Space debris cloud orbit evolution prediction method and system based on atmospheric resistance and group dynamics

Abstract

The invention relates to a space debris cloud orbit evolution prediction method and system based on atmospheric resistance and group dynamics, and the method comprises the steps: obtaining an initial orbit parameter of a target debris cloud, carrying out the subclass division of a debris group according to the initial orbit parameter, and simulating and generating a random initial state of the debris group; and when the random initial state of the fragment group is simulated and generated, the orbit evolution trajectory of each subclass fragment is calculated, and a three-dimensional orbit evolution result is generated. According to the method, the crossing of debris cloud evolution from static homogeneous prediction to dynamic heterogeneous simulation is realized, and a high-credibility time sequence planning basis is provided for active removal of space debris.

Description

Technical Field

[0001] The present invention relates to the technical field of space environment management, and particularly to a method and system for predicting the orbital evolution of space debris clouds based on atmospheric drag and swarm dynamics. Background Art

[0002] In recent years, the number of spacecraft launches has increased exponentially, reaching a record high in 2023, and the space debris environment has shown an exponential deterioration trend. According to the statistics of the European Space Agency, there are currently about 35,000 trackable space objects, of which about 9,100 are active payloads, and the other 26,000 are debris larger than 10 cm. However, the actual number of space debris larger than 1 cm exceeds 1 million, and these objects are sufficient to cause catastrophic damage. Moreover, the generation of secondary debris triggered by collisions and disintegrations is triggering the "Kessler syndrome" chain effect, which has posed a serious threat to spacecraft in orbit: in 2022 alone, the International Space Station carried out three collision avoidance maneuvers, and the number of abnormal events of low-Earth orbit satellites caused by micrometeorite impacts increased by 47% year-on-year.

[0003] The prediction of the orbital evolution of debris clouds is the core issue of space environment management, and its accuracy directly affects the timeliness of collision warning and the effectiveness of removal strategies. However, there are technical bottlenecks in traditional prediction methods:

[0004] Lack of heterogeneous modeling: Existing models mostly assume that debris has homogeneous physical properties (such as the same area-mass ratio). However, actual debris shows significant heterogeneity due to differences in materials and shapes.

[0005] Insufficient atmospheric drag modeling: Classical atmospheric density models (such as Harris-Priester) use static parameterization methods and do not integrate the solar activity index (F10.7) and geomagnetic disturbance (Ap index) in real time, making it difficult to capture the dynamic fluctuations of the altitude of the atmosphere top with space weather.

[0006] Lack of collision feedback mechanism: Debris collisions will generate a secondary debris swarm, and traditional Monte Carlo simulations do not establish a dynamic coupling mechanism between the generated debris and the parent swarm, resulting in the accumulation of long-term prediction errors.

[0007] In summary, there is an urgent need for a method and system for predicting the orbital evolution of space debris clouds that involve the non-linear effects of atmospheric drag and the heterogeneity of debris swarms. Summary of the Invention

[0008] To solve the problems existing in the above-mentioned prior art, the purpose of the present invention is to provide a method and system for predicting the orbital evolution of space debris clouds based on atmospheric drag and population dynamics. By establishing an atmospheric drag dynamic model dominated by the perigee altitude, a debris surface-to-mass ratio adaptive classification mechanism, and a population dynamics model with collision feedback, the prediction deviation problem caused by traditional methods ignoring debris diversity and atmospheric nonlinear effects is solved.

[0009] To achieve the above object, the present invention provides the following solutions:

[0010] A method for predicting the orbital evolution of a space debris cloud based on atmospheric drag and population dynamics, comprising:

[0011] Obtain the initial orbital parameters of the target debris cloud, divide the debris population into subclasses according to the initial orbital parameters, and simulate and generate the random initial state of the debris population;

[0012] While simulating and generating the random initial state of the debris population, calculate the orbital evolution trajectories of each subclass of debris to generate a three-dimensional orbital evolution result.

[0013] Optionally, the initial orbital parameters include: perigee altitude, apogee altitude, orbital inclination, and debris physical properties;

[0014] Among them, the debris physical properties include surface-to-mass ratio and material density.

[0015] Optionally, dividing the debris population into subclasses includes:

[0016] Based on the perigee altitude and the real-time atmospheric density model, calculate the deceleration effect of drag on the debris, that is, the attenuation rate;

[0017] Obtain the input features of the debris, input the input features into a machine learning model, and use the association rule between the surface-to-mass ratio and the attenuation rate;

[0018] As the classification condition, obtain the subclass division result and assign an independent attenuation coefficient to each subclass division result;

[0019] The machine learning model is obtained by training with a training set, and the training set includes: debris material, shape parameters, and initial orbital altitude;

[0020] During the subclass division process, dynamically adjust the classification threshold according to the deviation between the real-time attenuation rate and the predicted value.

[0021] Optionally, calculating the deceleration effect of drag on the debris includes:

[0022]

[0023] Among them, C dis the drag coefficient, ρ(h p ) is the atmospheric density at the perigee altitude, is the fragment area mass ratio, and v is the fragment velocity.

[0024] Optionally, the machine learning model includes: a random forest model or a K-means clustering algorithm model;

[0025] The random forest model uses Gini impurity or information gain as the classification criterion:

[0026]

[0027] Among them, Gini(p) is the Gini impurity, and p i is the proportion of the i-th type of sample in the node;

[0028]

[0029]

[0030] Among them, Entropy(S) is the information entropy, S is the current data set, p i is the proportion of the i-th type of sample in the data set, k is the total number of categories, IG is the information gain, S parent is the data set of the parent node, S j is the data set of the j-th child node after splitting by a certain feature, m is the number of split child nodes, is the proportion of the number of samples in the child node to the total number of samples;

[0031] The K-means clustering algorithm model divides the input features into multiple clusters, randomly selects multiple initial cluster centers, calculates the Euclidean distance from each sample in the cluster to the cluster center, and assigns the samples in the cluster to the nearest cluster center;

[0032] Calculating the Euclidean distance from each sample in the cluster to the cluster center includes:

[0033]

[0034] Among them, d(X,c i ) is the Euclidean distance from the sample point X to the cluster center c i , n is the feature dimension of the sample, x j is the feature value of the sample X in the j-th dimension, c ij is the cluster center c i at the coordinate in the j-th dimension.

[0035] Optionally, simulating and generating the random initial state of the debris population includes:

[0036] Calculating the collision probability according to the debris density and relative velocity:

[0037] P collision = nσv rel Δt

[0038] where n is the debris density in the target area; σ is the collision cross-section; v rel is the relative velocity; Δt is the time window;

[0039] If a collision occurs, set the kinetic energy to meet the conditions based on the law of conservation of momentum, and generate secondary debris. Further, based on the collision energy and material properties, generate the mass distribution of the secondary debris. Assume that the mass distribution of the secondary debris satisfies a power-law relationship:

[0040] N(m) ∝ m -α

[0041] where N(m) is the number of debris with mass m, and α is the empirical exponent;

[0042] According to the power-law distribution, use the method of random sampling to generate the mass of the secondary debris:

[0043]

[0044] where R is a uniformly distributed random number in the range from 0 to 1, and m min is the minimum debris mass;

[0045] Allocate the velocity vectors of the secondary debris through the conservation of total momentum:

[0046]

[0047] where N is the total number of generated secondary debris, m i is the mass of the i-th secondary debris after the collision, and v' i is the velocity vector of the i-th secondary debris after the collision;

[0048] Calculate the allocated energy of each secondary debris by the proportional distribution method based on the mass, calculate the velocity of the secondary debris, and randomly assign the direction of the debris velocity:

[0049]

[0050] where is the unit vector in the random direction;

[0051] Adjust the velocity direction, select some debris, and slightly adjust the velocity of the debris to maintain the conservation of total momentum.

[0052] Optionally, setting the kinetic energy to meet the conditions based on the law of conservation of momentum includes:

[0053]

[0054] Among them, m A , m B is the mass of debris A and B before collision, v A , v B is the velocity vector of debris A and B before collision.

[0055] Optionally, calculating the orbital evolution trajectories of each subclass of debris includes:

[0056] v′ i = v cm + Δv i

[0057]

[0058] Among them, v' i is the debris velocity, v cm is the center-of-mass velocity, and Δv i is the increment of the debris velocity relative to the center-of-mass velocity of the collision point.

[0059] Optionally, the three-dimensional orbital evolution results include: the overall decay curve, the collision risk heat map, and the reentry time probability distribution.

[0060] To achieve the above object, the present invention provides a space debris cloud orbital evolution prediction system based on atmospheric drag and population dynamics, including:

[0061] A data acquisition module for obtaining the initial orbital parameters of the target debris cloud;

[0062] A dynamic classification module for subclassifying the debris population according to the initial orbital parameters;

[0063] A data simulation and output module for simulating and generating the random initial states of the debris population, simultaneously calculating the orbital evolution trajectories of each subclass of debris, and generating three-dimensional orbital evolution results.

[0064] The beneficial effects of the present invention are:

[0065] Traditional methods for predicting the orbital evolution of debris clouds assume that all debris has the same physical properties, ignoring the heterogeneity effects caused by material and shape differences, resulting in model distortion; the static atmospheric model has insufficient accuracy and does not integrate solar activity indices and geomagnetic disturbances in real time; the Monte Carlo simulation does not establish a feedback mechanism for the generation of secondary debris from collisions. These defects lead to serious limitations of traditional methods in scenarios such as intense solar activity periods, heterogeneous debris clouds, and collision chain effects in long-term predictions. The method proposed in the present invention uses a dynamic classification algorithm for surface mass ratio based on machine learning in the classification of heterogeneous debris, and adaptively classifies the physical properties and orbital parameters of debris through random forest or K-means clustering, and adjusts the classification threshold in real time. It solves the model deviation caused by traditional homogeneous assumptions and improves the accuracy of population evolution modeling. Integrate the NRLMSISE-00 / JB2008 real-time atmospheric model, dynamically obtain the solar activity index (F10.7) and geomagnetic index (Ap) through the API, and construct an atmospheric density calculation model dominated by the perigee altitude. Capture the dynamic fluctuations of the atmosphere top with space weather, accurately quantify the drag acceleration, and overcome the error accumulation problem of static models for predicting long-term decay trajectories. Embed a collision generation mechanism in the Monte Carlo simulation, generate secondary debris based on momentum conservation and energy constraints, and dynamically incorporate subsequent evolution calculations to achieve a closed-loop feedback of debris collision-secondary debris generation-orbital evolution, avoiding long-term prediction distortion caused by traditional methods ignoring the Kessler syndrome chain effect.

[0066] In summary, the present invention realizes the leap from static homogeneous prediction to dynamic heterogeneous simulation of debris cloud evolution, providing a high-confidence chronological planning basis for the active removal of space debris. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0068] Figure 1 It is a flowchart of a method for predicting the orbital evolution of a space debris cloud based on atmospheric drag and population dynamics according to an embodiment of the present invention;

[0069] Figure 2 It is a schematic diagram of a system for predicting the orbital evolution of a space debris cloud based on atmospheric drag and population dynamics according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0070] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0071] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0072] As Figure 1 shown, this embodiment discloses a method for predicting the orbital evolution of a space debris cloud based on atmospheric drag and swarm dynamics, including: obtaining the initial orbital parameters of the target debris cloud, subclassifying the debris population according to the initial orbital parameters, and simulating and generating the random initial state of the debris population; while simulating and generating the random initial state of the debris population, calculating the orbital evolution trajectories of each subclass of debris to generate a three-dimensional orbital evolution result.

[0073] Specifically, this embodiment discloses a method for predicting the orbital evolution of a space debris cloud based on atmospheric drag and swarm dynamics, including:

[0074] Step 1, obtaining the initial orbital parameters of the target debris cloud, including the perigee altitude, apogee altitude, orbital inclination, and debris physical properties, including the area-to-mass ratio and material density;

[0075] Step 2, calculating the deceleration effect of atmospheric drag on the debris based on the perigee altitude and the real-time atmospheric density model, where the atmospheric density model is dynamically updated through the solar activity index;

[0076] Step 3, according to the heterogeneity of the debris area-to-mass ratio, using a dynamic classification algorithm to divide the debris population into multiple subclasses, and each subclass is assigned an independent attenuation coefficient;

[0077] Step 4, generating the random initial state of the debris population through Monte Carlo simulation and calculating the orbital evolution trajectories of each subclass in parallel;

[0078] Step 5, outputting the overall attenuation curve of the debris cloud, the collision risk heat map, and the probability distribution of the reentry time.

[0079] Furthermore, the initial orbital parameters include: perigee altitude, apogee altitude, orbital inclination, and debris physical properties; wherein, the debris physical properties include the area-to-mass ratio and material density.

[0080] Further, subclassifying the debris population includes: calculating the deceleration effect of drag on the debris, i.e., the decay rate, based on the perigee altitude and the real-time atmospheric density model; the real-time atmospheric density model uses the NRLMSISE-00 or JB2008 model and accesses the solar activity index (F10.7) and the geomagnetic index (Ap) in real time through an API interface; obtaining the input features of the debris, inputting the input features into a machine learning model, and using the correlation rule between the area-to-mass ratio and the decay rate as the classification condition to obtain the subclassifying result and assign an independent decay coefficient to each subclassifying result; the machine learning model is obtained by training with a training set, and the training set includes: debris material, shape parameters, and initial orbital altitude; during the subclassifying process, dynamically adjust the classification threshold according to the deviation between the real-time original decay rate and the predicted value.

[0081] The decay coefficient simplifies complex parameters such as the area-to-mass ratio, drag coefficient, and atmospheric density into a single dynamic index. Debris in different subclasses (high / medium / low decay groups) are assigned independent decay coefficients, which are directly input into the orbital evolution model to calculate future trajectories.

[0082] Further, calculating the deceleration effect of drag on the debris includes:

[0083]

[0084] where C d is the drag coefficient, ρ(h p ) is the atmospheric density at the perigee altitude, is the area-to-mass ratio of the debris, and v is the debris velocity.

[0085] Further, the machine learning model includes: a random forest model or a K-means clustering algorithm model; the random forest model uses Gini impurity or information gain as the classification criterion, and the K-means clustering algorithm model divides the input features into multiple clusters, randomly selects multiple initial cluster centers, calculates the Euclidean distance from each sample in the cluster to the cluster center, and assigns the samples in the cluster to the nearest cluster center.

[0086] Specifically, train the machine learning model based on historical reentry data to establish the correlation rule between the debris area-to-mass ratio and the decay rate; during the simulation, dynamically adjust the classification threshold according to the deviation between the real-time decay rate and the predicted value to optimize the subclassifying criterion.

[0087] Establishing the correlation rule between the debris area-to-mass ratio and the decay rate includes:

[0088] The decay rate of the semi-major axis of the debris orbit over time is caused by atmospheric drag and is given by:

[0089]

[0090] CD is the drag coefficient (usually taking values from 2.2 to 2.5); A / m is the area-to-mass ratio; ρ is the atmospheric density (which can be obtained from an atmospheric model, such as JB2008); v is the orbital velocity; μ is the Earth's gravitational constant (398600 km 3 / s 2 ): a is the semi-major axis; the semi-major axis a can be calculated from TLE data, and by fitting multiple TLE data, we can obtain and thus the area-to-mass ratio can be calculated inversely:

[0091]

[0092] The debris can be classified according to the numerical distribution, setting critical thresholds (such as high attenuation, medium attenuation, low attenuation), corresponding to different intervals. For example, for high, medium, and low area-to-mass ratio groups, it is stipulated that is the high area-to-mass ratio group, with fast attenuation corresponding to the high attenuation group, is the medium area-to-mass ratio group, with moderate attenuation corresponding to the medium attenuation group, is the low area-to-mass ratio group, with slow attenuation corresponding to the low attenuation group.

[0093] The machine learning model is a random forest or K-means clustering algorithm, and the input features include debris material, shape parameters, and initial orbital height.

[0094] Random forest is an ensemble learning algorithm based on decision trees, belonging to the category of supervised learning, suitable for classification and regression tasks. Its core idea is to build multiple decision trees and integrate the results to improve the generalization ability and robustness of the model. The construction of decision trees uses Gini Impurity or Information Gain as the splitting criterion.

[0095] Gini Impurity formula:

[0096]

[0097] In Equation (2), p i is the proportion of the i-th class of samples in the node.

[0098] Information Gain is essentially a measure of the contribution of a feature to the classification result by the reduction in entropy. Entropy calculation:

[0099]

[0100] Among them, Entropy(S) is the information entropy, S is the current dataset, p i is the proportion of the i-th class of samples in the dataset, k is the total number of classes, IG is the information gain, and Sparent is the parent node dataset, S j is the j-th child node dataset after splitting a certain feature, m is the number of child nodes after splitting, is the proportion of the number of samples in the child node to the total number of samples. Machine learning is used to assist classification and predict the probability that the debris belongs to the high / medium / low attenuation group.

[0101] The Gini impurity is first calculated according to the measured is divided into three categories: high / medium / low attenuation groups, and the Gini value at this time is calculated. Select other splitting conditions (such as perigee altitude, atmospheric density, semi-major axis decay rate and other attributes) to calculate the corresponding Gini values, compare the results, select the splitting condition with the smaller Gini value, and repeat this process to select the optimal splitting condition. The same is true for information gain, and select the splitting method with the largest information gain.

[0102] Information gain is used to screen out the features that are most effective for classification:

[0103] The selectable features include h p and material type, etc. First, according to the is divided into high / medium / low attenuation groups.

[0104] Information gain represents the reduction in data uncertainty after splitting by a certain feature:

[0105]

[0106] In formula (4), A is the feature to be split; Values(A) is the set of values of feature A; S v is the sub-dataset where the value of feature A is v; |S v | is the number of samples in the sub-dataset S v ; |S| is the total number of samples in the parent node.

[0107] The K-means clustering algorithm is an unsupervised learning algorithm used to divide data into K clusters. Its goal is to minimize the sum of the squared distances (SSE) from the samples within the cluster to the cluster center. Randomly select K initial cluster centers (e.g., K = 3, corresponding to high / medium / low categories), calculate the Euclidean distance from each sample to each cluster center, and assign it to the nearest cluster. The distance formula is:

[0108]

[0109] First, assume that the decay rate of the semi-major axis over time is used for clustering. Similarly, other decay rates can be calculated: the decay rate of the perigee altitude Orbital inclination change rate As a clustering feature vector, each debris is classified into different decay groups according to the Euclidean distance from the debris to the cluster center.

[0110] Dynamically adjusting the classification threshold according to the deviation between the real-time original decay rate and the predicted value includes:[[]]

[0111] If the prediction deviation of a certain subclass (such as the medium decay group) continues to be high, then relax its threshold range (such as from Adjusted to

[0112] Furthermore, simulating the generation of the random initial state of the debris population includes: calculating the collision probability according to the debris density and relative velocity. If a collision occurs, set the kinetic energy to meet the conditions based on the law of conservation of momentum, generate secondary debris, and further generate the mass distribution of the secondary debris based on the collision energy and material properties, and allocate the velocity vectors of the secondary debris through the total momentum conservation.

[0113] Specifically, calculate the collision probability according to the debris density and relative velocity. If a collision occurs, generate secondary debris based on the law of conservation of momentum and incorporate its physical property parameters into the orbital evolution of subsequent subclasses. Collision probability calculation:

[0114] P collison =n·σ·v rel ·Δt (6)

[0115] In formula (6), n is the debris density of the target area; σ is the collision cross-section; v rel is the relative velocity; Δt is the time window.

[0116] For two colliding objects (debris A and debris B), the total momentum before and after the collision is conserved:

[0117]

[0118] In formula (7), m A , m B are the masses of debris A and B before the collision; v A , v B are the velocity vectors of debris A and B before the collision; m i is the mass of the i-th secondary debris after the collision; v' i is the velocity vector of the i-th secondary debris after the collision; N is the total number of generated secondary debris.

[0119] Calculation steps:

[0120] Calculation of the total momentum before the collision:

[0121] P total =m A v A+m B v B (8)

[0122] Generate the fragment mass distribution (power-law distribution) according to the collision energy and material properties:

[0123] N(m) ∝ m -α (9)

[0124] where N(m) is the number of fragments with mass m, and α is the empirical exponent, ranging from 1.6 to 2.0;

[0125] Distribute the fragment velocity vectors according to the total momentum conservation:

[0126]

[0127] Calculate the allocated energy of each secondary fragment using the proportional distribution method based on the mass, and then calculate the velocity:

[0128]

[0129] where β is the velocity distribution exponent, usually taking values from 0.5 to 1.0, E total is the total energy before the collision, m total is the total mass before the collision, E i is the energy of the i-th secondary fragment after the collision, m i is the mass of the i-th secondary fragment after the collision;

[0130] On the premise of satisfying the momentum conservation, randomly distribute the directions of the fragment velocities:

[0131]

[0132] where, is the unit vector in the random direction;

[0133] Since the velocity directions are randomly generated, resulting in non-conservation of the total momentum, it is necessary to adjust the velocity directions:

[0134]

[0135] Further select some fragments and finely adjust the velocities of the fragments to make the total momentum conserved.

[0136] The increment Δv of the fragment velocity relative to the centroid velocity of the collision point i obeys the statistical distribution, and its centroid velocity calculation formula:

[0137]

[0138] Fragment velocity calculation formula:

[0139] v'i =v cm +Δv i (12)

[0140] Because of the law of conservation of energy, the total kinetic energy after the collision cannot be greater than the total kinetic energy before the collision. The following conditions must be met:

[0141]

[0142] like Figure 2 As shown, this embodiment also provides a space debris cloud orbit evolution prediction system based on atmospheric drag and group dynamics, including: a data acquisition module, used to obtain the initial orbit parameters of the target debris cloud; a dynamic classification module, used to divide the debris group into subcategories according to the initial orbit parameters; a data simulation and output module, used to simulate and generate a random initial state of the debris group, and simultaneously calculate the orbit evolution trajectory of each subcategory of debris to generate a three-dimensional orbit evolution result.

[0143] Specifically, this embodiment also provides a space debris cloud orbit evolution prediction system based on atmospheric drag and group dynamics, including: a data acquisition module: obtaining real-time orbital elements (TLE) and physical property data from a space detection database; a dynamic classification module: performing adaptive classification and parameter allocation of debris area-to-mass ratio; a data simulation engine: integrating a GPU-accelerated Monte Carlo simulator and a high-precision orbit integrator; a visualization output module: generating a three-dimensional orbit evolution animation and a collision risk heat map.

[0144] The numerical simulation engine supports multi-node distributed computing, and the task scheduling strategy is: high area-to-mass ratio fragments are assigned to high-priority threads to shorten their orbital integration time step; low area-to-mass ratio fragments use a low-precision fast integration algorithm.

[0145] The collision risk heat map of the visualization output module can be mapped to the risk level with a color gradient by dividing the orbital space into three-dimensional grid cells, counting the density and relative velocity of debris in each cell, or calculating the cell collision probability based on the collision cross-section formula.

[0146] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for predicting the orbital evolution of space debris clouds based on atmospheric drag and swarm dynamics, characterized in that, Including: Obtain the initial orbital parameters of the target debris cloud, subclassify the debris population according to the initial orbital parameters, and simulate and generate the random initial state of the debris population; While simulating and generating the random initial state of the debris population, calculate the orbital evolution trajectories of each subclass of debris and generate a three-dimensional orbital evolution result.

2. The method for predicting the orbital evolution of space debris clouds based on atmospheric drag and swarm dynamics according to claim 1, characterized in that The initial orbital parameters include: perigee altitude, apogee altitude, orbital inclination, and debris physical properties; Among them, the debris physical properties include surface-to-mass ratio and material density.

3. The method for predicting the orbital evolution of space debris clouds based on atmospheric drag and swarm dynamics according to claim 1, characterized in that Subclassifying the debris population includes: Based on the perigee altitude and the real-time atmospheric density model, calculate the deceleration effect of drag on the debris, i.e., the attenuation rate; Obtain the input features of the debris, input the input features into a machine learning model, and use the correlation rule between the surface-to-mass ratio and the attenuation rate; As the classification condition, obtain the subclass division result and assign an independent attenuation coefficient to each subclass division result; The machine learning model is obtained by training with a training set, and the training set includes: debris material, shape parameters, and initial orbital altitude; During the subclass division process, dynamically adjust the classification threshold according to the deviation between the real-time attenuation rate and the predicted value.

4. The method for predicting the orbital evolution of space debris clouds based on atmospheric drag and swarm dynamics according to claim 3, wherein Calculating the deceleration effect of drag on the debris includes: Among them, C d is the drag coefficient, ρ(h p ) is the atmospheric density at the perigee altitude, is the fragment area-mass ratio, and v is the fragment velocity.

5. The method for predicting the orbital evolution of space debris clouds based on atmospheric drag and swarm dynamics according to claim 3, wherein The machine learning model includes: a random forest model or a K-means clustering algorithm model; The random forest model uses Gini impurity or information gain as the classification criterion: Among them, Gini(p) is the Gini impurity, and p i is the proportion of the i-th type of samples in the node; Among them, Entropy(S) is the information entropy, S is the current data set, and p i is the proportion of the i-th class of samples in the data set, k is the total number of classes, IG is the information gain, and S parent is the data set of the parent node, and S j is the data set of the j-th child node after splitting by a certain feature, m is the number of child nodes after splitting, is the proportion of the number of child node samples to the total number of samples; The K-means clustering algorithm model divides the input features into multiple clusters, randomly selects multiple initial cluster centers, calculates the Euclidean distance from each sample in the cluster to the cluster center, and assigns the samples in the cluster to the nearest cluster center; Calculating the Euclidean distance from each sample in the cluster to the cluster center includes: Among them, d(X, c i ) is the Euclidean distance from the sample point X to the cluster center c i , n is the feature dimension of the sample, x j is the feature value of the sample X in the j-th dimension, c ij is the coordinate of the cluster center c i in the j-th dimension.

6. The method for predicting the orbital evolution of space debris clouds based on atmospheric drag and swarm dynamics according to claim 1, wherein Simulating and generating the random initial state of the debris population includes: Calculate the collision probability according to the debris density and relative velocity: P collision = nσv rel Δt where n is the debris density of the target area; σ is the collision cross-section; v rel is the relative velocity; Δt is the time window; If a collision occurs, set the kinetic energy to meet the conditions based on the law of conservation of momentum, generate secondary debris, and further generate the mass distribution of the secondary debris based on the collision energy and material characteristics. Assume that the mass distribution of the secondary debris satisfies a power-law relationship: N(m) ∝ m -α Among them, N(m) is the number of debris with mass m, and α is the empirical exponent; According to the power-law distribution, use the method of random sampling to generate the mass of the secondary debris: where R is a uniformly distributed random number with a value range between 0 and 1, and m min is the minimum fragment mass; Allocate the velocity vector of the secondary debris through the conservation of total momentum: where N is the total number of generated secondary fragments, and m i is the mass of the i-th secondary fragment after the collision, and v' i is the velocity vector of the i-th secondary fragment after the collision; Calculate the allocated energy of each secondary debris based on the mass using the proportional distribution method, calculate the velocity of the secondary debris, and randomly allocate the direction of the debris velocity: Among them, is a unit vector in a random direction; Adjust the velocity direction, select some debris, and fine-tune the velocity of the debris to maintain the conservation of total momentum.

7. The method for predicting the orbital evolution of space debris clouds based on atmospheric drag and swarm dynamics according to claim 6, characterized in that Setting the kinetic energy to meet the conditions based on the law of conservation of momentum includes: where m A , m B are the masses of debris A and B before the collision, and v A , v B are the velocity vectors of debris A and B before the collision.

8. The method for predicting the orbital evolution of space debris clouds based on atmospheric drag and swarm dynamics according to claim 1, characterized in that Calculating the orbital evolution trajectories of each subclass of debris includes: v i = v cm + Δv i where, v' i is the debris velocity, v cm is the centroid velocity, and Δv i is the increment of the debris velocity relative to the centroid velocity of the collision point.

9. The method for predicting the orbital evolution of space debris clouds based on atmospheric drag and swarm dynamics according to claim 1, characterized in that The three-dimensional orbital evolution result includes: an overall attenuation curve, a collision risk heat map, and a reentry time probability distribution.

10. A space debris cloud orbit evolution prediction system based on atmospheric drag and swarm dynamics, characterized in that, Including: A data acquisition module for obtaining the initial orbital parameters of the target debris cloud; A dynamic classification module for subclassifying the debris population according to the initial orbital parameters; The data simulation and output module is used to simulate and generate the random initial state of the debris population, and at the same time calculate the orbital evolution trajectories of each subclass of debris to generate three-dimensional orbital evolution results.

Images