Catalyst Relaxed Energy Prediction Using Two-Stage Graph Neural Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for predicting relaxed energy of catalyst systems using quantum mechanics are computationally complex and time-consuming, hindering the efficient selection of appropriate catalysts for chemical reactions.
Innovation Solution
A two-stage training process using graph neural networks to construct a catalyst system relaxed energy prediction model, first training a pre-trained energy prediction model and then refining it with a second training data set to achieve high precision and speed in predicting relaxed energy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If quantum mechanics method is used to predict relaxed energy of catalyst system, then prediction precision is improved, but calculation complexity and time consumption increase significantly
Solution Approach 1:
The patent creates a deep learning model that copies the predictive capabilities of quantum mechanics calculations. The model is trained on quantum mechanics calculation data to learn the mapping between catalyst structure and relaxed energy, then uses this learned knowledge to make predictions without performing actual quantum mechanics calculations, thus achieving similar precision with much lower computational complexity
Solution Approach 2:
The patent substitutes the quantum mechanics calculation system with a deep learning prediction system. Instead of using complex quantum mechanical equations and iterative calculations, the patent employs a trained neural network model that directly predicts relaxed energy from structural features, replacing the mechanical calculation process with an information processing approach
2Measurement precision
If quantum mechanics method is used to predict relaxed energy of catalyst system, then prediction precision is improved, but research and development efficiency deteriorates
Solution Approach 1:
The patent performs preliminary action by pre-training the deep learning model on a large dataset of quantum mechanics calculations before actual catalyst screening. This preliminary training phase captures the essential structure-energy relationships, enabling rapid predictions during subsequent catalyst evaluation without repeating expensive quantum mechanics calculations, thus improving overall R&D efficiency
Solution Approach 2:
The patent creates a computational copy of quantum mechanics prediction capabilities through the deep learning model. This copy can rapidly predict relaxed energy for new catalyst structures without performing actual quantum mechanics calculations, maintaining prediction precision while dramatically speeding up the evaluation process for catalyst discovery
3Measurement precision
If conventional relaxed energy prediction method is used, then calculation accuracy is maintained, but calculation amount increases hugely
Solution Approach 1:
The patent extracts the essential predictive knowledge from quantum mechanics calculations by training the deep learning model on calculation data. The model learns to extract relevant structural features and their relationship with relaxed energy, then uses this extracted knowledge to make predictions without performing full quantum mechanics calculations, thus reducing calculation amount while maintaining accuracy
Solution Approach 2:
The patent changes the computational parameters from quantum mechanical variables (electron densities, wave functions) to structural descriptors (atomic positions, bond lengths, angles). This parameter transformation allows the system to predict relaxed energy using simple structural inputs processed by the neural network, dramatically reducing the computational burden while preserving prediction accuracy
Data Source
AI summary
A method for constructing a catalyst system relaxed energy prediction model is performed by a computing device, and the method includes: obtaining a first training data set including a plurality of pieces of first training data, each piece including a first training sample and a corresponding first sample label; training a catalyst system energy prediction model by using the first training data set, to obtain a pre-trained catalyst system energy prediction model; constructing a catalyst system relaxed energy initial prediction model based on the pre-trained catalyst system energy prediction model; obtaining a second training data set including a plurality of pieces of second training data, each piece including a second training sample and a corresponding second sample label; and training the catalyst system relaxed energy initial prediction model by using the second training data set, to obtain a catalyst system relaxed energy prediction model.