Environmental box temperature field simulation method and system based on deep neural network

Abstract

The invention relates to the technical field of data processing, in particular to an environment box temperature field simulation method and system based on a deep neural network. The method comprises the following steps: acquiring a training data set obtained by performing temperature field simulation on different environmental boxes; training a pre-constructed deep neural network model by using the training data set to obtain a fitting relation model of the simulation parameters and the temperature field; simulation parameters of the environment box to be simulated are obtained; simulation parameters of the environment box to be simulated are input into the fitting relation model for prediction, and the temperature field of the environment box to be simulated is obtained. According to the method, the temperature field simulation efficiency can be improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for simulating the temperature field of an environmental chamber based on a deep neural network. Background Technology

[0002] In fields such as industrial manufacturing, scientific research experiments, and electronic product testing, environmental chambers are core equipment for simulating complex temperature and humidity conditions, and the precise control and rapid simulation of their temperature field distribution are of paramount importance.

[0003] Traditional methods for simulating the temperature field of an environmental chamber primarily rely on computational fluid dynamics (CFD) numerical simulation. This involves solving the partial differential equations and energy equations for the motion of viscous fluids to predict the temperature field. CFD simulation requires discretizing the three-dimensional space into a mesh and then solving for the temperature point by point. However, due to the large differences in the size of the various components in the environmental chamber, the number of meshes increases dramatically, leading to computation times that can last for hours or even days, making it difficult to meet the needs of rapid product iteration design.

[0004] Therefore, improving the simulation efficiency of the temperature field in the environmental chamber is a technical problem that urgently needs to be solved. Summary of the Invention

[0005] To address the technical problem of low efficiency in environmental chamber temperature field simulation, the present invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a method for simulating the temperature field of an environmental chamber based on a deep neural network, comprising: acquiring a training dataset obtained by simulating the temperature field of different environmental chambers; the training dataset including simulation parameters of the environmental chamber, three-dimensional coordinates and temperatures of various coordinate points within the environmental chamber, the simulation parameters including geometric feature parameters and boundary condition parameters; training a pre-constructed deep neural network model using the training dataset to obtain a fitting relationship model between the simulation parameters and the temperature field; acquiring simulation parameters of the environmental chamber to be simulated, and generating multiple coordinate points covering the internal space of the environmental chamber to be simulated based on the simulation parameters; inputting the simulation parameters of the environmental chamber to be simulated and the three-dimensional coordinates of the coordinate points into the fitting relationship model for prediction to obtain the temperature field of the environmental chamber to be simulated.

[0007] Beneficial effects: By directly using simulation data obtained from high-fidelity temperature field simulation as the training dataset, the problem of needing to use a large amount of labeled data to train neural network models in existing technologies is avoided; compared with using existing physical models for temperature field simulation, this invention uses a pre-trained fitting relationship model for temperature field simulation, without the need for mesh generation, and can quickly perform temperature field simulation on environmental chambers of different sizes and under different working conditions, thus improving the efficiency of environmental chamber temperature field simulation.

[0008] Furthermore, the simulation parameters include geometric feature parameters and boundary condition parameters; obtaining the training dataset obtained from temperature field simulations of different environmental chambers includes: constructing a parameterized 3D model of the environmental chamber; the parameterized 3D model defines the spatial structure of the environmental chamber through geometric feature parameters; defining boundary condition parameters; the boundary condition parameters are used to define the operating conditions of the environmental chamber, including vent temperature and vent velocity; performing temperature field simulations on the parameterized 3D models of different environmental chambers under different boundary condition parameters to obtain the 3D coordinates and temperatures of coordinate points corresponding to various locations inside different environmental chambers; establishing the correspondence between simulation parameters, 3D coordinates of coordinate points, and temperatures to obtain the training dataset.

[0009] Furthermore, constructing a parametric 3D model of the environmental enclosure includes: simplifying the 3D structure of the environmental enclosure into a cuboid composed of multiple cavities, and constructing an air outlet and an air return outlet on the cuboid; defining the dimensions of the cuboid as the geometric feature parameters; the geometric feature parameters include the length, width, and height of the cuboid.

[0010] Furthermore, under different boundary condition parameters, temperature field simulations are performed on the parameterized 3D models of different environmental chambers, including: obtaining the value ranges of the geometric feature parameters and condition boundary parameters of the environmental chambers; using a random cross-sampling algorithm to generate multiple sets of simulation parameter combinations within the value ranges; and calling the simulation solver to perform simulation solutions on each set of simulation parameter combinations to obtain the 3D coordinates and temperature of each coordinate point inside the corresponding environmental chamber.

[0011] Furthermore, the deep neural network model adopts a fully connected neural network architecture, including an input layer, multiple hidden layers, and an output layer; the input layer is configured to receive simulation parameters and the three-dimensional coordinates of the coordinate point to be measured; the output layer is configured to output the temperature of the coordinate point to be measured; and the hidden layers use a nonlinear activation function for feature transfer.

[0012] Furthermore, the nonlinear activation function is a hyperbolic tangent function, and the deep neural network model includes at least three hidden layers.

[0013] Furthermore, training the pre-built deep neural network model using the training dataset includes: dividing the training dataset into a training set and a validation set according to a preset ratio; training the deep neural network model using the training set, and then validating the deep neural network model using the validation set; stopping training when the training loss and validation loss reach a preset stopping condition.

[0014] Furthermore, training is stopped when the training loss and validation loss reach preset stopping conditions, including stopping the training of the deep neural network model when the training loss and validation loss are less than the corresponding preset thresholds or when the maximum number of iterations is reached.

[0015] Furthermore, after obtaining the temperature field of the environment chamber to be simulated, the method further includes: visualizing the temperature field of the environment chamber to be simulated.

[0016] In a second aspect, the present invention provides an environmental chamber temperature field simulation system based on a deep neural network, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the environmental chamber temperature field simulation method based on a deep neural network described in the first aspect is implemented.

[0017] The beneficial effects of this invention are as follows: Compared with the existing technology that uses physical models such as CFD for temperature field simulation, this invention uses a trained relation fitting model to simulate the temperature field of environmental chambers under different working conditions and different sizes. It can also ensure that the simulation results are close to the accuracy of the simulation results of traditional physical models, while greatly improving the simulation calculation efficiency. Attached Figure Description

[0018] Figure 1 This is a flowchart of the environmental chamber temperature field simulation method based on deep neural networks in an embodiment of the present invention; Figure 2 This is a temperature field distribution map of an environmental chamber obtained by the environmental chamber temperature field simulation method based on a deep neural network in an embodiment of the present invention. Figure 3 This is a structural block diagram of the environmental chamber temperature field simulation system based on deep neural networks in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0021] Figure 1 This is a flowchart of the environmental chamber temperature field simulation method based on deep neural networks in an embodiment of the present invention.

[0022] In existing technologies, environmental chamber temperature field simulation is mainly performed using physical models such as CFD. However, the simulation efficiency of physical models is low, making it difficult to meet the needs of rapid product iteration. Furthermore, data models are built using a large amount of experimental data to achieve rapid prediction of the environmental chamber temperature field. However, this method is highly dependent on data, requiring a large amount of labeled data to train the model, and obtaining publicly available experimental data for environmental chambers is costly. In addition, these models lack generalization ability and are difficult to adapt to the complex geometry and multi-physics coupling of environmental chambers. Therefore, to address one or more of the above technical problems, in the first aspect, this invention provides an environmental chamber temperature field simulation method based on deep neural networks. Specifically, as... Figure 1 As shown, the method of the present invention includes the following steps.

[0023] S1. Obtain the training dataset obtained by temperature field simulation from different environmental chambers.

[0024] In this embodiment, the training dataset includes simulation parameters of the environmental chamber, the three-dimensional coordinates of each coordinate point within the environmental chamber, and the temperature. The simulation parameters include geometric feature parameters and boundary condition parameters.

[0025] In one embodiment, obtaining a training dataset obtained by simulating the temperature field from different environmental chambers includes the following steps.

[0026] S11. Construct a parametric 3D model of the environment box.

[0027] In this embodiment, the parametric 3D model defines the spatial structure of the environment box through geometric feature parameters and defines the operating conditions of the environment box through boundary condition parameters.

[0028] Specifically, the geometry of the environmental chamber is simplified into a cuboid composed of multiple cavities, and an air outlet and return air inlet are constructed on this cuboid to obtain the simplified environmental chamber. For example, this cuboid can be composed of two cavities of different sizes. The larger cavity is the experimental area, and the smaller cavity is the mechanical area. During simplification, only the inner chamber, the air duct structure, and components that affect the temperature field simulation results of the environmental chamber are retained. The dimensions of other structures within the environmental chamber are constant values ​​or calculated using the dimensions of the cuboid; for example, the dimensions of the air outlet and return air inlet can be calculated using the corresponding proportions of the cuboid's dimensions.

[0029] Furthermore, the simplified environment box is modeled using 3D modeling software, and the dimensions of the cuboid are defined as geometric feature parameters. These geometric feature parameters include the length, width, and height of the cuboid, and are all variable parameters, meaning the user can set the length, width, and height of the cuboid according to actual needs. Moreover, 3D modeling software such as Solidworks and CATIA can be used; no specific restrictions are imposed here.

[0030] Simplifying the environment chamber reduces the number of meshes, thereby improving simulation efficiency. By parameterizing the 3D model of the environment chamber, users can flexibly set its size during subsequent simulations, allowing for temperature field simulations of environment chambers with different sizes to obtain different simulation data, which in turn improves the accuracy of predictions by subsequent deep neural network models.

[0031] S12, Define boundary condition parameters.

[0032] In this embodiment, the boundary condition parameters include the vent temperature and vent velocity, which are also variable parameters. Preferably, the vent temperature and vent velocity are the same as the outlet temperature and the outlet velocity, respectively. Furthermore, the boundary condition parameters may also include the material properties of the environmental chamber and the fluid. It should be noted that, except for the vent temperature and vent velocity, which are variables, the other boundary condition parameters are default values, manually set, and only the vent temperature and vent velocity are used for training the deep neural network model.

[0033] By defining boundary condition parameters as variables, temperature field simulations of environmental chambers under different working conditions can be performed, thereby improving the flexibility of environmental chamber temperature field simulations.

[0034] S13. Under different boundary condition parameters, perform temperature field simulation on the parameterized three-dimensional models of different environmental chambers to obtain the three-dimensional coordinates and temperature of the coordinate points corresponding to each position inside the different environmental chambers.

[0035] It is understandable that during simulation, an environment box corresponds to a set of geometric feature parameters and boundary condition parameters, that is, a set of simulation parameter combinations; different environment boxes correspond to different simulation parameter combinations.

[0036] Specifically, the parametric 3D model of the environment box is input into the simulation software, and then the range of values ​​for the geometric feature parameters and condition boundary parameters of the environment box is obtained. This range of values ​​is input by the user or uses the default values ​​that the user has set in advance on the simulation software.

[0037] In one embodiment, the length, width, and height of the cuboid can all range from 1000mm to 3000mm, the air outlet temperature can range from -40℃ to +150℃, and the air outlet velocity can range from 1m / s to 10m / s. The simulation software can be Ansys or Comsol. Furthermore, the initial values ​​for the length, width, and height of the cuboid are set to 1000mm, the initial values ​​for the air outlet temperature and the air outlet velocity are set to -40℃ and 1m / s, respectively. In optional embodiments, those skilled in the art can set the range of simulation parameters according to actual needs.

[0038] Furthermore, a random cross-sampling algorithm is employed to randomly generate multiple sets of simulation parameter combinations within the value range corresponding to each parameter, for example, generating 30 sets. One set of simulation parameter combinations is: a=1500mm, b=1500mm, c=2000mm, Tin=80℃, Vin=2m / s. Here, a is the length of the cuboid, b is the width of the cuboid, c is the height of the cuboid, Tin is the air outlet temperature, and Vin is the air outlet velocity. In optional embodiments, those skilled in the art can use other sampling methods, which are not limited here.

[0039] Furthermore, the simulation solution tool in the simulation software is invoked to perform simulation and solution for each set of simulation parameters, obtaining the three-dimensional coordinates and temperature of all coordinate points within the environmental chamber corresponding to each set of simulation parameter combinations. The three-dimensional coordinates of the coordinate points are denoted as (x, y, z), and the temperature of the coordinate points is denoted as T. It can be understood that the simulation solution yields the three-dimensional temperature field of the environmental chamber. Since this three-dimensional temperature field consists of several consecutive coordinate points and their corresponding temperatures, the three-dimensional coordinates and temperatures of all coordinate points can be clearly identified through this three-dimensional temperature field. These coordinate points represent various locations within the internal space of the environmental chamber. The CDF solver can be used for temperature field simulation and solution, i.e., the CDF simulation method is used to simulate the temperature field of the environmental chamber. In optional embodiments, those skilled in the art can also use other existing temperature field simulation methods for solution, which are not limited here.

[0040] Compared to obtaining training data through physical experiments, this invention uses simulation software to simulate environmental chambers with different boundary conditions and sizes, which can obtain tens of thousands of coordinate points and corresponding temperatures, thus obtaining a large amount of accurate and reliable training data.

[0041] S14. Establish the correspondence between simulation parameters, the three-dimensional coordinates of coordinate points, and temperature to obtain the training dataset.

[0042] Specifically, a data table is constructed based on simulation parameters, the 3D coordinates of coordinate points, and temperature. Each parameter is a column, and they correspond one-to-one, forming a training dataset. In essence, one environmental chamber corresponds to a set of simulation parameter combinations, multiple 3D coordinates of coordinate points, and temperature. Using this method, nearly eight million data sets can be obtained.

[0043] By using simulation data obtained through temperature field simulation as a training dataset for subsequent deep neural network model training, a large amount of accurate and reliable data can be obtained without physical experiments, thereby improving the accuracy of deep neural network model prediction. Furthermore, no data labeling is required; only a one-to-one correspondence data table needs to be established for training. By using simulation parameters and the three-dimensional coordinates of coordinate points as input data, the model can adapt to the geometry of the environmental box and multi-physics coupling, thereby improving the accuracy of model prediction.

[0044] S2. Train the pre-built deep neural network model using the training dataset to obtain a model showing the fitting relationship between simulation parameters and the temperature field.

[0045] Specifically, a deep neural network model is constructed. In this embodiment, the deep neural network model adopts a fully connected neural network architecture, including an input layer, multiple hidden layers, and an output layer. The input layer is configured to receive simulation parameters and the three-dimensional coordinates of the point to be measured; the output layer is configured to output the temperature of the point to be measured; and a nonlinear activation function is used for feature transfer between the hidden layers. In one embodiment, the nonlinear activation function can be the tanh (hyperbolic tangent) activation function, and the first layer can also use the tanh activation function.

[0046] Preferably, the deep neural network model adopts a 5-layer architecture, namely one input layer, three hidden layers, and one output layer. Specifically, the first layer, i.e., the input layer, receives 8 features, namely the length, width, and height of the cuboid, the air outlet temperature and air velocity, and the three-dimensional coordinates of the coordinate point to be measured, and outputs 32 features; the second layer receives the 32 features output by the first layer and outputs 16 features; the third layer receives the 16 features output by the second layer and outputs 8 features; the fourth layer receives the 8 features output by the third layer and outputs 4 features; the fifth layer, i.e., the output layer, receives the 4 features output by the fourth layer and outputs one feature, namely the temperature of the coordinate point to be measured.

[0047] Furthermore, the training dataset is divided into a training set and a validation set according to a preset ratio. In one embodiment, the training set and validation set can be divided in a 9:1 ratio.

[0048] Furthermore, the deep neural network model is trained using the training set and then validated using the validation set. The learning rate can be set to 0.001, the batch size to 512, and the number of training epochs to 1500.

[0049] Training of the deep neural network model is stopped when the training loss and validation loss reach preset stopping conditions. Specifically, training stops when the training loss and validation loss are less than the corresponding preset thresholds, the weighted sum of the training loss and validation loss is less than the preset threshold, or the maximum number of iterations is reached, thereby obtaining a fitting model of the simulation parameters and the temperature field.

[0050] In an optional embodiment, the training loss curve and validation loss curve in the convergence graph are monitored. When both the training loss and validation loss reach their minimum, or an equilibrium point is reached, the training of the fitted relation model is complete. This equilibrium point indicates that the deep neural network model has a good learning effect and can also generalize well to new data.

[0051] Furthermore, the trained fitting relationship model is configured into the simulation software as a function to replace the original physical model for temperature field simulation. In one embodiment, the function of the fitting relationship model can be: F(x,y,z,a,b,c,Tin,Vin), where x, y, and z are the three-dimensional coordinates of the coordinate points, a, b, and c are the length, width, and height of the cuboid in the geometric feature parameters, and Tin and Vin are the vent temperature and vent wind speed in the boundary condition parameters, respectively. S3. Obtain the simulation parameters of the environment box to be simulated, and generate multiple coordinate points covering the internal space of the environment box based on the simulation parameters.

[0052] Specifically, the simulation parameters of the environment box to be simulated, input by the user, are obtained. These simulation parameters need to be within a pre-defined range to ensure the reliability of the simulation. Further, based on the simulation parameters of the environment box to be simulated, multiple coordinate points covering the internal space of the environment box to be simulated are generated according to pre-defined generation rules. For example, the origin of the coordinate axis is defined as the vertex of the lower left corner of the cuboid, and the interval between each coordinate point is defined as 1 mm. This gives the three-dimensional coordinates of the coordinate points corresponding to each position inside the environment box to be simulated. This process is implemented by the simulation software, which can automatically generate a number of consecutive coordinate points based on the dimensions of the environment box, namely the length, width and height of the cuboid.

[0053] S4. Input the simulation parameters and three-dimensional coordinates of the coordinate points of the environment to be simulated into the fitting relationship model for prediction, and obtain the temperature field of the environment to be simulated.

[0054] For example, if the user inputs the simulation parameters as follows: a=1500mm, b=1500mm, c=2000mm, Tin=80℃, Vin=2m / s, the system will generate multiple coordinate points of the environment box to be simulated based on these simulation parameters. Then, the simulation parameters of the environment box and the three-dimensional coordinates of the coordinate points will be input into the fitting relationship model for simulation and solution, resulting in 4.5m. 3 The temperature field of the environmental chamber when the air outlet temperature is 80℃ and the air outlet velocity is 2m / s.

[0055] Furthermore, after obtaining the temperature field of the environment chamber to be simulated, the method of the present invention further includes: visualizing the temperature field to finally obtain a temperature field distribution map of the environment chamber to be simulated, such as... Figure 2 As shown.

[0056] By using this fitted relationship model to simulate the temperature field of an environmental chamber, not only can the predicted results of the temperature field of environmental chambers of different sizes be obtained quickly, but the results also show a high degree of agreement with the traditional simulation calculation results based on physical models. It can be seen that the method of the present invention improves the efficiency and realism of temperature field simulation.

[0057] Figure 3 This is a structural block diagram of the environmental chamber temperature field simulation system based on deep neural networks in an embodiment of the present invention.

[0058] In a second aspect, the present invention also provides an environmental chamber temperature field simulation system based on a deep neural network. For example... Figure 3 As shown, the system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the environmental chamber temperature field simulation method based on a deep neural network as described in the first aspect of this invention.

[0059] The system also includes other components well known to those skilled in the art, such as communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0060] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions that can be stored or otherwise maintained by such a computer-readable medium.

[0061] In the description of this specification, "multiple" means at least two, such as two, three or more, etc., unless otherwise explicitly specified. Furthermore, the division of steps in the above method is only for clarity of description; in implementation, it can be combined into one step or some steps can be split into multiple steps, as long as they include the same logical relationship. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0062] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A method for simulating the temperature field of an environmental chamber based on a deep neural network, characterized in that, include: Obtain training datasets obtained from temperature field simulations using different environmental chambers; The training dataset includes simulation parameters of the environmental chamber, the three-dimensional coordinates and temperature of each coordinate point inside the environmental chamber, and the simulation parameters include geometric feature parameters and boundary condition parameters. The pre-built deep neural network model is trained using the training dataset to obtain a model showing the fitting relationship between simulation parameters and the temperature field. Obtain the simulation parameters of the environment box to be simulated, and based on the simulation parameters of the environment box to be simulated, generate multiple coordinate points covering the internal space of the environment box to be simulated; The simulation parameters and three-dimensional coordinates of the coordinate points of the environment to be simulated are input into the fitting relationship model for prediction, and the temperature field of the environment to be simulated is obtained.

2. The environmental chamber temperature field simulation method based on deep neural networks according to claim 1, characterized in that, Obtain the training dataset obtained from temperature field simulations of different environmental chambers, including: A parametric 3D model of the environment box is constructed; the parametric 3D model defines the spatial structure of the environment box through geometric feature parameters; Define boundary condition parameters; the boundary condition parameters are used to define the operating conditions of the environmental chamber, including the air outlet temperature and air outlet velocity; Temperature field simulations were performed on parametric 3D models of different environmental chambers under different boundary condition parameters to obtain the 3D coordinates and temperature of the coordinate points at various locations inside the different environmental chambers. Establish the correspondence between simulation parameters, the three-dimensional coordinates of coordinate points, and temperature to obtain the training dataset.

3. The environmental chamber temperature field simulation method based on deep neural networks according to claim 2, characterized in that, Constructing a parametric 3D model of the environment box includes: The three-dimensional structure of the environmental chamber is simplified into a cuboid composed of multiple cavities, and an air inlet and a return air inlet are constructed on the cuboid. The dimensions of the cuboid are defined as the geometric feature parameters; the geometric feature parameters include the length, width, and height of the cuboid.

4. The environmental chamber temperature field simulation method based on deep neural networks according to claim 2, characterized in that, Temperature field simulations were performed on parametric 3D models of different environmental chambers under different boundary condition parameters, including: Obtain the range of values ​​for the geometric feature parameters and condition boundary parameters of the environment box; A random cross-sampling algorithm is used to generate multiple sets of simulation parameter combinations within the specified value range; The simulation solver is invoked to perform simulation and solve for each set of simulation parameters, thereby obtaining the three-dimensional coordinates and temperature of each coordinate point inside the corresponding environmental chamber.

5. The environmental chamber temperature field simulation method based on deep neural networks according to claim 1, characterized in that, The deep neural network model adopts a fully connected neural network architecture, including an input layer, multiple hidden layers, and an output layer; the input layer is configured to receive simulation parameters and the three-dimensional coordinates of the coordinate point to be measured; the output layer is configured to output the temperature of the coordinate point to be measured; and the hidden layers use a non-linear activation function for feature transfer.

6. The environmental chamber temperature field simulation method based on deep neural networks according to claim 5, characterized in that, The nonlinear activation function is a hyperbolic tangent function, and the deep neural network model includes at least three hidden layers.

7. The environmental chamber temperature field simulation method based on deep neural networks according to claim 1, characterized in that, Training a pre-built deep neural network model using the training dataset includes: The training dataset is divided into a training set and a validation set according to a preset ratio; The deep neural network model is trained using the training set and then validated using the validation set. Training stops when the training loss and validation loss reach the preset stopping conditions.

8. The environmental chamber temperature field simulation method based on deep neural networks according to claim 1, characterized in that, Training is stopped when the training loss and validation loss reach preset stopping conditions, including stopping the training of the deep neural network model when the training loss and validation loss are less than the corresponding preset thresholds or when the maximum number of iterations is reached.

9. The environmental chamber temperature field simulation method based on deep neural networks according to claim 1, characterized in that, After obtaining the temperature field of the environment chamber to be simulated, the process also includes: visualizing the temperature field of the environment chamber to be simulated.

10. A temperature field simulation system for an environmental chamber based on a deep neural network, characterized in that, It includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the environmental chamber temperature field simulation method based on any one of claims 1-9 is implemented.

Images