Self-adaptive efficient simulation method and system for electric propulsion plasma oscillation

Abstract

The invention provides a self-adaptive efficient simulation method and system for electric propulsion plasma oscillation, and relates to the technical field of electric propulsion plasma oscillation simulation. According to the method, a high-precision electric propulsion plasma numerical simulation model is adopted to ensure the precision of a simulation result; according to the method, the calculation tasks of the simulation model are decomposed through various adjustment parameters, that is, all the calculation tasks are completed in a parallel calculation mode, and therefore the simulation efficiency can be effectively improved. Moreover, when the simulation model is subjected to simulation solving, the electromagnetic field solving model is solved by adopting a multi-grid method, and the simulation calculation efficiency can be further improved based on the solving mode. Besides, the target neural network model is utilized to process the simulation result so as to update all the adjustment parameters in the process, so that a plurality of adjustment parameters used in the simulation process are continuously and adaptively optimized, and the technical effect of considering the simulation efficiency and precision of the electric propulsion plasma oscillation simulation model is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of electric propulsion plasma oscillation simulation, and in particular to an electric propulsion plasma oscillation adaptive and efficient simulation method and system. Background Art

[0002] Existing simulations of electric propulsion plasma oscillations typically employ the traditional particle-based approach. This approach relies on simplifying the physical, real-world electric thruster plasma discharge region into a numerical computational domain, dividing the continuous physical space into discrete geometric grids. Discrete Maxwell equations for the electromagnetic field or Poisson equations for the electrostatic field are solved on these grids to obtain the background electric and magnetic fields at each instant. The real plasma particles are then treated as macro-clusters formed by the aggregation of numerous particles, with the same specific charge as real particles. After setting initial values, a first-principles numerical solution is then used to investigate the dynamic characteristics of the formation and evolution of electric propulsion plasma oscillations.

[0003] However, in order to achieve high computational accuracy, it is often necessary to use hundreds of millions of particles and millions of grids. The large number of particles and grids makes this method very inefficient. Acceleration methods such as increasing weights may increase numerical noise, making it difficult for traditional electric propulsion plasma oscillation simulations to strike a balance between accuracy and efficiency. Summary of the Invention

[0004] The purpose of the present invention is to provide an adaptive and efficient simulation method and system for electric propulsion plasma oscillations, so as to alleviate the technical problem in the prior art that both the simulation efficiency and accuracy of electric propulsion plasma oscillation simulation cannot be taken into account.

[0005] In a first aspect, the present invention provides an adaptive and efficient simulation method for electric propulsion plasma oscillations, comprising: step 110, constructing a high-precision electric propulsion plasma numerical simulation model; wherein the simulation model comprises: a particle motion model, a charge deposition model, an electromagnetic field solution model, and a collision solution model; step 120, decomposing the computational tasks of the target model based on a first adjustment parameter set, and decomposing the computational tasks of the electromagnetic field solution model based on a second adjustment parameter set; wherein the first adjustment parameter set comprises: the number of first parallel computing tasks, the number of GPU cores required to call for each computing task, the load workload of each GPU core, and the time step of each particle push; the target model comprises: a particle motion model, a charge deposition model, and a collision solution model; the second adjustment parameter set comprises: the number of second parallel computing tasks, the number of CPU cores required to call for each computing task, and the load workload of each CPU core; step 130, simulating and solving the simulation model to obtain a simulation result. Results; wherein, the electromagnetic field solution model is solved by the multigrid method of the third adjustment parameter set; the third adjustment parameter set includes: the multigrid cycle form, the number of multigrid layers, the number of grid smoothing times per layer, the multigrid smoothing method and the grid spacing in each direction; step 140, using the target neural network model to extract features of the simulation results, so as to determine the updated first adjustment parameter set, the updated second adjustment parameter set and the updated third adjustment parameter set based on the feature extraction results; wherein, the target neural network model is a model obtained after training with the goal of maximizing simulation efficiency; the feature extraction results include: the electron density, electron temperature, elastic collision frequency between electrons and atoms, excitation collision frequency, ionization collision frequency, momentum exchange collision frequency and charge exchange collision frequency between ions and atoms, potential gradient and the growth rate of plasma oscillation over time of each grid node; step 150, repeating steps 120 to 140 until the preset simulation cutoff condition is reached.

[0006] Optionally, the simulation model is simulated and solved to obtain simulation results, including: initializing the number of particles in the electric propulsion plasma and the mass, charge, motion speed and position of each particle; depositing the charges of all charged particles to the corresponding grid nodes of the finest grid in the multi-grid based on the charge deposition model, and obtaining the charge amount on each grid node; wherein the grid nodes corresponding to the target charged particle are the 8 vertices of the grid cube closest to it; the target charged particle represents any charged particle among all charged particles; calculating the charge density of the target grid node based on the charge amount of the target grid node and the unit grid volume in the finest grid; wherein the target grid node represents any grid node in the finest grid; based on the multigrid method and the charge density of each grid node in the finest grid, the electromagnetic field solution model is solved to obtain the potential distribution in the electric propulsion plasma; the force on each particle is calculated based on the potential distribution to solve the updated position and movement speed of each particle based on the particle motion model; based on the updated movement speed of each particle and the collision solution model, the number of target collisions between particles is determined; wherein, target collisions include: elastic collisions between electrons and atoms, excitation collisions between electrons and atoms, ionization collisions between electrons and atoms, momentum exchange collisions between ions and charge exchange collisions between ions and atoms.

[0007] Optionally, based on the charge deposition model, the charges of all charged particles are deposited onto the corresponding grid nodes of the finest grid in the multigrid to obtain the charge amount on each grid node, including: dividing all charged particles evenly into group and copy the grid space of the finest grid in the multigrid to Partitions are created so that the GPU cores, charged particle groups, and grid spaces are in one-to-one correspondence; M represents the number of the first parallel computing tasks corresponding to the charge deposition model, and N represents the number of GPU cores required to be called for each computing task; each GPU core deposits the charges of the charged particles in its corresponding charged particle group into the corresponding grid space through the charge deposition model; all The charges of the grid spaces are accumulated to obtain the charge on each grid node.

[0008] Optionally, the particle motion model is: ;in, represents the mass of the particle, Indicates that the particle The speed of the movement in time steps, Indicates that the particle The speed of the movement in time steps, represents the time step of particle pushing, Indicates that the particle The force at each time step, Indicates that the particle The position of the time step, Indicates that the particle The position of the time step.

[0009] Optionally, the charge deposition model is: ;in, represents the grid spacing in the x direction, represents the grid spacing in the y direction, represents the grid spacing in the z direction, represents the position coordinates of the particle that undergoes charge deposition, represents the grid node coordinates, Indicates that the particle is at the grid node The charge distribution weight on , , , represents the number of grid nodes in the x direction, represents the number of grid nodes in the y direction, represents the number of mesh nodes in the z direction, Represents the macro particle weight.

[0010] Optionally, the electromagnetic field solution model is The discretized equation group consists of the objective equation of the grid nodes; the grid nodes The objective equation is expressed as: ;in, Represents a grid node The electric potential on Represents a grid node The charge density, represents the dielectric constant of vacuum.

[0011] Optionally, the collision solution model includes the following solution steps: using the formula Calculate the maximum collision frequency of all particles; where, represents the atomic number density, represents the velocity of the i-th particle, , represents the kinetic energy of particle i, represents the collision cross section of particle i undergoing the jth collision, Represents the maximum collision frequency of all particles; using the formula Calculate the probability of all particles having empty collisions in the current time step; where, represents the time step of particle pushing, Indicates the probability of all particles having empty collisions in the current time step; using the formula Count the number of collisions between particles; where, represents the total number of particles, Indicates the number of collisions between particles; randomly generated A random number corresponding to particles collide; if , then the kth colliding particle undergoes the first type of collision; among them, Indicates the random number corresponding to the kth colliding particle; represents the frequency of the jth collision of the kth particle; if , then the kth colliding particle will have the j+1th collision; if , then the kth colliding particle undergoes an empty collision.

[0012] Optionally, after determining the number of target collisions between particles, the method further includes: constructing a three-dimensional computational domain unit model and initializing particle parameters in the unit model; wherein the particle parameters include: atomic density, electron density, ion density, electron kinetic energy and ion kinetic energy; disabling the particle motion model, and counting the number of target collisions between particles within the time step of particle pushing; calculating the reaction rate error of each collision based on the number of each collision in the target collision and the particle parameters; when determining that the reaction rate error in the target collision is greater than a preset threshold, correcting the collision cross section of the corresponding collision, and / or correcting the time step of the particle pushing.

[0013] Optionally, before solving the electromagnetic field solution model using the multigrid method with the third adjustment parameter set, the method further includes: rewriting the discretized equation group into a linear equation group in the form of AS=b; wherein S represents an array of unknowns, A represents a coefficient matrix of the unknown array S, and b represents a source term array. ,and .

[0014] In a second aspect, the present invention provides an electric propulsion plasma oscillation adaptive and efficient simulation system, comprising: a construction module for constructing a high-precision electric propulsion plasma numerical simulation model; wherein the simulation model comprises: a particle motion model, a charge deposition model, an electromagnetic field solution model and a collision solution model; a decomposition module for decomposing the computing tasks of the target model based on a first adjustment parameter set, and decomposing the computing tasks of the electromagnetic field solution model based on a second adjustment parameter set; wherein the first adjustment parameter set comprises: the number of first parallel computing tasks, the number of GPU cores required to call for each computing task, the load workload of each GPU core and the time step of each particle push; the target model comprises: a particle motion model, a charge deposition model and a collision solution model; the second adjustment parameter set comprises: the number of second parallel computing tasks, the number of CPU cores required to call for each computing task and the load workload of each CPU core; the simulation module is used to simulate and solve the simulation model to obtain a simulation result The electromagnetic field solution model is solved by a multigrid method of a third adjustment parameter set; the third adjustment parameter set includes: a multigrid cycle form, a multigrid layer number, a grid smoothing number of times per layer, a multigrid smoothing method, and a grid spacing in each direction; a processing module is used to extract features from the simulation results using a target neural network model, so as to determine an updated first adjustment parameter set, an updated second adjustment parameter set, and an updated third adjustment parameter set based on the feature extraction results; wherein the target neural network model is a model obtained after training with the goal of maximizing simulation efficiency; the feature extraction results include: the electron density, electron temperature, elastic collision frequency between electrons and atoms, excitation collision frequency, ionization collision frequency, momentum exchange collision frequency and charge exchange collision frequency between ions and atoms, potential gradient, and the growth rate of plasma oscillation over time of each grid node; an optimization module is used to repeatedly call the decomposition module, the simulation module, and the processing module until a preset simulation cutoff condition is reached.

[0015] The present invention provides an adaptive and efficient simulation method for electric propulsion plasma oscillations. The method adopts a high-precision electric propulsion plasma numerical simulation model to ensure the accuracy of the simulation results, and decomposes the calculation tasks of the simulation model through a variety of adjustment parameters, that is, a parallel computing method is used to complete all calculation tasks, thereby effectively improving the simulation efficiency. Moreover, when simulating and solving the simulation model, a multi-grid method is specifically used to solve the electromagnetic field solution model. Based on this solution method, the simulation calculation efficiency can also be further improved. In addition, the present invention uses a target neural network model to process the simulation results to update all the adjustment parameters in the above process, thereby continuously and adaptively optimizing the multiple adjustment parameters used in the simulation process, thereby achieving a technical effect of taking into account the simulation efficiency and accuracy of the electric propulsion plasma oscillation simulation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 A flow chart of an adaptive and efficient simulation method for electric propulsion plasma oscillations provided by an embodiment of the present invention; Figure 2 A schematic diagram of a simulation cycle calculation framework provided by an embodiment of the present invention; Figure 3 A functional module diagram of an electric propulsion plasma oscillation adaptive and efficient simulation system provided by an embodiment of the present invention; Figure 4 A schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0019] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0020] The following embodiments of the present invention are described in detail with reference to the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments may be combined with each other.

[0021] Example 1 Figure 1 A flow chart of an adaptive and efficient simulation method for electric propulsion plasma oscillation provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the method specifically includes the following steps: Step 110: construct a high-precision electric propulsion plasma numerical simulation model.

[0022] For electric thruster plasma, its working process involves complex plasma reactions and wave-plasma interactions. Therefore, choosing a suitable simulation model is the theoretical basis for accurately describing the corresponding plasma process. Therefore, at the beginning of the method execution, it is first necessary to establish a high-precision electric propulsion plasma numerical simulation model to ensure the accuracy of the simulation results. Among them, the simulation model includes: particle motion model, charge deposition model, electromagnetic field solution model and collision solution model. The boundary conditions of each simulation model need to be defined according to the specific electric propulsion system. For example, non-metallic materials such as ceramics are defined as dielectric boundaries, and the anode and thruster wall are defined as the first type of boundary conditions with fixed potential. The embodiment of the present invention does not specifically limit the boundary conditions of the simulation model. The user needs to perform adaptive design according to the actual simulation system.

[0023] Step 120 : Decomposing the computational task of the target model based on the first adjustment parameter set, and decomposing the computational task of the electromagnetic field solution model based on the second adjustment parameter set.

[0024] The method proposed in this embodiment of the present invention is essentially an AI-optimized electric propulsion plasma oscillation simulation method. Specifically, it utilizes AI to optimize the design of electric propulsion plasma oscillation simulation. After determining the simulation model, to optimize the electric propulsion plasma oscillation simulation process through AI, it is necessary to reserve an extensible AI optimization interface within each computational module (i.e., the module used to simulate and solve the simulation model). This facilitates the subsequent efficient AI-optimized electric propulsion plasma oscillation simulation. Specifically, these interfaces are used to transmit relevant parameters that influence the solution of each simulation model.

[0025] The embodiment of the present invention uses a parallel computing method with adjustable load to optimize the solution of the simulation model to improve the simulation computing efficiency. The parallel task decomposition method of the simulation model needs to be specifically divided into the following two categories: 1. Particle Task Decomposition: Particle task decomposition divides computational tasks involving particle motion and particle-mesh interactions into spatial regions or particle groups, assigning them to different computing units (CPUs and GPUs) for parallel computing. Since the computations involved in particle motion primarily involve velocity and position updates, charge deposition, and particle collisions, these operations require high computing power but have weak inter-task coupling, making them particularly well-suited for accelerated computation using the GPU's multi-threaded architecture. GPUs can efficiently process large numbers of particle trajectories in parallel, significantly improving computational efficiency.

[0026] However, the parallelization process of particle motion needs to consider the load balance problem between particles to prevent certain computing units from becoming bottlenecks due to excessive load. In addition, by adjusting the time step of each particle push to match the plasma frequency, the optimal computing efficiency can be obtained while maintaining the computing accuracy. Therefore, when decomposing the particle tasks, the various parameters in the first adjustment parameter set should be optimized in combination with the task characteristics of the target model to achieve efficient particle simulation. Among them, the first adjustment parameter set includes: the number of first parallel computing tasks, the number of GPU cores required to call for each computing task, the load workload of each GPU core and the time step of each particle push; the target model includes: particle motion model, charge deposition model and collision solution model.

[0027] 2. Field solution task decomposition: Field solution task decomposition refers to the division of tasks involving electromagnetic field distribution calculations and their allocation to different computing units for parallel solution. Unlike particle tasks, the field solution process often involves numerical calculations on a global grid, and the coupling between tasks is strong. In particular, field smoothing and residual calculations in multi-grid methods require information exchange at multiple grid levels. Due to the CPU's advantages in handling complex task dependencies and memory management, field solution task decomposition is more suitable for CPU implementation. Specifically, field solution parallelization can be achieved by dividing the grid area.

[0028] However, data communication overhead between CPUs must be optimized. If the assigned task granularity is not reasonable, the increased communication overhead may offset the performance gains brought by parallelization. Therefore, when decomposing the computational tasks of the electromagnetic field solution model, it is necessary to optimize the parameters in the following second adjustment parameter set to ensure overall simulation efficiency improvement. The second adjustment parameter set includes: the number of second parallel computing tasks, the number of CPU cores required for each computing task, and the workload of each CPU core.

[0029] Step 130: Perform simulation on the simulation model to obtain simulation results.

[0030] Among them, the electromagnetic field solution model is solved using the multi-grid method of the third adjustment parameter set; the third adjustment parameter set includes: multi-grid cycle form (for example, V cycle, W cycle, F cycle), the number of multi-grid layers, the number of grid smoothing times per layer, the multi-grid smoothing method (Jacobi iteration or successive super-relaxation iteration method, etc.) and the grid spacing in each direction.

[0031] Specifically, after decomposing the computational tasks of each simulation model, the significant data communication overhead between particles and the grid is considered. Especially in high-resolution grids, large amounts of particle charge deposition and field information updates can lead to frequent host-device data transfers, thereby reducing overall simulation efficiency. Therefore, to improve simulation solution speed, this embodiment of the present invention utilizes a multigrid approach when solving electromagnetic field models. The relevant parameters in this algorithm (i.e., the aforementioned third set of adjustment parameters) are included in the list of parameters that require artificial intelligence adjustment.

[0032] The impact of the parameters included in the first, second, and third adjustment parameter sets on simulation computational efficiency is highly dependent on the computational conditions and the current state of the plasma simulation. Improper selection will not only fail to improve the efficiency of the electric propulsion plasma oscillation simulation, but may actually significantly reduce it. However, due to the complexity of the simulation parameters that must be considered and the wide range of plasma parameter variations over the simulation timeline, manual selection of simulation parameters is difficult to achieve in a timely manner. Therefore, embodiments of the present invention require artificial intelligence optimization of the electric propulsion plasma numerical simulation model.

[0033] Step 140 , performing feature extraction on the simulation results using the target neural network model, so as to determine an updated first adjustment parameter set, an updated second adjustment parameter set, and an updated third adjustment parameter set based on the feature extraction results.

[0034] Among them, the target neural network model is a model obtained after training with the goal of maximizing simulation efficiency; the feature extraction results include: the electron density of each grid node, electron temperature, elastic collision frequency between electrons and atoms, excitation collision frequency, ionization collision frequency, momentum exchange collision frequency and charge exchange collision frequency between ions and atoms, potential gradient and the growth rate of plasma oscillation over time.

[0035] The simulation results of the simulation model include: the amount of charge on each grid node, the charge density of each grid node, the potential distribution in the electric propulsion plasma, the movement speed and position of each particle, and the number of target collisions between particles. The target collisions include: elastic collisions between electrons and atoms, excitation collisions between electrons and atoms, ionization collisions between electrons and atoms, momentum exchange collisions between ions and atoms, and charge exchange collisions between ions and atoms.

[0036] Considering the following reasons, after obtaining the simulation results, the embodiment of the present invention needs to use the target neural network model to perform feature extraction on the simulation results, so as to update the first adjustment parameter set, the second adjustment parameter set, and the third adjustment parameter set according to the feature extraction results: 1. When particle weights are equal, the electron density affects the number of macroparticles and thus the solution efficiency of the particle motion model, charge deposition model, and collision solution model. The respective parallel workloads need to be distributed through the first adjustment parameter set. In addition, the electron density directly determines the plasma frequency. The time step of each particle push needs to be adjusted in the first adjustment parameter set to match the plasma frequency to obtain the optimal computational efficiency while maintaining the computational accuracy.

[0037] 2. Electron density and electron temperature directly determine the plasma Debye length. The plasma Debye length needs to be matched by adjusting the grid spacing in each direction in the second adjustment parameter set to achieve optimal computational efficiency while maintaining computational accuracy.

[0038] 3. Electron temperature, the elastic collision frequency between electrons and atoms, the excitation collision frequency, the ionization collision frequency, the momentum exchange collision frequency between ions and atoms, and the charge exchange collision frequency will affect the solution efficiency of the collision solution model. The parallel workload needs to be distributed through the first set of adjustment parameters.

[0039] 4. Electric potential gradient. Strong electric potential gradient will reduce the efficiency of the electromagnetic field solution model. It is necessary to adjust the second and third adjustment parameter sets to increase the computing resources (number of computing cores) occupied by the electromagnetic field solution model to improve computing efficiency.

[0040] 5. The growth rate of plasma oscillation over time may trigger more particle dynamic behaviors and increase computational complexity. Therefore, it is also used as one of the input parameters. The robustness of the neural network model needs to be improved by adjusting the first adjustment parameter set, the second adjustment parameter set, and the third adjustment parameter set.

[0041] 6. Various parameters contained in the feature extraction results, spatially non-uniform regions with drastic parameter changes (such as the plasma sheath region) may cause uneven loads, and it is necessary to increase the computing resource allocation (number of computing cores) in the computing density region through the first and second adjustment parameter sets.

[0042] The following is a detailed introduction to the method of determining each parameter in the above feature extraction results: By replacing the deposited charge with "1" or the directional components of the particle velocity or the particle kinetic energy, the particle density, directional components of the particle velocity, or the particle kinetic energy of any type of particle at the target grid node can be calculated. The directional components of the particle velocity, or the particle kinetic energy of any type of particle at the target grid node are referred to as electrons, ions, and atoms. Based on the particle density, directional components of the particle velocity, and kinetic energy of the electrons at the target grid node, the electron temperature at the target grid node can be calculated using the following algorithm: [2×(electron kinetic energy - electron directional motion kinetic energy) / (3×electron density)]. The electron directional motion kinetic energy is calculated using the directional components of the electron velocity at the target grid node. This method eliminates the contribution of directional drift to the electron temperature by eliminating the directional kinetic energy, accurately solving for the electron temperature, and determining the electron's thermal motion state.

[0043] Based on the collision solution model, the frequency of each target collision can be calculated by counting the number of target collisions per unit time. After obtaining the potential distribution in the electric propulsion plasma, the potential gradient can be calculated based on the potential distribution.

[0044] Build as Figure 2 The illustrated cyclic computational framework follows the charge deposition model, which is followed by an electromagnetic field solution model, a particle motion model, a collision solution model, and finally a charge deposition model. This loop generates simulation results for a certain number of iterations, where the number of iterations is determined by the desired simulation time. Based on these simulation results, the time-dependent sequence of electron density at each spatial point within a certain timeframe is taken, its absolute value taken, and then low-pass filtered to obtain the envelope signal. This signal is then numerically derived to calculate the growth rate of the plasma oscillation over time.

[0045] In an embodiment of the present invention, the target neural network model is a model obtained by training the initial neural network model with the goal of maximizing simulation efficiency. Although the input data of the neural network model is the simulation results of the aforementioned simulation model, the input features used during model training are essentially the feature extraction results corresponding to the simulation results, and the output labels are the first adjustment parameter set, the second adjustment parameter set, and the third adjustment parameter set. Simulation efficiency is inversely proportional to the simulation time of the simulation cycle (charge deposition model -> electromagnetic field solution model -> particle motion model -> collision solution model). That is, the longer the simulation time, the lower the simulation efficiency. During model training, the total time of each simulation cycle can be calculated, and the simulation efficiency can be determined based on the average of the total time of multiple adjacent simulation cycles to reduce fluctuation errors.

[0046] The training process of the neural network model is as follows: 1) Data Acquisition and Preparation: For a given input feature, uniformly sample the input within the parameter range of the output label. Run an electric propulsion plasma oscillation simulation without adaptive control and record the plasma parameters (feature extraction results) and computational time at each time step.

[0047] 2) Data Analysis and Feature Extraction: Analyze the data obtained in 1) to extract input features, the shortest computation time, and their corresponding output labels, and establish a mapping between input features and optimal output labels.

[0048] 3) Based on the data obtained in 2), train the initial neural network model until a predetermined end condition is met, enabling it to accurately predict the relationship between input features and optimal output labels. The predetermined end condition may include, for example, reaching a specified number of training cycles and achieving stable convergence in simulation efficiency.

[0049] Step 150 , repeating steps 120 to 140 until a preset simulation cutoff condition is reached.

[0050] After the target neural network model is obtained through training, the trained target neural network model is integrated into the electric propulsion plasma oscillation simulation system. During the electric propulsion plasma oscillation simulation process, the target neural network model can predict and dynamically adjust the output label by collecting simulation results in real time and calculating feature extraction results, thereby realizing adaptive and efficient simulation of electric propulsion plasma oscillation.

[0051] An embodiment of the present invention provides an adaptive and efficient simulation method for electric propulsion plasma oscillations. The method adopts a high-precision electric propulsion plasma numerical simulation model to ensure the accuracy of the simulation results, and decomposes the calculation tasks of the simulation model through a variety of adjustment parameters, that is, a parallel computing method is used to complete all calculation tasks, thereby effectively improving the simulation efficiency. Moreover, when simulating and solving the simulation model, a multi-grid method is specifically used to solve the electromagnetic field solution model. Based on this solution method, the simulation calculation efficiency can also be further improved. In addition, the present invention uses a target neural network model to process the simulation results to update all the adjustment parameters in the above process, thereby continuously and adaptively optimizing the multiple adjustment parameters used in the simulation process, thereby achieving a technical effect of taking into account the simulation efficiency and accuracy of the electric propulsion plasma oscillation simulation model.

[0052] In an optional embodiment, the above step 130, performing simulation on the simulation model to obtain simulation results, specifically includes the following steps: Step 1301: Initialize the number of particles in the electric propulsion plasma and the mass, charge, speed and position of each particle.

[0053] In the embodiment of the present invention, particles in the electric propulsion plasma include electrons, ions, and atoms. Before simulation, the number, mass, charge, movement speed, and position of each of the above particles need to be initialized.

[0054] Step 1302: Based on the charge deposition model, the charges of all charged particles are deposited onto the corresponding grid nodes of the finest grid in the multi-grid to obtain the charge amount on each grid node; wherein the grid nodes corresponding to the target charged particle are the eight vertices of the grid cube closest to it; the target charged particle represents any charged particle among all charged particles.

[0055] In the embodiment of the present invention, the charge deposition model is: ;in, represents the grid spacing in the x direction, represents the grid spacing in the y direction, represents the grid spacing in the z direction, represents the position coordinates of the particle that undergoes charge deposition, represents the grid node coordinates, Indicates that the particle is at the grid node The charge distribution weight on , , , represents the number of grid nodes in the x direction, represents the number of grid nodes in the y direction, represents the number of mesh nodes in the z direction, Represents the macro particle weight.

[0056] Based on the above charge deposition model, the charge of each charged particle can be deposited onto the grid nodes in the multigrid. This charge deposition model is applicable to three-dimensional multigrids. Therefore, the grid nodes corresponding to the target charged particle are the eight vertices of the grid cube closest to it. If a two-dimensional multigrid is used, the grid nodes corresponding to the target charged particle are the four vertices of the grid closest to it. Multiple particles may have their charges deposited on each grid node. The charge of any grid node is the net charge obtained by subtracting all negative charges from all positive charges on that grid node. In embodiments of the present invention, this charge deposition model is also applicable to processing particle density, average particle velocity, average particle energy, and the like.

[0057] Step 1303 : Calculate the charge density of the target grid node based on the charge of the target grid node and the unit grid volume in the finest grid; wherein the target grid node represents any grid node in the finest grid.

[0058] Specifically, the charge of the target grid node is divided by the unit grid volume in the finest grid, and the result is the charge density of the grid node.

[0059] Step 1304 : Solve the electromagnetic field solution model based on the multigrid method and the charge density of each grid node in the finest grid to obtain the electric potential distribution in the electric propulsion plasma.

[0060] Generally, electromagnetic field solutions can be roughly divided into two parts: solving the electric field and solving the magnetic field. Both involve differentially solving Maxwell's equations to obtain the spatial distribution of the electric and magnetic fields. In electric thrusters, especially Hall thrusters, the external magnetic field is much larger than the magnetic field generated by plasma coupling, so they can be considered a magnetostatic model. In this magnetostatic model, particles are affected by both electric and magnetic fields, but the induced magnetic field generated by particle motion is very small and can be ignored. In this case, the magnetic field can be considered a static magnetic field, and the calculation process only requires solving the Poisson equation for the electric field.

[0061] After obtaining the charge value at each grid node in the computational domain, the spatial potential distribution can be obtained through the Poisson equation: ,in, represents the electric potential, represents the charge value at the grid node, represents the dielectric constant of vacuum, represents the number density of ions, represents the number density of electrons. However, since the dimension of Poisson's equation is too high, it cannot be solved analytically by directly solving it. Therefore, it is necessary to discretize it into a system of linear equations through the difference method.

[0062] In an optional embodiment, the electromagnetic field solution model is The discretized equations are composed of the objective equations of the grid nodes; among them, the grid nodes The objective equation is expressed as: ;in, Represents a grid node The electric potential on Represents a grid node The charge density, represents the dielectric constant of vacuum.

[0063] The spatial distribution of electric potential can be obtained by solving the discretized equations. The spatial distribution of electric field can be further obtained by differential potential. The calculation formula of electric field intensity is: .

[0064] In order to improve the solution speed of the equation system and improve the simulation efficiency, in an embodiment of the present invention, before solving the electromagnetic field solution model using the multigrid method with the third adjustment parameter set, the following steps are also included: Rewrite the discretized equations into a linear equation system in the form of AS=b; where S represents the unknown array, A represents the coefficient matrix of the unknown array S, and b represents the source term array. ,and .

[0065] Regarding the numerical solution of the aforementioned linear equations, traditional direct solutions (Gaussian elimination, LU decomposition, etc.) and iterative solutions (Jacobi iteration, successive super-relaxation iteration, etc.) are too slow. Therefore, the embodiments of the present invention employ a multigrid method to improve computational efficiency. The multigrid method can be used to obtain the potential of each grid node in the spatial multigrid. Then, through interpolation, the potential at each particle's location can be obtained, thereby obtaining the potential distribution in the electric propulsion plasma.

[0066] Step 1305 : Calculate the force on each particle based on the electric potential distribution, and solve the updated position and motion speed of each particle based on the particle motion model.

[0067] Taking a particle as an example, after obtaining the electric potential at the particle, according to The electric field strength at the location of the particle can be calculated. The force exerted on a charged particle in the electric field is F=qE, where q is the charge of the particle.

[0068] During the simulation process, it is necessary to record the velocity and position of each particle, and perform calculations and update the data at each time step. When there are a large number of particles, this places high demands on the computer's storage space. In order to minimize storage space while ensuring calculation accuracy, in this embodiment of the present invention, the particle motion model is: ;in, represents the mass of the particle, Indicates that the particle The speed of the movement in time steps, Indicates that the particle The speed of the movement in time steps, represents the time step of particle pushing, Indicates that the particle The force at each time step, Indicates that the particle The position of the time step, Indicates that the particle The position of the particle is calculated at a time step. That is, the particle's velocity v and position x are at different time points. The time point for position calculation is selected according to an integer multiple of the time step, and the time point for velocity calculation is selected according to a half-integer multiple of the time step. The solution time of x and v is staggered by 0.5 , the two are solved alternately, thereby effectively reducing the requirements for computer storage space.

[0069] Step 1306, based on the updated motion speed and collision solution model of each particle, determines the number of target collisions between particles; wherein the target collisions include: elastic collisions between electrons and atoms, excitation collisions between electrons and atoms, ionization collisions between electrons and atoms, momentum exchange collisions between ions and atoms, and charge exchange collisions between ions and atoms.

[0070] In electric thrusters, various collisions occur between particles. Different types of collisions occur, including electron-atom, ion-atom, electron-ion, atom-atom, and ion-ion collisions. Due to the small dimensions of electric thrusters and the large mean free paths of most collisions, the probability of these collisions is low. The dominant collision processes in electric thrusters are elastic collisions, excitation collisions, and ionization collisions between electrons, as well as momentum exchange and charge exchange collisions between ions.

[0071] In an embodiment of the present invention, the collision solution model includes the following solution steps: Step A: Using the formula Calculate the maximum collision frequency of all particles; where, represents the atomic number density, represents the velocity of the i-th particle, , represents the kinetic energy of particle i, The collision cross section of particle i undergoing the jth collision is obtained by consulting the public experimental data and is pre-set in the simulation. Indicates the maximum collision frequency of all particles. In the embodiment of the present invention, according to the number of collision types included in the target collision, N is set to 5, representing 5 types of collisions.

[0072] The number density of atoms is obtained by the following method: referring to the charge deposition model to calculate the charge distribution weight, the charge distribution weight can be used as the atom distribution weight when calculating the number density of atoms. All "atoms" are deposited on each grid node to obtain the number of atoms on each grid node. The number of atoms on each grid node is then divided by the volume of the finest grid unit to obtain the number of atoms on each grid node. Indicates the maximum atomic number density among all mesh nodes in the space.

[0073] Step B: Using the formula Calculate the probability of all particles having empty collisions in the current time step; where, represents the time step of particle pushing, Represents the probability of null collision for all particles in the current time step.

[0074] Step C: Using the formula Count the number of collisions between particles; where, represents the total number of particles, Indicates the number of collisions between particles.

[0075] Step D: Random Generation A random number corresponding to particles that collide.

[0076] Step E: If , then the kth colliding particle undergoes the first type of collision; among them, Indicates the random number corresponding to the kth colliding particle; represents the frequency of the jth collision of the kth particle.

[0077] Step F: If , then the kth colliding particle undergoes the j+1th collision.

[0078] Step G: If , then the kth colliding particle undergoes an empty collision.

[0079] The above method is used to deal with the collision between particles. The calculation of collision of individual particles is greatly reduced, which greatly reduces the number of particles involved in the collision calculation and improves the calculation efficiency.

[0080] In an optional implementation manner, after determining the number of target collisions between particles, the embodiment of the present invention further includes the following steps: Step 201 : construct a three-dimensional computational domain unit model and initialize the particle parameters in the unit model; wherein the particle parameters include: atomic density, electron density, ion density, electron kinetic energy and ion kinetic energy.

[0081] Step 202 : disable the particle motion model and count the number of target collisions between particles within the time step of particle propulsion.

[0082] Step 203 : Calculate the reaction rate error of each collision based on the number of each collision and the particle parameters in the target collision.

[0083] Step 204 : When it is determined that the reaction rate error in the target collision is greater than a preset threshold, the collision cross section of the corresponding collision is corrected, and / or the time step of the particle propulsion is corrected.

[0084] Specifically, after using parallel computing to improve the efficiency of electric propulsion plasma oscillation simulation, in order to further improve the simulation accuracy, the embodiment of the present invention proposes to calibrate the collision model. Optionally, a three-dimensional computational domain unit model with a size of 1mm×1mm×1mm (1mm is an exemplary value) is constructed, and the atomic density in the unit model is initialized. , electron density , ion density , electron kinetic energy and ion kinetic energy .

[0085] Next, in order to disable the change in particle velocity, loss, and generation after the collision model occurs, the embodiment of the present invention disables the aforementioned particle motion model, and within the time step of simulating a given particle push, counts the number of events of each collision occurring during this period.

[0086] In conjunction with the collision types included in the embodiments of the present invention, it can be seen that target collisions can be divided into two categories: collisions between electrons and atoms (first-type collisions) and collisions between ions and atoms (second-type collisions). Therefore, when calibrating the collision solution model, it is necessary to refer to the following method: If the jth collision of the first type of collision occurs (j=1, 2, 3), then the actual reaction rate of the jth collision (the number of collision events per unit time and unit volume) is calculated by the formula The theoretical reaction rate of the j-th collision is calculated by the formula Calculate, where represents the number of j-th collisions, represents the macro-particle weight (the real particle weight represented by a macro-particle in the simulation of the macro-particle method), represents the elementary charge, Represents the mass of an electron. and The error is the reaction rate error of the j-th collision in the first type of collision.

[0087] If the jth collision of the second type of collision occurs (j=1, 2), then the actual reaction rate of the jth collision (the number of collision events per unit time and unit volume) is calculated by the formula The theoretical reaction rate of the j-th collision is calculated by the formula Calculate, where Indicates the mass of the ion. and The error is the reaction rate error of the j-th collision in the second type of collision.

[0088] In other words, the theoretical reaction rate formula of target collision can be expressed as: , where x=e if a first-type collision occurs, and x=i if a second-type collision occurs.

[0089] The above method can be used to calculate the reaction rate error for each type of collision in the target collision. If the error is less than a preset threshold (e.g., 5%), the computational accuracy of the parallelized collision model is considered to meet the requirements. If there is a reaction rate error greater than the preset threshold, the collision cross section of the corresponding collision is corrected, and / or the time step of the particle push is corrected (because an excessively large time step may cause a collision that should have occurred to not occur). For example, if the reaction rate error of an elastic collision between an electron and an atom is greater than a preset threshold, it is necessary to correct the collision cross section of the elastic collision between the electron and the atom, and / or correct the time step of the electron and atom push.

[0090] To ensure efficient computational efficiency in the charge deposition model, simply converting the serial code to parallel execution would result in read / write conflicts between different compute cores when processing the same grid node, severely distorting the simulation results. A common approach in the prior art is to have the compute cores wait. If two compute cores attempt to read or write to the same grid node simultaneously, one core will read or write first, while the other waits for the first core to complete its read / write. While this approach ensures accurate simulation results, when conflicts between different compute cores occur frequently, the waiting time between the cores will slow down the efficiency of the entire charge deposition model.

[0091] Therefore, in the embodiment of the present invention, the above step 1302, which deposits the charges of all charged particles onto the corresponding grid nodes of the finest grid in the multi-grid based on the charge deposition model to obtain the charge amount on each grid node, specifically includes the following steps: Step 13021, divide all charged particles evenly into group and copy the grid space of the finest grid in the multigrid to so that there is a one-to-one correspondence between GPU cores, charged particle groups, and grid spaces; where M represents the number of the first parallel computing tasks corresponding to the charge deposition model, and N represents the number of GPU cores required to be called for each computing task.

[0092] In step 13022, each GPU core deposits the charges of the charged particles in its corresponding charged particle group into the corresponding grid space through the charge deposition model.

[0093] Step 13023, all The charges of the grid spaces are accumulated to obtain the charge on each grid node.

[0094] Based on the method described above, during the actual computation process, each GPU core deposits its own group of charged particles onto the corresponding grid space using the charge deposition model. This ensures that the computations performed by each GPU core do not conflict with each other. After the computation is complete, the charge distribution in the grid space on each GPU core is summed, resulting in an accurate charge distribution while improving computational efficiency. This method can significantly improve the computational efficiency of the charge deposition model when conflicts between different cores are high.

[0095] In summary, the embodiments of the present invention can greatly improve the simulation efficiency of the electric propulsion plasma oscillation simulation model while ensuring simulation accuracy, and can combine artificial intelligence to adaptively optimize the adjustment parameters of multiple key links in the simulation process, thereby reducing the consumption of computing resources.

[0096] Example 2 An embodiment of the present invention further provides an electric propulsion plasma oscillation adaptive and efficient simulation system, which is mainly used to execute the electric propulsion plasma oscillation adaptive and efficient simulation method provided in the above-mentioned embodiment 1. The electric propulsion plasma oscillation adaptive and efficient simulation system provided in the embodiment of the present invention is specifically introduced below.

[0097] Figure 3 This is a functional module diagram of an electric propulsion plasma oscillation adaptive and efficient simulation system provided by an embodiment of the present invention, such as Figure 3 As shown, the system mainly includes: a construction module 10, a decomposition module 20, a simulation module 30, a processing module 40, and an optimization module 50, wherein: The construction module 10 is used to construct a high-precision electric propulsion plasma numerical simulation model; wherein the simulation model includes: a particle motion model, a charge deposition model, an electromagnetic field solution model and a collision solution model.

[0098] The decomposition module 20 is used to decompose the computing tasks of the target model based on the first adjustment parameter set, and to decompose the computing tasks of the electromagnetic field solution model based on the second adjustment parameter set; wherein the first adjustment parameter set includes: the number of first parallel computing tasks, the number of GPU cores required to call for each computing task, the load workload of each GPU core, and the time step of each particle push; the target model includes: a particle motion model, a charge deposition model, and a collision solution model; the second adjustment parameter set includes: the number of second parallel computing tasks, the number of CPU cores required to call for each computing task, and the load workload of each CPU core.

[0099] The simulation module 30 is used to simulate and solve the simulation model to obtain simulation results. The electromagnetic field solution model is solved using a multigrid method with a third set of adjustment parameters. The third set of adjustment parameters includes: a multigrid loop form, a number of multigrid layers, a number of grid smoothing times per layer, a multigrid smoothing method, and grid spacing in each direction.

[0100] The processing module 40 is used to extract features from the simulation results using the target neural network model, so as to determine the updated first adjustment parameter set, the updated second adjustment parameter set and the updated third adjustment parameter set based on the feature extraction results; wherein the target neural network model is a model obtained after training with the goal of maximizing simulation efficiency; the feature extraction results include: the electron density, electron temperature, elastic collision frequency between electrons and atoms, excitation collision frequency, ionization collision frequency, momentum exchange collision frequency and charge exchange collision frequency between ions and atoms, potential gradient and the growth rate of plasma oscillation over time of each grid node.

[0101] The optimization module 50 is used to repeatedly call the decomposition module, the simulation module and the processing module until a preset simulation cutoff condition is reached.

[0102] The embodiment of the present invention provides an electric propulsion plasma oscillation adaptive and efficient simulation system, which adopts a high-precision electric propulsion plasma numerical simulation model to ensure the accuracy of the simulation results, and decomposes the calculation tasks of the simulation model through a variety of adjustment parameters, that is, adopts a parallel computing method to complete all calculation tasks, thereby effectively improving the simulation efficiency. Moreover, when simulating and solving the simulation model, a multi-grid method is specifically adopted to solve the electromagnetic field solution model, and the simulation calculation efficiency can be further improved based on this solution method. In addition, the embodiment of the present invention uses a target neural network model to process the simulation results to update all the adjustment parameters in the above process, thereby continuously and adaptively optimizing the multiple adjustment parameters used in the simulation process, thereby achieving the technical effect of taking into account the simulation efficiency and accuracy of the electric propulsion plasma oscillation simulation model.

[0103] Optionally, the simulation module 30 includes: The initialization unit is used to initialize the number of particles in the electric propulsion plasma and the mass, charge, movement speed and position of each particle.

[0104] The charge deposition unit is used to deposit the charges of all charged particles onto the corresponding grid nodes of the finest grid in the multigrid based on the charge deposition model to obtain the charge amount on each grid node; the grid nodes corresponding to the target charged particle are the 8 vertices of the grid cube closest to it; the target charged particle represents any charged particle among all charged particles.

[0105] The first calculation unit is used to calculate the charge density of the target grid node based on the charge amount of the target grid node and the unit grid volume in the finest grid; wherein the target grid node represents any grid node in the finest grid.

[0106] The electromagnetic field solving unit is used to solve the electromagnetic field solving model based on the multi-grid method and the charge density of each grid node in the finest grid to obtain the electric potential distribution in the electric propulsion plasma.

[0107] The second calculation unit is used to calculate the force on each particle based on the electric potential distribution, so as to solve the updated position and movement speed of each particle based on the particle motion model.

[0108] The determination unit is used to determine the number of target collisions between particles based on the updated motion speed and collision solution model of each particle; wherein the target collisions include: elastic collisions between electrons and atoms, excitation collisions between electrons and atoms, ionization collisions between electrons and atoms, momentum exchange collisions between ions and atoms, and charge exchange collisions between ions and atoms.

[0109] Optionally, the charge deposition unit is specifically used for: Divide all charged particles evenly into group and copy the grid space of the finest grid in the multigrid to so that there is a one-to-one correspondence between GPU cores, charged particle groups, and grid spaces; where M represents the number of the first parallel computing tasks corresponding to the charge deposition model, and N represents the number of GPU cores required to be called for each computing task.

[0110] Each GPU core deposits the charges of the charged particles in its corresponding charged particle group into the corresponding grid space through the charge deposition model.

[0111] All The charges of the grid spaces are accumulated to obtain the charge on each grid node.

[0112] Optionally, the particle motion model is: ;in, represents the mass of the particle, Indicates that the particle The speed of the movement in time steps, Indicates that the particle The speed of the movement in time steps, represents the time step of particle pushing, Indicates that the particle The force at each time step, Indicates that the particle The position of the time step, Indicates that the particle The position of the time step.

[0113] Optionally, the charge deposition model is: ;in, represents the grid spacing in the x direction, represents the grid spacing in the y direction, represents the grid spacing in the z direction, represents the position coordinates of the particle that undergoes charge deposition, represents the grid node coordinates, Indicates that the particle is at the grid node The charge distribution weight on , , , represents the number of grid nodes in the x direction, represents the number of grid nodes in the y direction, represents the number of mesh nodes in the z direction, Represents the macro particle weight.

[0114] Optionally, the electromagnetic field solution model is The discretized equation group consists of the objective equation of the grid nodes; the grid nodes The objective equation is expressed as: ;in, Represents a grid node The electric potential on Represents a grid node The charge density, represents the dielectric constant of vacuum.

[0115] Optionally, the collision solution model includes the following solution steps: Using formula Calculate the maximum collision frequency of all particles; where, represents the atomic number density, represents the velocity of the i-th particle, , represents the kinetic energy of particle i, represents the collision cross section of particle i undergoing the jth collision, Indicates the maximum collision frequency of all particles.

[0116] Using formula Calculate the probability of all particles having empty collisions in the current time step; where, represents the time step of particle pushing, Represents the probability of null collision for all particles in the current time step.

[0117] Using formula Count the number of collisions between particles; where, represents the total number of particles, Indicates the number of collisions between particles.

[0118] Random Generation A random number corresponding to particles that collide.

[0119] if , then the kth colliding particle undergoes the first type of collision; among them, Indicates the random number corresponding to the kth colliding particle; represents the frequency of the jth collision of the kth particle.

[0120] if , then the kth colliding particle undergoes the j+1th collision.

[0121] if , then the kth colliding particle undergoes an empty collision.

[0122] Optionally, after determining the number of target collisions between particles, the apparatus is further configured to: Construct a three-dimensional computational domain unit model and initialize the particle parameters in the unit model; the particle parameters include: atomic density, electron density, ion density, electron kinetic energy, and ion kinetic energy.

[0123] Disables the particle motion model and counts the number of target collisions between particles during the time step of the particle push.

[0124] Calculates the reaction rate error for each collision based on the number of each collision and the particle parameters in the target collision.

[0125] When it is determined that a reaction rate error greater than a preset threshold exists in the target collision, the collision cross section of the corresponding collision is corrected, and / or the time step of the particle propulsion is corrected.

[0126] Optionally, before solving the electromagnetic field solution model using the multigrid method with the third adjustment parameter set, the device is further configured to: Rewrite the discretized equations into a linear equation system in the form of AS=b; where S represents the unknown array, A represents the coefficient matrix of the unknown array S, and b represents the source term array. ,and .

[0127] Example 3 See also Figure 4An embodiment of the present invention provides an electronic device, which includes: a processor 60, a memory 61, a bus 62 and a communication interface 63, wherein the processor 60, the communication interface 63 and the memory 61 are connected via the bus 62; the processor 60 is used to execute an executable module stored in the memory 61, such as a computer program.

[0128] Memory 61 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk drive. Communication between the system network element and at least one other network element is achieved via at least one communication interface 63 (which may be wired or wireless), and may utilize the Internet, a wide area network, a local area network, a metropolitan area network, or the like.

[0129] The bus 62 may be an ISA bus, a PCI bus, or an EISA bus. The bus may be divided into an address bus, a data bus, a control bus, and the like. For ease of representation, Figure 4 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0130] Among them, the memory 61 is used to store programs, and the processor 60 executes the program after receiving the execution instruction. The method executed by the device defined by the process disclosed in any embodiment of the above-mentioned embodiment of the present invention can be applied to the processor 60 or implemented by the processor 60.

[0131] The processor 60 may be an integrated circuit chip with signal processing capabilities. During implementation, the steps of the above method may be performed by hardware integrated logic circuits or software instructions within the processor 60. The processor 60 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It may implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention may be directly executed by a hardware decoding processor or by a combination of hardware and software modules within the decoding processor. The software modules may be located in storage media well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or the like. The storage medium is located in the memory 61 , and the processor 60 reads the information in the memory 61 and completes the steps of the above method in combination with its hardware.

[0132] An embodiment of the present invention provides a computer program product for an adaptive and efficient simulation method and system for electric propulsion plasma oscillations, comprising a computer-readable storage medium storing non-volatile program code executable by a processor. The program code includes instructions that can be used to execute the methods described in the previous method embodiments. For specific implementation, please refer to the method embodiments and will not be repeated here.

[0133] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0134] If the functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0135] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not require further definition or explanation in subsequent drawings.

[0136] In the description of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer" and the like indicate positions or locations based on the positions shown in the accompanying drawings, or the positions or locations in which the inventive product is typically placed when in use. These terms are intended solely to facilitate the description of the present invention and to simplify the description, and are not intended to indicate or imply that the devices or components referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limitations on the present invention. Furthermore, the terms "first," "second," and "third," etc., are used solely to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0137] Furthermore, terms such as "horizontal," "vertical," and "overhanging" do not necessarily imply that a component must be absolutely horizontal or overhanging, but rather that it can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but rather that it can be slightly tilted.

[0138] In the description of the present invention, it should also be noted that, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; they may refer to mechanical connections or electrical connections; they may refer to direct connections or indirect connections through an intermediate medium; and they may refer to internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0139] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An adaptive and efficient simulation method for electric propulsion plasma oscillation, characterized in that: include: Step 110: construct a high-precision electric propulsion plasma numerical simulation model; wherein the simulation model includes: a particle motion model, a charge deposition model, an electromagnetic field solution model, and a collision solution model; Step 120: Decompose the computational tasks of the target model based on the first adjustment parameter set, and decompose the computational tasks of the electromagnetic field solution model based on the second adjustment parameter set; wherein the first adjustment parameter set includes: the number of first parallel computing tasks, the number of GPU cores required to call each computing task, the load workload of each GPU core, and the time step of each particle propulsion; the target model includes: the particle motion model, the charge deposition model, and the collision solution model; the second adjustment parameter set includes: the number of second parallel computing tasks, the number of CPU cores required to call each computing task, and the load workload of each CPU core; Step 130: simulating and solving the simulation model to obtain simulation results; wherein the electromagnetic field solution model is solved using a multigrid method with a third set of adjustment parameters; the third set of adjustment parameters includes: a multigrid loop form, a number of multigrid layers, a number of grid smoothing times per layer, a multigrid smoothing method, and grid spacing in each direction; Step 140: extracting features from the simulation results using a target neural network model to determine an updated first adjustment parameter set, an updated second adjustment parameter set, and an updated third adjustment parameter set based on the feature extraction results; wherein the target neural network model is a model trained with the goal of maximizing simulation efficiency; and the feature extraction results include: electron density, electron temperature, elastic collision frequency between electrons and atoms, excitation collision frequency, ionization collision frequency, momentum exchange collision frequency and charge exchange collision frequency between ions and atoms, potential gradient, and growth rate of plasma oscillation over time for each grid node; Step 150 , repeating steps 120 to 140 until a preset simulation cutoff condition is reached.

2. The electric propulsion plasma oscillation adaptive and efficient simulation method according to claim 1, characterized in that: The simulation model is simulated and solved to obtain simulation results, including: Initialize the number of particles in the electric propulsion plasma and the mass, charge, speed and position of each particle; Depositing the charges of all charged particles onto corresponding grid nodes of the finest grid in the multigrid based on the charge deposition model to obtain the charge amount on each grid node; wherein the grid node corresponding to the target charged particle is the eight vertices of the grid cube closest to the target charged particle; and the target charged particle represents any charged particle among all the charged particles; Calculating a charge density of the target grid node based on the charge of the target grid node and a unit grid volume in the finest grid; wherein the target grid node represents any grid node in the finest grid; Solving the electromagnetic field solution model based on the multigrid method and the charge density of each grid node in the finest grid to obtain the electric potential distribution in the electric propulsion plasma; Calculating the force on each particle based on the electric potential distribution to solve the updated position and motion speed of each particle based on the particle motion model; Based on the updated motion speed of each particle and the collision solution model, the number of target collisions between particles is determined; wherein the target collisions include: elastic collisions between electrons and atoms, excitation collisions between electrons and atoms, ionization collisions between electrons and atoms, momentum exchange collisions between ions and atoms, and charge exchange collisions between ions and atoms.

3. The electric propulsion plasma oscillation adaptive and efficient simulation method according to claim 2, characterized in that: Based on the charge deposition model, the charges of all charged particles are deposited onto the corresponding grid nodes of the finest grid in the multigrid, and the charge amount on each grid node is obtained, including: All the charged particles are evenly divided into group and copy the grid space of the finest grid in the multigrid to so that the GPU cores, charged particle groups, and grid spaces are in a one-to-one correspondence; wherein M represents the number of first parallel computing tasks corresponding to the charge deposition model, and N represents the number of GPU cores required to be called for each computing task; Each of the GPU cores deposits the charges of the charged particles in its corresponding charged particle group into the corresponding grid space through the charge deposition model; All The charges of the grid spaces are accumulated to obtain the charge on each grid node.

4. The electric propulsion plasma oscillation adaptive and efficient simulation method according to claim 1, characterized in that: The particle motion model is: ;in, represents the mass of the particle, Indicates that the particle The speed of the movement in time steps, Indicates that the particle The speed of the movement in time steps, represents the time step of particle pushing, Indicates that the particle The force at each time step, Indicates that the particle The position of the time step, Indicates that the particle The position of the time step.

5. The self-adaptive and efficient simulation method for electric propulsion plasma oscillation according to claim 1, characterized in that: The charge deposition model is: ; in, represents the grid spacing in the x direction, represents the grid spacing in the y direction, represents the grid spacing in the z direction, represents the position coordinates of the particle that undergoes charge deposition, represents the grid node coordinates, Indicates that the particle is at the grid node The charge distribution weight on , , , represents the number of grid nodes in the x direction, represents the number of grid nodes in the y direction, represents the number of mesh nodes in the z direction, Represents the macro particle weight.

6. The electric propulsion plasma oscillation adaptive and efficient simulation method according to claim 5, characterized in that: The electromagnetic field solution model is The discretized equation group consists of the objective equations of the grid nodes; Mesh Node The objective equation is expressed as: ;in, Represents a grid node The electric potential on Represents a grid node The charge density, represents the dielectric constant of vacuum.

7. The self-adaptive and efficient simulation method for electric propulsion plasma oscillation according to claim 1, characterized in that: The collision solution model includes the following solution steps: Using formula Calculate the maximum collision frequency of all particles; where, represents the atomic number density, represents the velocity of the i-th particle, , represents the kinetic energy of particle i, represents the collision cross section of particle i undergoing the jth collision, represents the maximum collision frequency of all particles; Using formula Calculate the probability of all particles having empty collisions in the current time step; where, represents the time step of particle pushing, represents the probability of empty collision for all particles in the current time step; Using formula Count the number of collisions between particles; where, represents the total number of particles, Indicates the number of collisions between particles; Random Generation A random number corresponding to particles that collide; if , then the kth colliding particle undergoes the first type of collision; among them, Indicates the random number corresponding to the kth colliding particle; represents the frequency of the jth collision of the kth particle; if , then the kth colliding particle will undergo the j+1th collision; if , then the kth colliding particle undergoes an empty collision.

8. The self-adaptive and efficient simulation method for electric propulsion plasma oscillation according to claim 7, characterized in that: After determining the number of target collisions between particles, it also includes: Constructing a three-dimensional computational domain unit model and initializing particle parameters in the unit model; wherein the particle parameters include: atomic density, electron density, ion density, electron kinetic energy, and ion kinetic energy; Disabling the particle motion model and counting the number of target collisions between particles within a time step of particle propulsion; Calculating a reaction rate error for each collision based on the number of each collision in the target collision and the particle parameters; When it is determined that a reaction rate error greater than a preset threshold exists in the target collision, the collision cross section of the corresponding collision is corrected, and / or the time step of the particle propulsion is corrected.

9. The self-adaptive and efficient simulation method for electric propulsion plasma oscillation according to claim 6, characterized in that: Before solving the electromagnetic field solution model using the multigrid method with the third adjustment parameter set, the following steps are also included: The discretized equations are rewritten as a linear equation in the form of AS=b, where S represents an array of unknowns, A represents a coefficient matrix of the unknown array S, and b represents an array of source terms. ,and .

10. An electric propulsion plasma oscillation adaptive and efficient simulation system, characterized in that: include: A construction module for constructing a high-precision electric propulsion plasma numerical simulation model; wherein the simulation model includes: a particle motion model, a charge deposition model, an electromagnetic field solution model, and a collision solution model; A decomposition module is configured to decompose the computational tasks of the target model based on a first adjustment parameter set, and to decompose the computational tasks of the electromagnetic field solution model based on a second adjustment parameter set; wherein the first adjustment parameter set includes: the number of first parallel computational tasks, the number of GPU cores required to be called for each computational task, the load workload of each GPU core, and the time step of each particle propulsion; the target model includes: the particle motion model, the charge deposition model, and the collision solution model; and the second adjustment parameter set includes: the number of second parallel computational tasks, the number of CPU cores required to be called for each computational task, and the load workload of each CPU core; A simulation module, configured to simulate and solve the simulation model to obtain simulation results; wherein the electromagnetic field solution model is solved using a multigrid method with a third set of adjustment parameters; the third set of adjustment parameters includes: a multigrid loop form, a number of multigrid layers, a number of grid smoothing times per layer, a multigrid smoothing method, and grid spacing in each direction; a processing module for performing feature extraction on the simulation results using a target neural network model to determine an updated first adjustment parameter set, an updated second adjustment parameter set, and an updated third adjustment parameter set based on the feature extraction results; wherein the target neural network model is a model obtained after training with the goal of maximizing simulation efficiency; and the feature extraction results include: electron density, electron temperature, elastic collision frequency between electrons and atoms, excitation collision frequency, ionization collision frequency, momentum exchange collision frequency and charge exchange collision frequency between ions and atoms, potential gradient, and growth rate of plasma oscillation over time for each grid node; The optimization module is used to repeatedly call the decomposition module, the simulation module and the processing module until a preset simulation cutoff condition is reached.

Images