Full-process numerical simulation method and system for Hall thruster

Abstract

The present invention relates to the field of computer-aided design, and more specifically to a full-process numerical simulation method and system for a Hall thruster. The method comprises: obtaining a target area during a current simulation, predicting a first predicted value of each grid at the next moment using a time series of simulation values ​​of each grid within the target area, recording the simulation value of each grid in the same target area at the next moment in the previous simulation as the second predicted value of each grid in the current simulation; and obtaining an evolution area during the current simulation, performing a weighted fusion of the first predicted values ​​and second predicted values ​​of all grids within the target area to obtain the simulation values ​​of all grids in the target area at the next moment. The present invention reduces simulation time while ensuring the accuracy of the simulation process.

Description

Technical Field

[0001] The present invention relates to the field of computer-aided design technology, and in particular to a full-process numerical simulation method and system for a Hall thruster. Background Art

[0002] Using computer-aided design methods to simulate the working process of a Hall thruster is a common method for designing efficient Hall thrusters. For example, the particle physics method (PIC) is used to simulate the three-dimensional model of a Hall thruster. By simulating the working process of the Hall thruster under multiple sets of parameters, the optimal Hall thruster parameters (such as discharge chamber structure, electromagnetic field distribution, etc.) with the best indicators (such as thrust, specific impulse, working efficiency, life, etc.) are obtained.

[0003] To accurately simulate the operation of a Hall thruster and obtain optimal parameters, the 3D model must be divided into a large number of grids and simulated sequentially under multiple sets of Hall thruster parameters. This requires a significant amount of computation and simulation time. While simulation time can be shortened by reducing the number of grids and parameters in the 3D model, this can lead to inaccurate simulations and an inability to obtain the globally optimal Hall thruster parameters. Summary of the Invention

[0004] In order to solve the above-mentioned problem that simulation time and accuracy cannot be achieved simultaneously, the present invention provides a full-process numerical simulation method and system for a Hall thruster.

[0005] The full-process numerical simulation method and system for Hall thrusters of the present invention adopt the following technical solutions:

[0006] One embodiment of the present invention provides a full-process numerical simulation method for a Hall thruster, the method comprising the following steps:

[0007] The parameters of the Hall thruster are changed multiple times, and a simulation is performed each time the parameters are changed; during each simulation, each grid in the three-dimensional model of the Hall thruster generates a simulation value at each moment;

[0008] During the current simulation, a target area is obtained, and the difference in simulation values ​​of all grids in the target area is minimal, and the average simulation value of all grids in the target area has the same change trend as the average simulation value of all grids in the same target area in the previous simulation; a first prediction value of each grid at the next moment is predicted using the time series of the simulation value of each grid in the target area, and the simulation value of each grid in the same target area at the next moment in the previous simulation is recorded as the second prediction value of each grid in the current simulation;

[0009] During the current simulation, an evolving region is obtained, where the evolving region is adjacent to the target region and has the largest difference from the average simulation value of the target region;

[0010] The first predicted value and the second predicted value of all grids in the target area are weightedly fused to obtain the simulated values ​​of all grids in the target area at the next moment. The direction of the maximum distribution difference of the simulated values ​​of all grids in the target area at the next moment is consistent with the direction of the maximum distribution difference of the simulated values ​​of all grids in the evolution area.

[0011] Preferably, the acquiring of the target area includes the following specific steps: the difference in simulation values ​​of all grids in the target area is minimal, and the average simulation value of all grids in the target area has the same change trend as the average simulation value of all grids in the same target area in the previous simulation process.

[0012] At the current moment in the current simulation process, K-Means clustering is performed on the feature points of all grids at the current moment to obtain several first categories; the area formed by all grids in each first category is recorded as the first area; during the current simulation process, the first average value of the simulation values ​​of all grids in the first area at each moment is obtained; during the current simulation process, the first average values ​​of all moments before the current moment constitute the first sequence of each first area;

[0013] During the last simulation, a second average value of simulation values ​​of all grids in each first region is obtained, wherein the second average values ​​of all moments before the current moment in the last simulation constitute a second sequence of each first region;

[0014] Obtain the trend similarity between the first sequence and the second sequence of each first region. When the trend similarity is greater than or equal to a first preset threshold, determine that the average simulation value of all grids in the first region has the same change trend as the average simulation value of all grids in the same region in the previous simulation process, and record the first region as the target region.

[0015] Preferably, the acquiring of the evolution region, wherein the evolution region is adjacent to the target region and has the largest difference from the average simulation value of the target region, comprises the following specific steps:

[0016] In the current simulation process, in all first regions, all first regions adjacent to each target region are obtained and recorded as target adjacent regions; and the average simulation value of all grids in each target adjacent region is obtained;

[0017] The difference between the average simulation value of the target area and the average simulation value of the target adjacent areas is obtained, and several target adjacent areas with the largest differences are recorded as the evolution areas of each target area.

[0018] Preferably, the weighted fusion of the first predicted values ​​and the second predicted values ​​of all grids in the target area to obtain the simulated values ​​of all grids in the target area at the next moment includes the following specific steps:

[0019] Several groups of fusion weights are preset. For each group of fusion weights (w1, w2), the first predicted values ​​and the second predicted values ​​of all grids in the target area are weighted and summed using w1 and w2 respectively. w1 is the weighted value of the first predicted values ​​of all grids, and w2 is the weighted value of the second predicted values ​​of all grids to obtain the comprehensive predicted simulation value of all grids.

[0020] Obtain the maximum distribution difference direction of the comprehensive prediction simulation values ​​of all grids in the target area, which is simply recorded as the first direction; the maximum distribution difference direction of the simulation values ​​of all grids in the evolution area is simply recorded as the second direction; use the similarity of the first direction and the second direction to obtain the selection index of each group of fusion weights; obtain a group of fusion weights with the largest selection index, use this group of fusion weights to perform weighted summation of the first prediction value and the second prediction value of all grids, and the obtained comprehensive prediction simulation value of all grids is used as the simulation value of all grids in the target area at the next moment in the current simulation process.

[0021] Preferably, the specific steps for obtaining the maximum distribution difference direction of the simulation values ​​of all grids in the evolution area are as follows:

[0022] The K-Means algorithm is used to cluster the feature points of all grids in the evolution area to obtain several second categories. The average simulation value in each second category is obtained, the difference between the average simulation values ​​of any two second categories is obtained, and several pairs of second categories with the largest differences are obtained. The direction of the displacement vector between the category center points of several pairs of second categories is taken as the direction of the maximum distribution difference of the simulation values ​​of all grids in the evolution area.

[0023] Preferably, the specific steps of obtaining the trend similarity between the first sequence and the second sequence of each first region are as follows:

[0024] The first and second sequences are linearly normalized respectively, and the DTW distance between the first and second sequences after linear normalization is denoted as x, and exp(-x) is denoted as the trend similarity, where exp() represents an exponential function with a natural constant as the base.

[0025] Preferably, the characteristic point of the grid is a vector formed by combining the coordinates of the center point of the grid and the simulation value of the grid.

[0026] Preferably, the method of obtaining the selection index of each group of fusion weights by using the similarity between the first direction and the second direction includes the following specific steps:

[0027] Perform KM matching on all first directions and all second directions to obtain several matching pairs. Each matching pair contains a first direction and a second direction. Calculate the cosine similarity between the first direction and the second direction of each matching pair. The sum of the cosine similarities of all matching pairs is recorded as the selection index of each group of fusion weights.

[0028] Preferably, the parameters include the inner diameter and outer diameter of the annular discharge chamber of the Hall thruster at different positions, the current of the coil, the voltage between the accelerating electrodes, the propellant flow density of the particle emitter, and the power of the Hall thruster.

[0029] Another embodiment of the present invention provides a full-process numerical simulation system for a Hall thruster, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the full-process numerical simulation method for a Hall thruster are implemented as described above.

[0030] The beneficial effects of the technical solution of the present invention are:

[0031] The present invention performs a weighted fusion of the first and second predicted values ​​for all grids within the target area to obtain the simulated values ​​for all grids within the target area at the next moment. This process fuses the simulation results of the previous simulation process with the simulation results of the current simulation process to directly obtain the simulated values ​​for the next moment of the current simulation process. This avoids the computational complexity and long calculation time associated with simulating a large number of grids using the particle granularity method (PIC). Furthermore, it enables accurate simulation results for multiple sets of Hall effect thruster parameters to be obtained in a shorter timeframe (the greater the number of grids, the more accurate the simulation).

[0032] Among them, the first prediction value is obtained by predicting the time series of the simulation value of each grid in the target area, the difference in the simulation values ​​of all grids in the target area is minimal, and the average simulation value of all grids in the target area has the same change trend as the average simulation value of all grids in the same target area in the previous simulation process; this process takes into account that since the difference between the simulation values ​​of all grids in the target area is relatively small, the collision, ionization, acceleration and other processes of the particles in the grids in the target area may not be disturbed by the collision, ionization, acceleration and other processes of the particles in other grids. Therefore, in the current simulation process, the time series of the simulation value of each grid in the target area is used to predict the first prediction value of each grid at the next moment, and the first prediction value provides an accurate reference value for the simulation value of each grid at the next moment.

[0033] Among them, since the evolving region is adjacent to the target region and has the largest average simulation value with the target region, there is an obvious correlation between the target region and the evolving region. For example, due to the collision, ionization, and acceleration of particles, there is particle exchange, charge offset, and evolution of particle velocity distribution between the target region and the evolving region.

[0034] Based on the evolution region, the direction of the maximum distribution difference in the simulated values ​​of all grids within the target region at the next moment is further aligned with the direction of the maximum distribution difference in the simulated values ​​of all grids within the evolution region. This process ensures that, from a larger perspective (i.e., the larger local area formed by the target region and the evolution region), the direction of changes in particle velocity, charge, ionization, and other factors will not suddenly change within a short period of time. By aligning the predicted direction of these changes with the direction of changes in the evolution region, the consistency of the change direction over a larger range (i.e., the absence of sudden changes) is guaranteed, thereby ensuring the accuracy of the predicted simulation values.

[0035] In summary, the present invention reduces simulation time while ensuring the accuracy of the simulation process, making it possible to efficiently obtain Hall thruster parameters that meet design expectations using the method of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only 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.

[0037] Figure 1 A flowchart of the steps of a full-process numerical simulation method for a Hall thruster provided by one embodiment of the present invention;

[0038] Figure 2 A first design solution for the discharge chamber of a Hall thruster provided in one embodiment of the present invention;

[0039] Figure 3 A second design solution for the discharge chamber of a Hall thruster provided in one embodiment of the present invention;

[0040] Figure 4 This is a third design scheme for the discharge chamber of a Hall thruster provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0041] To further illustrate the technical means and effectiveness of the present invention in achieving its intended objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, details the implementation, structure, features, and effectiveness of the full-process numerical simulation method and system for Hall thrusters proposed in accordance with the present invention. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0042] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0043] The specific scheme of the full-process numerical simulation method and system for Hall thrusters provided by the present invention is described in detail below with reference to the accompanying drawings.

[0044] Example 1:

[0045] See also Figure 1 , which shows a flowchart of a full-process numerical simulation method for a Hall thruster provided by an embodiment of the present invention, the method comprising the following steps:

[0046] Step S101 : performing a full-process numerical simulation using a three-dimensional model of the Hall thruster. During the simulation, each grid in the three-dimensional model of the Hall thruster generates a simulation value at each moment.

[0047] Hall thrusters use electromagnetic fields to ionize and accelerate neutral particles (such as xenon atoms) to generate thrust. Parameters such as the structure (e.g., the inner and outer diameters of the annular discharge chamber at different locations) and the electromagnetic field distribution (e.g., the coil position and current at different locations) of the Hall thruster significantly influence its thrust, specific impulse, efficiency, and lifespan.

[0048] In order to design a Hall thruster with better indicators, it is necessary to simulate the operation of Hall thrusters with different parameters and then screen out the Hall thruster parameters that optimize the above indicators.

[0049] This example first constructs a three-dimensional model of a Hall thruster. This model includes a ring-shaped discharge chamber, accelerating electrodes, and coils. The coils generate a magnetic field within the discharge chamber, confining ions within it. The accelerating electrodes, consisting of a cathode and an anode, generate an electric field within the discharge chamber to accelerate the ions. Furthermore, a particle emitter is included, generating propellant. The propellant consists of neutral particles and electrons, which collide, ionize, and accelerate within the electromagnetic field of the discharge chamber.

[0050] The inner and outer diameters of the annular discharge chamber at different positions (i.e., different cross-sections) in the three-dimensional model, the position of the coil, the current magnitude (or magnetic field magnitude) of the coil at different positions, the voltage (or electric field magnitude) between the accelerating electrodes, and the propellant flow density of the particle emitter are used as a set of parameters of the Hall thruster.

[0051] As an optional example, multiple sets of parameters are randomly generated. The corresponding three-dimensional Hall thruster model for each set of parameters can simulate corresponding Hall thruster indicators. Different sets of parameters may correspond to different indicators. In this embodiment, the set of parameters that achieves optimal Hall thruster indicators (e.g., maximum thrust and lifespan, or maximum operating efficiency) is used as the optimal design parameters for the Hall thruster.

[0052] As a preferred example, several groups of parameters are set manually (for example, experts set several groups of parameters based on experience).

[0053] Obtaining an optimal set of parameters by using the indicators simulated under different sets of parameters is a conventional technique and is not the focus of this embodiment, so it will not be described in detail in this embodiment.

[0054] In this embodiment, the three-dimensional model of the Hall thruster is divided into several grids (specifically, the discharge chamber of the Hall thruster is divided into several grids). Each grid is a cube, and the volumes of different grids are the same. In this embodiment, the volume of each grid is 0.08 cubic millimeters.

[0055] This embodiment uses the particle-interconnected mesh (PIC) method to simulate a three-dimensional model of a Hall thruster under a given set of parameters. This method tracks the motion trajectories of a large number of particles (neutral atoms, ions, and electrons) and counts collision events within each grid at each moment. Ultimately, it determines the number of positively charged ions, negatively charged ions, and neutral particles in each grid, as well as particle density and the average velocity of all particles. This embodiment concatenates this data within each grid into a vector, which is recorded as a simulated value generated for each grid at each moment.

[0056] When the simulation process continues for a period of time (for example, one hour), the specific impulse, thrust, working efficiency, life and other indicators of the Hall thruster are simulated.

[0057] The particle granularity analysis (PIC) method is a well-known technique and will not be described in detail in this embodiment.

[0058] In other embodiments, only the particle density or the average velocity of the particles may be used as the simulation value. In still other embodiments, the ratio of the number of positive ions to the number of negative ions to the number of all particles may be recorded as the ionization rate, and the ionization rate may be used as the simulation value.

[0059] As an optional example, one second is a moment, and in other embodiments, one 5-second moment may be a moment. It should be noted that, in each simulation, the moment after all grids obtain simulation values ​​for the first time is recorded as the first moment (ie, the starting moment).

[0060] As a preferred example, considering that each simulation is computationally intensive (especially when there are many grids), it's impossible to obtain the simulated value for each grid in real time. For example, after all grids have simulated their values ​​at the current time point, it may take some time (e.g., 2 to 4 seconds) for the next corresponding simulated value for all grids to be simulated. Therefore, in this example, the time when all grids have simulated their values ​​is recorded as one moment. For example, the simulated value of each grid at moment i in this example actually represents the simulated value obtained after the i-th simulation of each grid during the same simulation process.

[0061] At this point, given a set of Hall thruster parameters, the simulation values ​​of each grid at different times can be obtained.

[0062] Step S102: Continuously change the parameters of the Hall thruster three-dimensional model. During the current simulation, obtain a target area, wherein the difference in simulation values ​​of all grids within the target area is minimal, and the average simulation value of all grids within the target area has the same change trend as the average simulation value of all grids within the same target area during the previous simulation.

[0063] In order to obtain simulation results under different parameters, this implementation continuously changes the parameters of the Hall thruster three-dimensional model multiple times. A full-process numerical simulation is performed after each parameter change. During each simulation, each grid in the Hall thruster three-dimensional model generates a simulation value at each moment.

[0064] This step takes into account the fact that when there are many sets of parameters or an excessive number of grids, a large amount of computational effort and time is required. In other words, it's impossible to accurately obtain simulation results for multiple sets of parameters within a limited timeframe (a larger number of grids improves accuracy). This, in turn, makes it impossible to guarantee that the simulated indicators are globally optimal (a larger number of grids and parameter sets guarantee global optimization). Consequently, the designed Hall effect thruster's thrust, specific impulse, lifespan, and other indicators fall short of expectations.

[0065] Therefore, this step improves the calculation speed of the simulation process by combining two adjacent simulation processes, so that as many parameters and grids as possible can be used to obtain the parameters with the best indicators within a limited time.

[0066] The target area obtained in this embodiment has the same or similar change trend as that of the previous simulation process, so the simulation results of the previous simulation process (ie, the simulation value of each grid) can be used to assist in obtaining the simulation results in the current simulation process.

[0067] In particular, when there is no previous simulation process, the particle granularity method (PIC) is used to simulate the current process.

[0068] As an example, in the current simulation process, a target area is obtained, the difference in simulation values ​​of all grids in the target area is minimal, and the average simulation value of all grids in the target area has the same change trend as the average simulation value of all grids in the same target area in the previous simulation process, including the following methods:

[0069] At the current moment in the simulation, the simulated value for each grid is obtained. The coordinates of each grid (the coordinates of the grid's center point) and the simulated values ​​at each and the previous moment are concatenated into a vector, which is recorded as the feature point of each grid. K-Means clustering is performed on the feature points of all grids at the current moment. The number of resulting categories is set to K1, and each resulting category is recorded as the first category. Grids in the same category are spatially connected and have minimal differences in their simulated values.

[0070] In order to eliminate the difference in dimensions of vectors corresponding to feature points, this embodiment uses the ZCA algorithm to perform whitening processing on the feature points of all grids.

[0071] In this embodiment, K1 is equal to the number of all grids divided by 20 (rounded up). In other embodiments, K1 may be set to other values, which are not specifically limited in this embodiment.

[0072] The region formed by all grids in each first category is called the first region. During the current simulation, the first average of the simulated values ​​of all grids in the first region at each time instant is obtained. During the current simulation, the first average values ​​of all times up to (including) the current time instant constitute the first sequence for each first region.

[0073] In particular, if there are less than 5 moments before the current moment, all the methods of this embodiment are not executed, but the particle granularity method (PIC) is directly used to simulate the three-dimensional model of the Hall thruster.

[0074] During the last simulation, the second average value of the simulation values ​​of all grids in each first region is obtained. During the last simulation, the second average values ​​of all moments before the current moment constitute a second sequence for each first region.

[0075] Obtain the trend similarity between the first and second sequences for each first region. When the trend similarity is greater than or equal to a first preset threshold th1, determine that the average simulated value of all grids in the first region has the same trend as the average simulated value of all grids in the same region during the previous simulation. In this case, the first region is marked as the target region.

[0076] This embodiment is described by taking th1=0.6 as an example. In other embodiments, th1 may be set to other values, which is not specifically limited in this embodiment.

[0077] Specifically, if the number of grids in the first region is less than 10, the first region is deleted.

[0078] So far, several target areas have been obtained.

[0079] As another example, during the current simulation process, a target area is obtained, the difference in simulation values ​​of all grids in the target area is minimal, and the average simulation value of all grids in the target area has the same change trend as the average simulation value of all grids in the same target area in the previous simulation process, including the following methods:

[0080] All first regions whose trend similarity is less than the first preset threshold th1 are recorded as first regions to be merged.

[0081] For any two adjacent first regions to be merged (that is, first regions to be merged with a common boundary), the two first regions to be merged are merged into one region, which is recorded as the first region. Then, the trend similarity of the first region is calculated again using the method of the above example, and recorded as the merged trend similarity of any two adjacent first regions to be merged.

[0082] The first regions to be merged whose merging trend similarity is greater than or equal to th1 are merged, and the merged region is also recorded as the target region.

[0083] In particular, when the similarity of the merging trend between any first region to be merged and multiple other first regions to be merged is greater than or equal to th1, the two first regions to be merged with the greatest similarity of the merging trend are preferably merged to obtain the target region.

[0084] As an example, a method for obtaining trend similarity between the first sequence and the second sequence includes:

[0085] If the simulated values ​​in the first sequence and the second sequence are vectors, PCA is performed to reduce the simulated values ​​in the first sequence and the second sequence to one dimension, so that the values ​​in the first sequence and the second sequence are one-dimensional.

[0086] The first and second sequences are linearly normalized. The DTW distance between the first and second sequences after linear normalization is denoted as x. Exp(-x) is denoted as the trend similarity, where exp() represents an exponential function with a natural constant as the base. The DTW distance is obtained using the well-known DTW algorithm.

[0087] In other embodiments, the cosine similarity between the first sequence and the second sequence after linear normalization may be recorded as trend similarity.

[0088] Step S103: During the current simulation, the time series of the simulation values ​​of each grid in the target area is used to predict the first prediction value of each grid at the next moment, and the simulation value of each grid in the same target area at the next moment in the previous simulation is recorded as the second prediction value of each grid in the current simulation.

[0089] Since the differences between the simulation values ​​of all grids in the target area are relatively small, the collision, ionization, acceleration and other processes of the particles in the grids in the target area may not be interfered with by the collision, ionization, acceleration and other processes of the particles in other grids. Therefore, in the current simulation process, the time series sequence of the simulation values ​​of each grid in the target area is used to predict the first predicted value of each grid at the next moment, and the first predicted value can be used as a reference for the simulation value of each grid at the next moment.

[0090] The time series of simulation values ​​of each grid specifically refers to a time series consisting of simulation values ​​of all moments within a preset time period (eg, within 5 moments) before the current moment in the current simulation process.

[0091] As an example, the first predicted value of each grid at the next moment is predicted, including the following method:

[0092] The Kalman filter algorithm is used to make predictions to obtain a first prediction value at the next moment. In other embodiments, the Kalman filter algorithm may be replaced with an extended Kalman filter algorithm.

[0093] As another example, the first predicted value of each grid at the next moment is predicted, including the following method:

[0094] For any of the multiple historical simulations, after obtaining a first region, the time series of simulated values ​​for each grid within the first region is used as a sample. When a sufficient number of samples is obtained (for example, greater than 300), these samples are combined into a dataset, which is used to train an LSTM neural network. The LSTM neural network takes the time series of simulated values ​​as input and outputs the simulated value at the next moment.

[0095] For the time series of simulation values ​​of each grid obtained in the current simulation process, the trained LSTM neural network is used to make predictions to obtain the first predicted value at the next moment.

[0096] The structure, training, and use methods of the LSTM neural network are well known and will not be described in detail in this embodiment.

[0097] Furthermore, since the average simulation value of all grids in the target area has the same changing trend as the average simulation value of all grids in the same target area in the previous simulation process, it means that the changing law of the grid simulation value in the current simulation process is the same as the changing law of the previous simulation process. Therefore, the simulation value of the grid at the next moment in the previous simulation process can be used to calculate the simulation value of the grid at the next moment in the current simulation process, that is: the simulation value of each grid in the same target area in the previous simulation process at the next moment is recorded as the second predicted value of each grid in the current simulation process, and the second predicted value can also be used as a reference for the simulation value of each grid at the next moment.

[0098] Step S104 : During the current simulation process, an evolving region is obtained, where the evolving region is adjacent to the target region and has the largest average simulation value with the target region.

[0099] Since the evolving region is adjacent to the target region and has the largest average simulation value with the target region, there is an obvious correlation between the target region and the evolving region. For example, due to particle collision, ionization, and acceleration, there is particle exchange, charge offset, and evolution of particle velocity distribution between the target region and the evolving region.

[0100] As an example, the method for obtaining the evolution region includes:

[0101] In the current simulation process, in all first areas, all first areas adjacent to each target area are obtained and recorded as target adjacent areas; the average value of all simulation values ​​in each target adjacent area is obtained, that is, the average simulation value.

[0102] For the target adjacent regions of each target region, the difference between the average simulation value of the target region and the average simulation value of the target adjacent regions is obtained, and the N1 target adjacent regions with the largest difference are recorded as the evolution regions of each target region.

[0103] In this embodiment, N1 is half the number of target adjacent areas (rounded up). In other embodiments, N1 may be set to other values, for example, N1=1.

[0104] The difference between the average simulated values ​​refers to the absolute value of the difference. If the simulated values ​​are vectors, the difference between the average simulated values ​​refers to the Euclidean distance.

[0105] Step S105: perform weighted fusion on the first predicted values ​​and the second predicted values ​​of all grids in the target area to obtain the simulation values ​​of all grids in the target area at the next moment. The maximum distribution difference direction of the simulation values ​​of all grids in the target area at the next moment is consistent with the maximum distribution difference direction of the simulation values ​​of all grids in the evolution area.

[0106] The first predicted value and the second predicted value of all grids in the target area are weightedly fused to obtain the simulation value of all grids in the target area at the next moment. The process fuses the simulation results of the previous simulation process with the simulation results of the current simulation process to directly obtain the simulation value of the next moment of the current simulation process, avoiding the problems of large amount of calculation and long calculation time when simulating a large number of grids through the particle grid method (PIC). On the other hand, it enables accurate simulation results to be obtained in a shorter time (the more grids, the more accurate).

[0107] Furthermore, the direction of the maximum distribution difference of the simulation values ​​at the next moment obtained by the fusion of this embodiment and the direction of the maximum distribution difference of the simulation values ​​of all grids in the evolution area represent the direction of spread of changes in the speed, charge, ionization conditions, etc. of particles in the local area.

[0108] In this embodiment, the direction of the maximum distribution difference in the simulated values ​​of all grids in the target region at the next moment is consistent with the direction of the maximum distribution difference in the simulated values ​​of all grids in the evolution region. In this process, from a larger scope (i.e., the larger local area formed by the target region and the evolution region), the particle velocity, charge, ionization state, and other factors will not suddenly change in a short period of time when the direction of change propagates. Therefore, by aligning the predicted direction of change with the direction of change propagates in the evolution region, this process ensures consistency of the direction of change propagates over a larger scope (i.e., no sudden changes), thereby ensuring the accuracy of the predicted simulated values.

[0109] As an example, the specific method for obtaining the maximum distribution difference direction of the simulation values ​​of all grids in the evolution area is:

[0110] For all the feature points of the grid in the evolution area, the PCA algorithm is used to obtain the principal component directions of all the feature points, and these principal component directions are used as the directions of the maximum distribution differences of the simulation values.

[0111] As another example, the specific method for obtaining the maximum distribution difference direction of the simulation values ​​of all grids in the evolution area is:

[0112] For all feature points of the grids within the evolution area, the K-Means algorithm is used again to cluster all feature points. The resulting categories are recorded as second categories, and the number of second categories obtained is set to N2. This embodiment is described using N2=10 as an example. The mean of all feature points in each second category is recorded as the category center point of the second category. The average simulation value within each second category is obtained, the difference between the average simulation values ​​of any two second categories is obtained, and several pairs (for example, 5 pairs) of second categories with the largest differences are obtained. The 5 pairs of vectors formed between the category center points of these 5 pairs of second categories are also displacement vectors between the 5 pairs of category center points. In this embodiment, the direction of the displacement vector is from the feature point with a smaller modulus length to the feature point with a larger modulus length; the direction of these 5 pairs of vectors is used as the direction of the maximum distribution difference of the simulation value.

[0113] Compared with the above examples, the maximum distribution difference directions obtained in this example are not limited to orthogonal directions, and the number of maximum distribution difference directions can be set manually, so the amount of calculation is relatively small.

[0114] The direction of the maximum distribution difference of the simulation values ​​obtained in all examples is expressed as a unit vector in this direction (denoted as the second direction).

[0115] As a preferred example, the first predicted values ​​and the second predicted values ​​of all grids in each target area are weightedly fused to obtain the simulated values ​​of all grids in the target area at the next moment, including the following method:

[0116] Given several sets of fusion weights, each set of fusion weights is represented by w1 and w2, w1 + w2 = 1. For example, given 10 sets of fusion weights, (w1, w2) in each set of fusion weights are (0.0, 1.0), (0.1, 0.9), (0.2, 0.8), ..., (1.0, 0) respectively.

[0117] For each set of fusion weights (w1, w2), the first prediction values ​​and the second prediction values ​​of all grids are weighted and summed using w1 and w2 respectively, w1 is the weighted weight of the first prediction values ​​of all grids, and w2 is the weighted weight of the second prediction values ​​of all grids, thereby obtaining the comprehensive prediction simulation value of all grids.

[0118] Obtain the direction of maximum distribution difference of the simulated values ​​of the integrated predictions for all grids within the target area (using the same method as in the previous example). Express this direction as a unit vector, referred to as the first direction. The unit vector corresponding to the direction of maximum distribution difference of the simulated values ​​for all grids within the evolution area is referred to as the second direction.

[0119] Perform KM matching on all first directions and all second directions to obtain several matching pairs. Each matching pair contains a first direction and a second direction. For each matching pair, a cosine similarity is calculated between the first direction and the second method. The sum of the absolute values ​​of the cosine similarities of all matching pairs obtained by the KM algorithm is the largest.

[0120] In this embodiment, the sum of the absolute values ​​of the cosine similarities of all matching pairs is recorded as the selection index for each set of fusion weights. The set of fusion weights with the largest selection index is obtained. The first and second predicted values ​​of all grids are weighted summed using this set of fusion weights. The resulting comprehensive predicted simulation value for all grids is used as the simulation value for all grids in the target area at the next moment in the current simulation process.

[0121] At this time, the direction of the maximum distribution difference of the simulation values ​​of all grids in the target area at the next moment is consistent with the direction of the maximum distribution difference of the simulation values ​​of all grids in the evolution area.

[0122] This example is relatively computationally light.

[0123] As another optional example, the first predicted values ​​and the second predicted values ​​of all grids in each target area are weightedly fused to obtain the simulated values ​​of all grids in the target area at the next moment, including the following methods:

[0124] A set of fusion weights are randomly initialized, and the first predicted values ​​and the second predicted values ​​of all grids in each target area are weightedly summed using the fusion weights to obtain the comprehensive predicted simulation value of all grids in each target area. The selection index of the set of fusion weights is obtained according to the maximum distribution difference direction of the comprehensive predicted simulation values ​​of all grids in the target area and the maximum distribution difference direction of the simulation values ​​of all grids in the evolution area. The simulated annealing algorithm is used to obtain the fusion weight when the selection index is maximum. The first predicted values ​​and the second predicted values ​​of all grids are weightedly summed using the set of fusion weights. The obtained comprehensive predicted simulation value of all grids is used as the simulation value of all grids in the target area at the next moment in the current simulation process.

[0125] At this time, the direction of the maximum distribution difference of the simulation values ​​of all grids in the target area at the next moment is consistent with the direction of the maximum distribution difference of the simulation values ​​of all grids in the evolution area.

[0126] Although this example has a large amount of calculation, the simulation value obtained at the next moment is more accurate.

[0127] At this point, in the current simulation process, the simulation value of each grid in the target area at the next moment is obtained, reducing the computational complexity of the simulation process. For each grid outside the target area, the particle physics (PIC) method is still used to simulate the simulation value of the next moment.

[0128] Repeating steps S102 to S105 of this embodiment ensures that the simulated values ​​for all grids within the target area at the next moment are obtained by fusing the simulated values ​​from the previous simulation process with the simulated values ​​already simulated in the current simulation process. This process eliminates the need to simulate the values ​​of each grid using the particle physics (PIC) method at each moment, reducing the computational effort. This allows for accurate simulation results (i.e., the simulated values ​​for each grid at each moment) to be obtained using a larger number of grids and a larger number of Hall thruster parameter sets within a limited timeframe. This ensures more accurate Hall thruster metrics (such as specific impulse, thrust, and lifespan) obtained through simulation, thereby ensuring that optimal Hall thruster parameters are obtained, leading to the design of a Hall thruster that meets expectations. At the same time, in the specific process of this embodiment, the first predicted value and the second predicted value of all grids in the target area are weightedly fused, and the maximum distribution difference direction of the simulation values ​​of all grids in the target area at the next moment is consistent with the maximum distribution difference direction of the simulation values ​​of all grids in the evolution area; the maximum distribution difference direction represents the change and spread direction of the speed, charge, ionization condition, etc. of particles in the local area; this process ensures the consistency of the change and spread direction over a larger range (that is, there is no mutation) by making the predicted change and spread direction consistent with the change and spread direction of the evolution area, thereby ensuring the accuracy of the predicted simulation value.

[0129] This concludes the present embodiment.

[0130] Example 2:

[0131] To further ensure the accuracy of the simulation process, this embodiment uses the method of Example 1 to obtain the simulated values ​​of all grids in the target area at the next moment. Then, the particle granular mesh method (PIC) is used to simulate the simulated values ​​of all grids at subsequent moments. After the particle granular mesh method (PIC) has simulated the simulated values ​​of all grids in the target area at subsequent moments, the method of Example 1 is used again to obtain the simulated values ​​of all grids in the target area at subsequent moments. This process is repeated until the corresponding simulation process for all parameters is completed. This process balances calculation speed and simulation accuracy by alternately using the particle granular mesh method (PIC) and the method of Example 1 to obtain the simulated values ​​of all grids.

[0132] Example 3:

[0133] The Hall thruster developed in this embodiment is required to have higher working efficiency than the existing EHT-400 thruster.

[0134] The greater the propellant flow density (propellant flow rate / discharge chamber cross-sectional area) of a Hall thruster, the closer the plasma discharge region (i.e., the area where the particles exhibit their electrical center) is to the discharge chamber nozzle. At this point, the interaction between the plasma (i.e., the group of particles containing charged ions) and the discharge chamber walls decreases, the less resistance to ion ejection, and the higher the Hall thruster's efficiency. As the power of the Hall thruster decreases, the propellant flow density decreases. To maintain the plasma discharge, the plasma discharge region gradually moves inward into the discharge chamber. At this point, the interaction between the plasma and the discharge chamber walls gradually intensifies, increasing the resistance to ion ejection from the walls and gradually reducing the efficiency of the Hall thruster.

[0135] In order to suppress the phenomenon that the plasma discharge region shrinks into the discharge chamber when the power of the Hall thruster is reduced and the propellant flow rate is reduced, the multiple sets of parameters of the Hall thruster provided in this embodiment include three sets of parameters. These three sets of parameters specifically refer to three annular discharge chambers with different structures. The three annular discharge chambers with different structures have different inner and outer diameters at different positions (i.e., different cross-sections). Specifically, Figure 2 、 Figure 3 、 Figure 4 The three sets of parameters differ only in Figure 2 、 Figure 3 、 Figure 4 The structure shown.

[0136] Figure 3 、 Figure 4 Compared with the designed discharge chamber Figure 2 The difference is as follows: a variable cross-section area is constructed near the nozzle of the Hall thruster discharge chamber, so that the cross-sectional area of ​​the discharge channel gradually decreases from the outside to the inside.

[0137] After simulating all the parameters in Example 1, the Figure 4 A set of parameters for the structure shown in , compared to the one containing Figure 2 、 Figure 3 The parameters of the structure shown have higher working efficiency (compared to Figure 2 Work efficiency increased by 11.8 times).

[0138] Example 4:

[0139] This embodiment provides a full-process numerical simulation system for a Hall thruster, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the full-process numerical simulation method for a Hall thruster described in the first embodiment are implemented.

[0140] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A full-process numerical simulation method for Hall thrusters, characterized in that: The method comprises the following steps: The parameters of the Hall thruster are changed multiple times, and a simulation is performed each time the parameters are changed; during each simulation, each grid in the three-dimensional model of the Hall thruster generates a simulation value at each moment; During the current simulation, a target area is obtained, and the difference in simulation values ​​of all grids in the target area is minimal, and the average simulation value of all grids in the target area has the same change trend as the average simulation value of all grids in the same target area in the previous simulation; a first prediction value of each grid at the next moment is predicted using the time series of the simulation value of each grid in the target area, and the simulation value of each grid in the same target area at the next moment in the previous simulation is recorded as the second prediction value of each grid in the current simulation; During the current simulation, an evolving region is obtained, where the evolving region is adjacent to the target region and has the largest difference from the average simulation value of the target region; The first predicted value and the second predicted value of all grids in the target area are weightedly fused to obtain the simulated values ​​of all grids in the target area at the next moment. The direction of the maximum distribution difference of the simulated values ​​of all grids in the target area at the next moment is consistent with the direction of the maximum distribution difference of the simulated values ​​of all grids in the evolution area.

2. The full-process numerical simulation method for Hall thruster according to claim 1, characterized in that: The target area is obtained, the difference in simulation values ​​of all grids in the target area is minimal, and the average simulation value of all grids in the target area has the same change trend as the average simulation value of all grids in the same target area in the previous simulation process, including the following specific steps: At the current moment in the current simulation process, K-Means clustering is performed on the feature points of all grids at the current moment to obtain several first categories; the area formed by all grids in each first category is recorded as the first area; during the current simulation process, the first average value of the simulation values ​​of all grids in the first area at each moment is obtained; during the current simulation process, the first average values ​​of all moments before the current moment constitute the first sequence of each first area; During the last simulation, a second average value of simulation values ​​of all grids in each first region is obtained, wherein the second average values ​​of all moments before the current moment in the last simulation constitute a second sequence of each first region; Obtain the trend similarity between the first sequence and the second sequence of each first region. When the trend similarity is greater than or equal to a first preset threshold, determine that the average simulation value of all grids in the first region has the same change trend as the average simulation value of all grids in the same region in the previous simulation process, and record the first region as the target region.

3. The full-process numerical simulation method for Hall thruster according to claim 2, characterized in that: The obtaining of the evolution region, wherein the evolution region is adjacent to the target region and has the largest difference from the average simulation value of the target region, comprises the following specific steps: In the current simulation process, in all first regions, all first regions adjacent to each target region are obtained and recorded as target adjacent regions; and the average simulation value of all grids in each target adjacent region is obtained; The difference between the average simulation value of the target area and the average simulation value of the target adjacent areas is obtained, and several target adjacent areas with the largest differences are recorded as the evolution areas of each target area.

4. The full-process numerical simulation method for Hall thruster according to claim 1, characterized in that: The weighted fusion of the first predicted values ​​and the second predicted values ​​of all grids in the target area to obtain the simulated values ​​of all grids in the target area at the next moment includes the following specific steps: Several groups of fusion weights are preset. For each group of fusion weights (w1, w2), the first predicted values ​​and the second predicted values ​​of all grids in the target area are weighted and summed using w1 and w2 respectively. w1 is the weighted value of the first predicted values ​​of all grids, and w2 is the weighted value of the second predicted values ​​of all grids to obtain the comprehensive predicted simulation value of all grids. Obtain the maximum distribution difference direction of the comprehensive prediction simulation values ​​of all grids in the target area, which is simply recorded as the first direction; the maximum distribution difference direction of the simulation values ​​of all grids in the evolution area is simply recorded as the second direction; use the similarity of the first direction and the second direction to obtain the selection index of each group of fusion weights; obtain a group of fusion weights with the largest selection index, use this group of fusion weights to perform weighted summation of the first prediction value and the second prediction value of all grids, and the obtained comprehensive prediction simulation value of all grids is used as the simulation value of all grids in the target area at the next moment in the current simulation process.

5. The full-process numerical simulation method for Hall thruster according to claim 1, characterized in that: The specific steps for obtaining the maximum distribution difference direction of the simulation values ​​of all grids in the evolution area are as follows: The K-Means algorithm is used to cluster the feature points of all grids in the evolution area to obtain several second categories. The average simulation value in each second category is obtained, the difference between the average simulation values ​​of any two second categories is obtained, and several pairs of second categories with the largest differences are obtained. The direction of the displacement vector between the category center points of several pairs of second categories is taken as the direction of the maximum distribution difference of the simulation values ​​of all grids in the evolution area.

6. The full-process numerical simulation method for Hall thruster according to claim 2, characterized in that: The specific steps of obtaining the trend similarity between the first sequence and the second sequence of each first region are as follows: The first and second sequences are linearly normalized respectively, and the DTW distance between the first and second sequences after linear normalization is denoted as x, and exp(-x) is denoted as the trend similarity, where exp() represents an exponential function with a natural constant as the base.

7. The full-process numerical simulation method for Hall thruster according to claim 2 or 5, characterized in that: The characteristic points of the grid are: the coordinates of the center point of the grid and the vector formed by splicing the simulation values ​​of the grid at adjacent moments.

8. The full-process numerical simulation method for Hall thruster according to claim 4, characterized in that: The specific steps of obtaining the selection index of each group of fusion weights by using the similarity between the first direction and the second direction are as follows: Perform KM matching on all first directions and all second directions to obtain several matching pairs. Each matching pair contains a first direction and a second direction. Calculate the cosine similarity between the first direction and the second direction of each matching pair. The sum of the cosine similarities of all matching pairs is recorded as the selection index of each group of fusion weights.

9. The full-process numerical simulation method for Hall thruster according to claim 8, characterized in that: The parameters include the inner diameter and outer diameter of the annular discharge chamber of the Hall thruster at different positions, the current of the coil, the voltage between the accelerating electrodes, and the propellant flow density of the particle emitter.

10. A full-process numerical simulation system for a Hall thruster, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the full-process numerical simulation method for a Hall thruster according to any one of claims 1 to 9 are implemented.

Images