemented method using machine learning to encode chemical molecules into SMILES representations, extract features, predict molecular properties, and generate new structures that satisfy target properties for carbon dioxide separation, utilizing an Inverse Materials Design (IMD) approach to automate the discovery process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If High-Throughput Computational Materials Screening and Design (HCMSD) method is used, then materials discovery speed is improved, but computational cost increases
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing molecular properties and features in a training dataset before the actual materials discovery process. This allows the machine learning model to be trained in advance on encoded molecular structures and their properties, so that during runtime, new molecules can be rapidly screened without performing costly ab-initio calculations for each candidate, thus improving discovery speed while managing computational cost
Solution Approach 2:
The patent uses copying by creating encoded representations (SMILES strings) of molecular structures that capture essential chemical information in a computationally efficient format. These encoded copies allow the machine learning model to process and analyze molecular structures without requiring the full computational resources needed for quantum mechanical calculations, enabling rapid screening of large numbers of candidate materials
2Productivity
If automated creation of training data is implemented, then training efficiency is improved, but data quality and relevance may deteriorate
Solution Approach 1:
The patent implements feedback by using the predicted molecular properties from the machine learning model to guide the generation of new molecular structures that satisfy target properties. The model continuously learns from the relationship between encoded molecular structures and their calculated properties, refining its predictions to improve both training efficiency and data quality in an iterative manner
Solution Approach 2:
The patent applies parameter changes by transforming molecular structures into encoded representations (SMILES strings) that preserve essential chemical information while enabling efficient computational processing. This encoding transforms complex molecular data into a format suitable for machine learning, maintaining data quality while dramatically improving training efficiency
3Reliability
If automated validation of physical performance is implemented, then discovery process completeness is improved, but computational resources required increase
Solution Approach 1:
The patent applies preliminary action by pre-calculating molecular properties such as gas permeability, selectivity, glass transition temperature, and decomposition temperature using ab-initio methods or literature data before training the machine learning model. This allows the validation process to be performed once during data preparation rather than repeatedly for each candidate, ensuring comprehensive validation while managing computational resource consumption
Solution Approach 2:
The patent uses copying by creating encoded representations of molecular structures that can be rapidly processed by the machine learning model for validation. Once a molecule is validated through the model, its encoded representation can be reused and referenced without requiring repeated full validation calculations, improving efficiency while maintaining validation completeness
Data Source
AI summary
A method and system of discovering materials for use in carbon dioxide separation includes extracting references to chemical molecules from online sources. The extracted references are encoded into chemical formulas. Molecular properties are calculated from the encoded chemical formulas. Features are extracted from the chemical formulas. Molecular properties of predicted molecular structures are predicted through a machine learning engine. The predicted molecular properties are based on the calculated molecular properties and extracted features. Target properties for predicted molecular structures are defined. Synthesized molecular structures are generated. The synthesized molecular structures include predicted molecular properties satisfying the defined target properties.