On-orbit spacecraft temperature prediction method and device based on physical information neural network

Abstract

The invention discloses an on-orbit spacecraft temperature prediction method and device based on a physical information neural network. The method comprises the following steps: S1, constructing an object model and an orbit heat source model according to a spacecraft to be predicted; s2, solving a space external heat flow of the orbit heat source model, and converting the space external heat flow into a volume heat flow of the object model; s3, constructing a physical information neural network; wherein a three-dimensional transient heat conduction equation is embedded in the physical information neural network; s4, designing a total loss function according to the volume heat flow and the physical information neural network; and S5, training the physical information neural network according to the total loss function to obtain a temperature prediction result of the spacecraft to be predicted. According to the method, the situation that model grids need to be divided in traditional finite element simulation is avoided, the calculation efficiency is improved, the calculation time is shortened, the temperature result can be predicted in real time, and the generalization ability and accuracy of the model are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of spacecraft structure thermal performance simulation, and in particular relates to a method and device for predicting the temperature of an on-orbit spacecraft based on a physical information neural network. Background Art

[0002] Temperature control is a crucial aspect of spacecraft design. Spacecraft in orbit are subject to significant fluctuations in surface temperature due to direct solar radiation, infrared radiation from the Earth, reflected radiation from the Earth, and infrared and reflected radiation from other planets. Furthermore, they are also affected by passages into and out of the Earth's shadow. These significant temperature fluctuations can cause thermal shock, impacting the spacecraft's health and service life. Therefore, temperature prediction is essential during design to ensure optimal operating temperature. Currently, commonly used methods for predicting the temperature of spacecraft in orbit include numerical simulation methods such as the finite element method, finite difference method, and finite volume method. However, these methods require meshing, impose high computational requirements, and take long to solve, making them inadequate for real-time prediction. Therefore, research is urgently needed to develop methods for rapidly and accurately predicting the surface temperature field of spacecraft in orbit. Summary of the Invention

[0003] In order to solve the above problems existing in the prior art, the present invention provides a method and device for predicting the temperature of an on-orbit spacecraft based on a physical information neural network.

[0004] The technical problem to be solved by the present invention is achieved through the following technical solutions:

[0005] In a first aspect, the present invention provides a method for predicting the temperature of an on-orbit spacecraft based on a physical information neural network, the method comprising:

[0006] S1. Build an object model and orbital heat source model based on the spacecraft to be predicted;

[0007] S2. solving the external heat flow of the track heat source model and converting the external heat flow into the volume heat flow of the object model;

[0008] S3. Constructing a physical information neural network; wherein the physical information neural network is embedded with a three-dimensional transient heat conduction equation;

[0009] S4. Designing a total loss function based on the volume heat flow and the physical information neural network;

[0010] S5. Train the physical information neural network according to the total loss function to obtain a temperature prediction result of the spacecraft to be predicted.

[0011] Optionally, after S5, the method further includes:

[0012] S6. Calculating the temperature difference between the temperature prediction result and the simulated comparison temperature;

[0013] S7. When the temperature difference is greater than a preset difference threshold, adjust the weights of each loss function in the total loss function and return to S4;

[0014] S8. When the temperature difference is less than or equal to the preset difference threshold, output the temperature prediction result as a final prediction result.

[0015] Optionally, the total loss function includes: initial condition loss, PDE residual loss and radiation boundary loss.

[0016] Optionally, the total loss function is expressed as follows:

[0017] L total =L pde +β·L initial +γ·L rad ;

[0018] Among them, L total Represents the total loss function, L pde represents the PDE residual loss, β represents the initial condition loss L initial The weight of γ represents the radiation boundary loss L rad The weight of .

[0019] Optionally, the PDE residual loss is expressed as follows:

[0020]

[0021] in, represents the number of sampling points obtained by sampling the nodes in the object model, T i represents the temperature field of the i-th sampling point, t represents the time, (xi,yi,zi) represents the coordinates of the i-th sampling point in the object model, α represents the thermal diffusion coefficient of the material of the spacecraft to be predicted, Q i represents the volume heat flux term at the i-th sampling point.

[0022] Optionally, the initial condition loss is expressed as follows:

[0023]

[0024] Among them, N0 represents the number of sampling points at the initial moment, T j Represents the temperature field of the jth sampling point at the initial moment.

[0025] Optionally, the radiation boundary loss is expressed as follows:

[0026]

[0027] in, is the number of boundary sampling points, T k Represents the temperature field of the kth boundary sampling point, n is the normal gradient, ∈σT k 4 is the Stefan-Boltzmann radiation term, ∈ is the surface emissivity of the material, and σ is the Stefan-Boltzmann constant.

[0028] In a second aspect, the present invention provides an on-orbit spacecraft temperature prediction device based on a physical information neural network, the device comprising:

[0029] A model building module, used for building an object model and an orbital heat source model according to the spacecraft to be predicted;

[0030] a heat flow acquisition module, configured to solve the outer-space heat flow of the track heat source model and convert the outer-space heat flow into the volume heat flow of the object model;

[0031] A neural network building module, used to build a physical information neural network; wherein the physical information neural network is embedded with a three-dimensional transient heat conduction equation;

[0032] a loss function design module, configured to design a total loss function based on the volume heat flow and the physical information neural network;

[0033] The temperature prediction module is used to train the physical information neural network according to the total loss function to obtain the temperature prediction result of the spacecraft to be predicted.

[0034] The technical solutions provided by the embodiments of the present invention may have the following beneficial effects:

[0035] In this technical solution, an object model and an orbital heat source model are constructed. The orbital heat source model is used to obtain the object model's volumetric heat flow. A loss function is then designed to train a physical information neural network embedded with a three-dimensional transient heat conduction equation to obtain temperature predictions for the spacecraft being predicted. This method eliminates the need for traditional finite element simulations to divide the model into grids, improving computational efficiency and reducing computation time. It also enables real-time temperature predictions, enhancing the model's generalization and accuracy.

[0036] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of a method for predicting temperature of an on-orbit spacecraft based on a physical information neural network according to an embodiment of the present invention;

[0038] Figure 2 is a schematic diagram of an object model provided by an embodiment of the present invention;

[0039] Figure 3 is a schematic diagram of an antenna operating track provided by an embodiment of the present invention;

[0040] Figure 4 is a schematic diagram of a heat flow calculation result outside a space provided by an embodiment of the present invention;

[0041] Figure 5 This is a schematic diagram of the structure of a physical information neural network provided by an embodiment of the present invention;

[0042] Figure 6 is a schematic diagram of a temperature prediction result provided by an embodiment of the present invention;

[0043] Figure 7 is a schematic diagram of a simulated temperature comparison provided by an embodiment of the present invention;

[0044] Figure 8 An embodiment of the present invention provides an on-orbit spacecraft temperature prediction device based on a physical information neural network. DETAILED DESCRIPTION

[0045] The present invention will be further described in detail below with reference to specific examples, but the embodiments of the present invention are not limited thereto.

[0046] Figure 1 This is a flow chart of a method for predicting temperature of an on-orbit spacecraft based on a physical information neural network provided by an embodiment of the present invention. Figure 1 As shown, the method may include the following steps:

[0047] S1. Build an object model and orbital heat source model based on the spacecraft to be predicted.

[0048] For example, Figure 2 is a schematic diagram of an object model provided by an embodiment of the present invention. The spacecraft to be predicted can be a peripheral truss deployable antenna. The coordinates and connection relationship of the rod nodes and the connection between the parabola and the rod are generated by a function. The antenna diameter can be 20m, the height can be 5m, and the rod diameter can be 0.02m. There are 18 upper chords, lower chords, vertical webs, and diagonal webs each. The object model of the peripheral truss deployable antenna is shown in FIG. Figure 2 .

[0049] Taking the perimeter truss deployable antenna as an example, Figure 3This is a schematic diagram of an antenna operating track provided by an embodiment of the present invention. Taking into account the complex environment of the surrounding truss deployable antenna on orbit, the antenna operating track is determined by the six orbit elements, namely, the orbit semi-major axis, orbit eccentricity, orbit inclination, ascending node right ascension, perigee argument and true anomaly.

[0050] S2. Calculate the external heat flux of the orbital heat source model and convert the external heat flux into the volume heat flux of the object model.

[0051] It is understandable that Figure 4 This is a schematic diagram of the calculation results of the external heat flux provided by an embodiment of the present invention. Taking into account the direct solar radiation, the earth's infrared radiation, the earth's reflected radiation, etc., the external heat flux q to which the antenna is subjected during its orbital operation period is output. Figure 4 .

[0052] S3. Construct a physical information neural network; wherein the physical information neural network is embedded with a three-dimensional transient heat conduction equation.

[0053] It is understandable that Figure 5 This is a schematic diagram of the structure of a physical information neural network provided by an embodiment of the present invention. The present invention establishes a 4-layer fully connected network. Figure 5 The input layer variables are spatiotemporal data, node coordinates, and time. The hidden layer consists of linear transformations and Tanh activation functions. The output layer result is the temperature field. The governing equation is the PDE residual loss, the boundary conditions are the boundary losses, the initial conditions are the losses that constitute the initial conditions, and the initial temperature is -260°C. When constructing the forward propagation physical information neural network, the three-dimensional transient heat conduction equation is embedded:

[0054]

[0055] Wherein, T is the temperature field, (x, y, z) is the node coordinate, t is the time, α is the thermal diffusion coefficient of the material, and the truss of the peripheral truss deployable antenna in this embodiment uses carbon fiber rods of 6.2×10 -5 m 2 / s, the reflective surface uses gold-plated molybdenum wire: 7.12×10 -5 m 2 / s, Q is the volume heat source, which is converted from the heat flow outside the space.

[0056]

[0057] A=π·r 2 ;

[0058] Where q is the external heat flux in space, A is the cross-sectional area of ​​the rod, is the volume heat capacity of the material, and r is the radius of the rod.

[0059] S4. Design the total loss function based on the volume heat flow and physical information neural network.

[0060] S5. Train the physical information neural network according to the total loss function to obtain the temperature prediction result of the spacecraft to be predicted.

[0061] It is understandable that Figure 6 This is a schematic diagram of a temperature prediction result provided by an embodiment of the present invention. Taking the peripheral truss deployable antenna model as an example, data is first generated by sampling from the object model. The specific sampling method is to sample along the rods, sampling 5 points on each rod and all nodes of the parabola. The orbital heat source is calculated, and the calculated external heat flow is converted into volume heat flow and loaded to the sampling points. Training is performed based on the loss function, and the temperature results are calculated and displayed, as shown in the figure. Figure 6 .

[0062] Optionally, after S5, the method further includes:

[0063] S6, calculating the temperature difference between the temperature prediction result and the simulated comparison temperature;

[0064] S7. When the temperature difference is greater than the preset difference threshold, adjust the weights of each loss function in the total loss function and return to S4;

[0065] S8. When the temperature difference is less than or equal to the preset difference threshold, the temperature prediction result is output as the final prediction result.

[0066] It is worth mentioning that Figure 7 It is a schematic diagram of a simulated comparison temperature provided by an embodiment of the present invention. The temperature prediction result calculated by the present invention can be compared with the simulated comparison temperature. If the temperature difference is less than the preset difference threshold 1, the temperature prediction result is output. Otherwise, the parameters are back-propagated to dynamically adjust the respective weights of the radiation boundary loss and the initial condition loss in the loss function.

[0067] Optionally, the total loss function includes: initial condition loss, PDE residual loss and radiation boundary loss.

[0068] The total loss function is expressed as follows:

[0069] L total =L pde +β·L initial +γ·L rad ;

[0070] Among them, L total Represents the total loss function, L pde represents the PDE residual loss, β represents the initial condition loss L initial The weight of γ represents the radiation boundary loss L rad The weight of .

[0071]

[0072] The PDE residual loss is expressed as follows:

[0073]

[0074] in, Indicates the number of sampling points obtained by sampling the nodes in the object model, T i represents the temperature field of the i-th sampling point, t represents the time, (xi,yi,zi) represents the coordinates of the i-th sampling point in the object model, α represents the thermal diffusion coefficient of the material of the spacecraft to be predicted, Q i represents the volume heat flux term at the i-th sampling point.

[0075] The initial condition loss is expressed as follows:

[0076]

[0077] Among them, N0 represents the number of sampling points at the initial moment, T j Represents the temperature field of the jth sampling point at the initial moment.

[0078] The radiation boundary loss is expressed as follows:

[0079]

[0080] in, is the number of boundary sampling points, T k Represents the temperature field of the kth boundary sampling point, n is the normal gradient, ∈σT k 4 is the Stefan-Boltzmann radiation term, ∈ is the surface emissivity of the material, and σ is the Stefan-Boltzmann constant.

[0081] The present invention adopts a physical information neural network to predict the surface temperature field of the spacecraft, and embeds the three-dimensional transient heat conduction equation as a soft constraint into the physical information neural network, avoiding the grid generation required by traditional numerical methods. At the same time, it reduces the requirements of traditional deep learning algorithms on the number of samples, improves computational efficiency and reduces calculation time, and can predict temperature results in real time, thereby improving the generalization ability and accuracy of the model.

[0082] Figure 8 An embodiment of the present invention provides an on-orbit spacecraft temperature prediction device based on a physical information neural network. The device 800 may include:

[0083] A model building module 801 is used to build an object model and an orbital heat source model according to the spacecraft to be predicted;

[0084] The heat flow acquisition module 802 is used to solve the external heat flow of the track heat source model and convert the external heat flow into the volume heat flow of the object model;

[0085] A neural network construction module 803 is used to construct a physical information neural network; wherein the physical information neural network is embedded with a three-dimensional transient heat conduction equation;

[0086] a loss function design module 804 for designing a total loss function based on the volume heat flow and the physical information neural network;

[0087] The temperature prediction module 805 is used to train the physical information neural network according to the total loss function to obtain the temperature prediction result of the spacecraft to be predicted.

[0088] In the description of this specification, the reference terms "one embodiment," "some embodiments," "example," "specific example," or "some examples" mean that the specific features or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features or characteristics described can be combined in any suitable manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification.

[0089] Although the present invention is described herein in conjunction with various embodiments, in the process of implementing the claimed invention, those skilled in the art may understand and implement other variations of the disclosed embodiments by reviewing the drawings and the disclosed content. In the description of the present invention, the word "comprising" does not exclude other components or steps, "one" or "a" does not exclude multiple situations, and "multiple" means two or more, unless otherwise clearly and specifically limited. In addition, certain measures are recorded in different embodiments, but this does not mean that these measures cannot be combined to produce good results.

[0090] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0091] It should be noted that the device of the embodiment of the present invention is a device that applies the above-mentioned on-orbit spacecraft temperature prediction method based on physical information neural network. All embodiments of the above-mentioned on-orbit spacecraft temperature prediction method based on physical information neural network are applicable to the device and can achieve the same or similar beneficial effects.

[0092] The above is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. For those skilled in the art of the present invention, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should be considered to fall within the scope of protection of the present invention.

Claims

1. A method for predicting the temperature of an on-orbit spacecraft based on a physical information neural network, characterized in that: The method comprises: S1. Build an object model and orbital heat source model based on the spacecraft to be predicted; S2. solving the external heat flow of the track heat source model and converting the external heat flow into the volume heat flow of the object model; S3. Constructing a physical information neural network; wherein the physical information neural network is embedded with a three-dimensional transient heat conduction equation; S4. Designing a total loss function based on the volume heat flow and the physical information neural network; S5. Train the physical information neural network according to the total loss function to obtain a temperature prediction result of the spacecraft to be predicted.

2. The method for predicting temperature of an on-orbit spacecraft based on a physical information neural network according to claim 1, characterized in that: After S5, the method further includes: S6. Calculating the temperature difference between the temperature prediction result and the simulated comparison temperature; S7. When the temperature difference is greater than a preset difference threshold, adjust the weights of each loss function in the total loss function and return to S4; S8. When the temperature difference is less than or equal to the preset difference threshold, output the temperature prediction result as a final prediction result.

3. The method for predicting temperature of an on-orbit spacecraft based on a physical information neural network according to claim 1, characterized in that: The total loss function includes: initial condition loss, PDE residual loss and radiation boundary loss.

4. The method for predicting temperature of an on-orbit spacecraft based on a physical information neural network according to claim 3, characterized in that: The total loss function is expressed as follows: L total =L pde +β·L initial +γ·L rad ; Among them, L total Represents the total loss function, L pde represents the PDE residual loss, β represents the initial condition loss L initial The weight of γ represents the radiation boundary loss L rad The weight of .

5. According to the on-orbit spacecraft temperature prediction method based on physical information neural network according to claim 4, the PDE residual loss is expressed as follows: in, represents the number of sampling points obtained by sampling the nodes in the object model, T i represents the temperature field of the i-th sampling point, t represents the time, (xi,yi,zi) represents the coordinates of the i-th sampling point in the object model, α represents the thermal diffusion coefficient of the material of the spacecraft to be predicted, Q i represents the volume heat flux term at the i-th sampling point.

6. The on-orbit spacecraft temperature prediction method based on physical information neural network according to claim 5, wherein the initial condition loss is expressed as follows: in, N0 represents the number of sampling points at the initial moment, T j Represents the temperature field of the jth sampling point at the initial moment.

7. The on-orbit spacecraft temperature prediction method based on physical information neural network according to claim 5, wherein the radiation boundary loss is expressed as follows: in, is the number of boundary sampling points, T k Represents the temperature field of the kth boundary sampling point, n is the normal gradient, ∈σT k 4 is the Stefan-Boltzmann radiation term, ∈ is the surface emissivity of the material, and σ is the Stefan-Boltzmann constant.

8. An on-orbit spacecraft temperature prediction device based on physical information neural network, characterized in that: The device comprises: A model building module, used for building an object model and an orbital heat source model according to the spacecraft to be predicted; a heat flow acquisition module, configured to solve the outer-space heat flow of the track heat source model and convert the outer-space heat flow into the volume heat flow of the object model; A neural network building module, used to build a physical information neural network; wherein the physical information neural network is embedded with a three-dimensional transient heat conduction equation; a loss function design module, configured to design a total loss function based on the volume heat flow and the physical information neural network; The temperature prediction module is used to train the physical information neural network according to the total loss function to obtain the temperature prediction result of the spacecraft to be predicted.

Images