Liquid metal electromagnetic pump multi-physics simulation method based on stochastic neural network
By constructing divergence-free stochastic neural networks and standard stochastic neural networks, and combining them with the least squares method, the problems of high computational cost and low accuracy in multiphysics simulation of liquid metal electromagnetic pumps are solved, achieving efficient and accurate simulation results. This method is applicable to the simulation of liquid metal electromagnetic pumps in complex geometric regions.
CN121189180BActive Publication Date: 2026-07-21XI AN JIAOTONG UNIV
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- XI AN JIAOTONG UNIV
- Filing Date
- 2025-09-26
- Publication Date
- 2026-07-21
Smart Images
Figure CN121189180B_ABST
Abstract
A kind of liquid metal electromagnetic pump multi-physics field simulation method based on random neural network, constructs two structure consistent divergence-free random neural networks and a standard random neural network, in turn for simulating magnetic induction intensity, liquid metal velocity field and liquid metal pressure in electromagnetic pump at any time;Obtain the geometric information of electromagnetic pump and the material information of liquid metal, determine the boundary condition;In the boundary and space-time calculation region, obtain training points according to uniform distribution random sampling;Each random neural network is brought into electromagnetic pump mathematical model, and the corresponding linear equation set is obtained, and the simulation result is obtained by solving;Or, each random neural network is brought into linearized electromagnetic pump mathematical model, and the corresponding linear equation set is obtained, and the weight of the output layer of each random neural network is updated until nonlinear iteration converges, and the multi-physics field is predicted using the random neural network at convergence.
Need to check novelty before this filing date? Find Prior Art