DeepM&Mnet Neural Network for Nuclear-Thermal Coupling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current nuclear-thermal coupling calculation methods are limited by slow convergence speed, low numerical accuracy, and dependency on specific neutron physics and thermal-hydraulic calculation programs, making them inefficient for solving new problems without extensive databases and long pre-training processes.
Innovation Solution
The implementation of a nuclear-thermal coupling method based on a deep multi-physics and multi-scale neural network (DeepM&Mnet), which uses numerical solvers for material temperature and neutron physics fields, and a deep operator network (DeepONet) to achieve fast and accurate calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If existing coupling iterative calculation methods are used, then nuclear-thermal coupling calculation can be performed, but the convergence speed is slow
Solution Approach 1:
The patent replaces traditional mechanical iterative calculation methods with a neural network-based computational model. The DeepM&Mnet neural network learns the complex coupling relationships between neutron physics and thermal-hydraulic fields during training, then directly predicts results for new scenarios without requiring iterative convergence, thus substituting the iterative mechanical process with a direct neural network inference process.
Solution Approach 2:
The patent performs preliminary training of the neural network using extensive coupling calculation data before actual application. During this pre-training phase, the network learns the underlying physical relationships and coupling mechanisms, so that when deployed, it can directly provide accurate results without requiring iterative convergence for each new calculation, effectively performing the learning work in advance.
2Measurement precision
If existing coupling iterative calculation methods are used, then nuclear-thermal coupling calculation can be performed, but the numerical accuracy is low
Solution Approach 1:
The patent changes the fundamental parameters and approach of the calculation system by transitioning from traditional iterative numerical methods to a neural network-based approach. The neural network is trained to directly output high-accuracy results for temperature and neutron flux fields, bypassing the accuracy limitations and slow convergence of traditional iterative methods while maintaining computational efficiency.
3Reliability
If DeepONet pre-training process is used, then the neural network can be trained, but the pre-training process is long
Solution Approach 1:
The patent performs preliminary training of the neural network using extensive coupling calculation data before actual application. During this pre-training phase, the network learns the underlying physical relationships and coupling mechanisms, so that when deployed, it can directly provide accurate results without requiring iterative convergence for each new calculation, effectively performing the learning work in advance.
4Adaptability or versatility
If traditional iterative methods are used, then calculation can be performed, but dependency on existing neutron physics and thermal-hydraulic programs is high
Solution Approach 1:
The patent creates a universal neural network model that can handle various nuclear-thermal coupling problems through a single unified framework. The DeepM&Mnet is trained on diverse coupling scenarios and can generalize to new problems without requiring specific iterative programs, making the system adaptable to different geometries, materials, and operating conditions while reducing dependency on specialized traditional codes.
Data Source
AI summary
A method for implementing nuclear-thermal coupling based on the DeepM&Mnet neural network is disclosed; the proposed nuclear-thermal coupling implementation method comprises the following steps; first construct DeepM&Mnet neural network inclusive of numerical solvers for the material temperature and neutron physics fields; then the DeepM&Mnet neural network is trained by utilizing physics constraints of the numerical solvers for the material temperature and neutron physics fields; finally, the formation of the network loss function is adjusted based on the training results to achieve nuclear-thermal coupling simulation; according to the method, a numerical solver for material temperature and neutron physics fields, or a deep operator network (DeepONet) is applied for fitting a numerical solution process, and a DeepM&Mnet is constructed on the basis.