Component one-dimensional simulation calculation method and system based on finite element analysis simulation and neural network prediction

Abstract

The invention discloses a finite element analysis simulation and neural network prediction-based component one-dimensional simulation calculation method and system, relates to the technical field of one-dimensional simulation, and is used for solving the technical problems of large calculation amount and low simulation efficiency of the existing three-dimensional simulation technology. The component one-dimensional simulation calculation method based on finite element analysis simulation and neural network prediction comprises the steps that a first heterogeneous system is determined; establishing a three-dimensional geometric model of each component by using COMSOL software; performing thermal simulation on each component by using COMSOL software to obtain three-dimensional simulation data of each component; training a neural network by taking the three-dimensional simulation data as a data set for training the neural network; after training is finished, a neural network prediction model is obtained; inputting simulation data of a to-be-calculated component into the neural network prediction model, and outputting characteristic parameters of the corresponding component by the neural network prediction model; and inputting the characteristic parameters into a one-dimensional simulation platform, and carrying out component calculation by the one-dimensional simulation platform according to the characteristic parameters.

Description

Technical Field

[0001] The present invention relates to the technical field of one-dimensional simulation, and more specifically, to a one-dimensional simulation calculation method and system for components based on finite element analysis simulation and neural network prediction. Background Art

[0002] With the continuous progress of existing engineering and scientific research, simulation technology plays a crucial role in the design, optimization, and verification processes of various physical systems; especially in multi-physics field simulation, three-dimensional simulation models (such as finite element analysis software like COMSOL) are widely used in fields such as heat, mechanics, electromagnetics, and fluid mechanics to simulate complex physical phenomena.

[0003] Due to the gradually increasing complexity of simulation problems, especially in the case of multi-physics field coupling, traditional three-dimensional simulation technology usually requires very fine mesh division and a large amount of computing resources. Especially in the coupled analysis of multi-physics fields, the amount of calculation will increase exponentially, resulting in a large amount of calculation and low simulation efficiency. Summary of the Invention

[0004] The purpose of the present invention is to provide a one-dimensional simulation calculation method and system for components based on finite element analysis simulation and neural network prediction, which is used to solve the technical problems of large calculation amount and low simulation efficiency of existing three-dimensional simulation technology. In view of this, the present invention is realized through the following solutions.

[0005] In the first aspect, the present invention provides a one-dimensional simulation calculation method for components based on finite element analysis simulation and neural network prediction, including: Determine a first heterogeneous system; the first heterogeneous system includes multiple components; Use COMSOL software to establish a three-dimensional geometric model for each component; each component has pressure loss and heat transfer characteristics; Use the COMSOL software to perform thermal simulation on each component to obtain three-dimensional simulation data of each component; the three-dimensional simulation data includes the working conditions of the corresponding component, as well as pressure drop and temperature rise; Use the three-dimensional simulation data as a data set for training a neural network to train the neural network, and use the working conditions as the input of the neural network. The neural network outputs the pressure drop and temperature rise corresponding to the working conditions; after the training is completed, a neural network prediction model is obtained; Input the simulation data of the component to be calculated into the neural network prediction model, and the neural network prediction model outputs the characteristic parameters of the corresponding component; the characteristic parameters include pressure drop and temperature rise; Input the characteristic parameters into a one-dimensional simulation platform, and the one-dimensional simulation platform performs component calculation according to the characteristic parameters.

[0006] Compared with the prior art, in the one-dimensional simulation calculation method of components based on finite element analysis simulation and neural network prediction of the present invention, the first heterogeneous system includes multiple components; after establishing a three-dimensional geometric model of each component through COMSOL software, thermal simulation is performed on each component, and three-dimensional simulation data of each component can be obtained; further, this three-dimensional simulation data will be used as a data set for training the neural network to train the neural network. During the process of training the neural network, the working conditions are used as the input of the neural network, and the neural network outputs the voltage drop and temperature rise corresponding to the working conditions, thereby obtaining the neural network prediction model; after inputting the simulation data of the component to be calculated into the neural network prediction model, the neural network prediction model can output the characteristic parameters of the corresponding component; further, these characteristic parameters will be input into the one-dimensional simulation platform, and the one-dimensional simulation platform can perform component calculations according to the characteristic parameters. In the above process, through the neural network prediction model obtained by training, key features and effective parameters can be extracted from a large amount of three-dimensional simulation data, converting the complex three-dimensional simulation into a one-dimensional simulation with a smaller amount of calculation, thereby significantly reducing the amount of calculation and improving the simulation efficiency. In the present invention, the finite element analysis is implemented through COMSOL software. Through the above technical solution of the present invention, the technical problems of large calculation amount and low simulation efficiency of the existing three-dimensional simulation technology are solved.

[0007] Further, in the one-dimensional simulation calculation method of components based on finite element analysis simulation and neural network prediction of the present invention, the one-dimensional simulation platform performs component calculations according to the characteristic parameters, including: The one-dimensional simulation platform performs integrated calculations on the second heterogeneous system according to the characteristic parameters and performs thermal analysis on the second heterogeneous system; wherein, the second heterogeneous system is composed of multiple components to be calculated.

[0008] Further, in the one-dimensional simulation calculation method of components based on finite element analysis simulation and neural network prediction of the present invention, the neural network is a BP neural network.

[0009] Further, in the one-dimensional simulation calculation method of components based on finite element analysis simulation and neural network prediction of the present invention, during the process of using the three-dimensional simulation data as a data set for training the neural network to train the neural network, the BP neural network performs layer-by-layer calculations through the input layer, hidden layer, and output layer, and uses the gradient descent method to optimize the weights and biases of the network.

[0010] Further, in the one-dimensional simulation calculation method of components based on finite element analysis simulation and neural network prediction of the present invention, before using the three-dimensional simulation data as a data set for training the neural network to train the neural network, it further includes: After normalizing the three-dimensional simulation data, it is divided into a training set, a validation set, and a test set.

[0011] Further, in the one-dimensional simulation calculation method of components based on finite element analysis simulation and neural network prediction of the present invention, the proportions of the training set, the validation set, and the test set in the three-dimensional simulation data are 70%, 15%, and 15% respectively; among them, the training set is used to train the neural network, the validation set is used to adjust the hyperparameters of the neural network, and the test set is used to evaluate the performance of the neural network prediction model obtained after training.

[0012] Further, in the one-dimensional simulation calculation method of components based on finite element analysis simulation and neural network prediction of the present invention, the number of neurons in the input layer of the neural network is consistent with the feature dimension of the input data; the number of neurons in the output layer of the neural network is consistent with the dimension of the output data; the number of neurons in the hidden layer of the neural network is 15.

[0013] Further, in the one-dimensional simulation calculation method of components based on finite element analysis simulation and neural network prediction of the present invention, the neural network is created using the newff function in MATLAB.

[0014] Further, in the one-dimensional simulation calculation method of components based on finite element analysis simulation and neural network prediction of the present invention, after obtaining the neural network prediction model, it further includes: Using the validation set to evaluate the performance of the neural network prediction model.

[0015] Further, in the one-dimensional simulation calculation method of components based on finite element analysis simulation and neural network prediction of the present invention, the process of using the validation set to evaluate the performance of the neural network prediction model is: determining the evaluation index as the mean square error, and determining the prediction accuracy of the neural network prediction model by obtaining the mean square error of the neural network prediction model; when the prediction accuracy does not meet the requirements, re-determine the three-dimensional simulation data to input the neural network prediction model for training until the prediction accuracy meets the requirements, and obtain a new neural network prediction model.

[0016] In the second aspect, the present invention provides a one-dimensional simulation calculation system of components based on finite element analysis simulation and neural network prediction, including: A heterogeneous system determination module, configured to: determine a first heterogeneous system; the first heterogeneous system includes multiple components; A geometric model establishment module, configured to: establish a three-dimensional geometric model of each component using COMSOL software; each of the components has a pressure loss and a heat transfer characteristic; The simulation data acquisition module is used for: performing thermal simulation on each component by using the COMSOL software to obtain three-dimensional simulation data of each component; the three-dimensional simulation data includes the working conditions of the corresponding component, as well as the voltage drop and temperature rise. The neural network training module is used for: using the three-dimensional simulation data as a dataset for training a neural network to train the neural network, and using the working conditions as the input of the neural network, and the neural network outputs the voltage drop and temperature rise corresponding to the working conditions; after the training is completed, a neural network prediction model is obtained. The characteristic parameter acquisition module is used for: inputting the simulation data of the component to be calculated into the neural network prediction model, and the neural network prediction model outputs the characteristic parameters of the corresponding component; the characteristic parameters include the voltage drop and temperature rise. The component calculation module is used for: inputting the characteristic parameters into a one-dimensional simulation platform, and the one-dimensional simulation platform performs component calculation according to the characteristic parameters. Description of the Drawings

[0017] The drawings described herein are used to provide a further understanding of the present invention, and constitute a part of the present invention. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 It is a schematic flow chart of the one-dimensional simulation calculation method of components based on finite element analysis simulation and neural network prediction of the present invention; Figure 2 It is a schematic composition diagram of the one-dimensional simulation calculation system of components based on finite element analysis simulation and neural network prediction of the present invention. Detailed Embodiments

[0018] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0019] It should be noted that when an element is referred to as being "fixed to" or "disposed on" another element, it can be directly on the other element or indirectly on the other element. When an element is referred to as being "connected to" another element, it can be directly connected to the other element or indirectly connected to the other element.

[0020] In addition, the terms "first" and "second" are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, "a plurality of" means two or more, unless otherwise specifically defined. "Several" means one or more, unless otherwise specifically defined.

[0021] Due to the gradually increasing complexity of simulation problems, especially in the case of multi-physics field coupling, traditional three-dimensional simulation technologies usually require very fine mesh generation and a large amount of computing resources. Especially in the coupled analysis of multi-physics fields, the amount of calculation will increase exponentially, resulting in a large amount of calculation and low simulation efficiency.

[0022] To solve the above technical problems, please refer to Figure 1 , the present invention provides a one-dimensional simulation calculation method for components based on finite element analysis simulation and neural network prediction, including: S100, determining a first heterogeneous system; the first heterogeneous system includes a plurality of components; S200, establishing a three-dimensional geometric model of the component; specifically, using COMSOL software to establish a three-dimensional geometric model of each component; each of the components has pressure loss and heat transfer characteristics; S300, obtaining three-dimensional simulation data of the component; specifically, using the COMSOL software to perform thermal simulation on each component to obtain three-dimensional simulation data of each component; the three-dimensional simulation data includes the working conditions of the corresponding component, as well as pressure drop and temperature rise; S400, obtaining a neural network prediction model; specifically, using the three-dimensional simulation data as a data set for training the neural network to train the neural network, and using the working conditions as the input of the neural network, and the neural network outputs the pressure drop and temperature rise corresponding to the working conditions; after the training is completed, a neural network prediction model is obtained; S500, obtaining characteristic parameters of the component; specifically, inputting the simulation data of the component to be calculated into the neural network prediction model, and the neural network prediction model outputs the characteristic parameters of the corresponding component; the characteristic parameters include pressure drop and temperature rise; S600, performing component calculation; specifically, inputting the characteristic parameters into a one-dimensional simulation platform, and the one-dimensional simulation platform performs component calculation according to the characteristic parameters.

[0023] In the case of adopting the above technical solution, in the one-dimensional simulation calculation method of components based on finite element analysis simulation and neural network prediction of the present invention, the first heterogeneous system includes multiple components; after establishing a three-dimensional geometric model of each component through COMSOL software, thermal simulation is performed on each component, and three-dimensional simulation data of each component can be obtained; further, the three-dimensional simulation data will be used as a data set for training the neural network to train the neural network. During the process of training the neural network, the working conditions are used as the input of the neural network, and the neural network outputs the pressure drop and temperature rise corresponding to the working conditions, so as to obtain the neural network prediction model; after inputting the simulation data of the component to be calculated into the neural network prediction model, the neural network prediction model can output the characteristic parameters of the corresponding component; further, the characteristic parameters will be input into the one-dimensional simulation platform, and the one-dimensional simulation platform can perform component calculations according to the characteristic parameters. In the above process, through the trained neural network prediction model, key features and effective parameters can be extracted from a large amount of three-dimensional simulation data, and the complex three-dimensional simulation is transformed into a one-dimensional simulation with a smaller amount of calculation, thereby significantly reducing the amount of calculation and improving the simulation efficiency. Through the above technical solution of the present invention, the technical problems of large calculation amount and low simulation efficiency of the existing three-dimensional simulation technology are solved.

[0024] To better understand the present invention, the content of the present invention will be further clarified below in conjunction with specific embodiments, but the content of the present invention is not limited to the following embodiments. Embodiment 1

[0025] This embodiment provides a one-dimensional simulation calculation method of components based on finite element analysis simulation and neural network prediction, including: Step 1, determine the first heterogeneous system; the first heterogeneous system includes multiple components; Step 2, use COMSOL software to establish a three-dimensional geometric model of each component; each component has pressure loss and heat transfer characteristics; Step 3, use the COMSOL software to perform thermal simulation on each component to obtain three-dimensional simulation data of each component; the three-dimensional simulation data includes the working conditions of the corresponding component, as well as the pressure drop and temperature rise; Step 4, use the three-dimensional simulation data as a data set for training the neural network to train the neural network, and use the working conditions as the input of the neural network, and the neural network outputs the pressure drop and temperature rise corresponding to the working conditions; after the training is completed, a neural network prediction model is obtained; Step 5, input the simulation data of the component to be calculated into the neural network prediction model, and the neural network prediction model outputs the characteristic parameters of the corresponding component; the characteristic parameters include the pressure drop and temperature rise; Step 6, input the characteristic parameters into a one-dimensional simulation platform, and the one-dimensional simulation platform performs component calculations based on the characteristic parameters. Embodiment 2

[0026] This embodiment provides a one-dimensional simulation calculation method for components based on finite element analysis simulation and neural network prediction, including: S100, determine a first heterogeneous system; the first heterogeneous system includes multiple components; S200, use COMSOL software to establish a three-dimensional geometric model for each component; each component has pressure loss and heat transfer characteristics; S300, use the COMSOL software to perform thermal simulation on each component to obtain three-dimensional simulation data of each component; the three-dimensional simulation data includes the working conditions of the corresponding component, as well as pressure drop and temperature rise; S400, after normalizing the three-dimensional simulation data, divide it into a training set, a validation set, and a test set; the proportions of the training set, validation set, and test set in the three-dimensional simulation data are 70%, 15%, and 15% respectively; Among them, the training set is used to train the neural network, the validation set is used to adjust the hyperparameters of the neural network, and the test set is used to evaluate the performance of the neural network prediction model obtained after training; S500, use the three-dimensional simulation data as the dataset for training the neural network to train the neural network, and use the working conditions as the input of the neural network, and the neural network outputs the pressure drop and temperature rise corresponding to the working conditions; after training, a neural network prediction model is obtained; Among them, the neural network is a BP neural network; the BP neural network performs layer-by-layer calculations through an input layer, a hidden layer, and an output layer, and uses the gradient descent method to optimize the weights and biases of the network; the number of neurons in the input layer of the neural network is consistent with the feature dimension of the input data; the number of neurons in the output layer of the neural network is consistent with the dimension of the output data; the number of neurons in the hidden layer of the neural network is 15; the neural network is created using the newff function in MATLAB; S600, use the validation set to evaluate the performance of the neural network prediction model; S601, determine the evaluation index as the mean square error, and determine the prediction accuracy of the neural network prediction model by obtaining the mean square error of the neural network prediction model; S602, when the prediction accuracy does not meet the requirements, re-determine the three-dimensional simulation data to input the neural network prediction model for training until the prediction accuracy meets the requirements, and obtain a new neural network prediction model; S700. Input the simulation data of the component to be calculated into the neural network prediction model, and the neural network prediction model outputs the characteristic parameters of the corresponding component; the characteristic parameters include voltage drop and temperature rise. S800. Input the characteristic parameters into a one-dimensional simulation platform. The one-dimensional simulation platform performs integrated calculations on the second heterogeneous system according to the characteristic parameters and conducts thermal analysis on the second heterogeneous system; wherein, the second heterogeneous system is composed of multiple components to be calculated. Embodiment 3

[0027] In a first aspect, this embodiment provides a one-dimensional simulation calculation method for components based on finite element analysis simulation and neural network prediction, including: S100. Determine the first heterogeneous system; the first heterogeneous system includes multiple components; the components include heat sources, heat sinks, and heat conduction channels, and the heat conduction channels include square straight channels, circular elbows, square tapers, and circular mutations. S200. Use COMSOL software to establish a three-dimensional geometric model for each component; each component has pressure loss and heat transfer characteristics. S300. Use the COMSOL software to perform thermal simulation on each component, simulate the pressure drop loss, temperature rise, and the resulting heat dissipation when ethylene glycol passes through, and obtain the three-dimensional simulation data of each component; the three-dimensional simulation data includes the working conditions of the corresponding component, as well as the voltage drop and temperature rise. S400. After normalizing the three-dimensional simulation data, divide it into a training set, a validation set, and a test set; the proportions of the training set, validation set, and test set in the three-dimensional simulation data are 70%, 15%, and 15% respectively. Among them, the training set is used to train the neural network, the validation set is used to adjust the hyperparameters of the neural network, and the test set is used to evaluate the performance of the neural network prediction model obtained after training. S500. Use the three-dimensional simulation data as the dataset for training the neural network to train the neural network, and use the working conditions as the input of the neural network. The neural network outputs the voltage drop and temperature rise corresponding to the working conditions; after the training is completed, a neural network prediction model is obtained. Among them, the neural network is a backpropagation (BP) neural network; the BP neural network performs layer-by-layer calculations through an input layer, a hidden layer, and an output layer, and uses the gradient descent method to optimize the weights and biases of the network; the number of neurons in the input layer of the neural network is the same as the feature dimension of the input data; the number of neurons in the output layer of the neural network is the same as the dimension of the output data; the number of neurons in the hidden layer of the neural network is 15; the neural network is created using the newff function in MATLAB. Furthermore, in this embodiment, the BP neural network model is trained. The BP neural network performs layer-by-layer calculations through the input layer, hidden layer, and output layer, and uses the gradient descent method to optimize the weights and biases of the network to minimize the prediction error. The specific steps of the training process are as follows: S501, Data collection: Obtain the simulation data of the components through COMSOL simulation. The simulation data includes input features (geometric dimensions, flow resistance coefficient) and output targets (pressure drop, temperature rise); Data normalization: To avoid biases in the training process due to differences in data scales, normalize the simulation data; Dataset division: Divide the processed simulation data into a training set, a validation set, and a test set, and divide them according to the ratio of 70%, 15%, and 15%. The training set is used for network training, the validation set is used to adjust hyperparameters, and the test set is used to finally evaluate the model performance; S502, Network topology: Construct a BP neural network with a single hidden layer; The number of neurons in the input layer is consistent with the feature dimension of the input data; The number of neurons in the output layer is the same as the dimension of the target data; The number of neurons in the hidden layer is set according to experimental experience and is set to 15; Network model establishment: Use the newff function in MATLAB to create a neural network; The newff function is used to construct a feedforward neural network and is suitable for regression problems; The activation function is the Sigmoid function or the ReLU function; S503, Training method selection: Adopt the Levenberg-Marquardt algorithm as the backpropagation training algorithm; This algorithm minimizes the mean squared error (MSE) by adjusting the network weights and biases, effectively improving the network convergence speed; Training process: Use the train function in MATLAB to train the neural network; During the training process, the network will continuously update the parameters through forward propagation and backpropagation to reduce the prediction error; Training process monitoring: Monitor the network performance during the training process by plotting the training error, validation error, and test error curves to ensure that overfitting or underfitting phenomena are avoided; S600, Use the validation set to evaluate the performance of the neural network prediction model; S601, Determine that the evaluation index is the mean squared error, and determine the prediction accuracy of the neural network prediction model by obtaining the mean squared error of the neural network prediction model; S602, When the prediction accuracy does not meet the requirements, re-determine the three-dimensional simulation data and input it into the neural network prediction model for training until the prediction accuracy meets the requirements, and obtain a new neural network prediction model; Among them, the mean squared error (MSE) is a common index to measure the error between the model prediction result and the actual value; Its calculation formula is as follows: ; Among them, represents the mean square error, n is the total number of samples, is the i true value of the th sample, i is the predicted value of the th sample; Exemplarily, by obtaining the mean square error of the neural network prediction model to determine the prediction accuracy of the neural network prediction model, it can be: setting the threshold of the target accuracy to 0.05, comparing the MSE value of the model with the predetermined target accuracy value. If the MSE of the model ≤ 0.05, then it is considered that the neural network prediction model meets the accuracy requirements. If the MSE of the model is greater than the preset target threshold, it indicates that the prediction accuracy of the model does not meet the requirements and needs to be further optimized; S700, input the simulation data of the component to be calculated into the neural network prediction model, and the neural network prediction model outputs the characteristic parameters of the corresponding component; the characteristic parameters include voltage drop and temperature rise; S800, input the characteristic parameters into a one-dimensional simulation platform, and the one-dimensional simulation platform performs integrated calculation on the second heterogeneous system according to the characteristic parameters and conducts thermal analysis on the second heterogeneous system; wherein, the second heterogeneous system is composed of multiple components to be calculated; In step S800, the input of the one-dimensional simulation platform: the characteristic parameters (such as voltage drop, temperature rise, etc.) of each component predicted by the neural network prediction model are input into the one-dimensional simulation platform; this one-dimensional simulation platform can accept the characteristic parameters of multiple components and simulate their working states in the entire second heterogeneous system; specifically, the system-level simulation of the one-dimensional simulation platform is: on the one-dimensional simulation platform, integrated calculation is performed on the second heterogeneous system; through the characteristic parameters of the components provided by the neural network, this one-dimensional simulation platform can quickly conduct thermal analysis of the entire system, such as calculating the overall temperature rise, energy efficiency, and possible overheating or overload conditions of the system; It should be noted that the above one-dimensional simulation platform is used to simulate heat transfer, heat dissipation effect, and temperature distribution in the heat dissipation calculation of the heterogeneous system; by using one-dimensional simulation, the thermal management performance of different materials, components, and heat dissipation structures under specific working conditions can be effectively predicted and optimized; the calculation process of the one-dimensional simulation platform can be divided into five parts, which are respectively: Obtain actual conditions: The one-dimensional simulation platform can be an electronic device composed of multiple different materials, which contains different electronic components (such as processors, graphics cards, etc.); these components generate heat, and heat conduction, diffusion, and dissipation to the environment are carried out through different materials (such as copper, aluminum, etc.); it mainly includes a heat source area, a heat conduction channel, a heat dissipation area, and the external environment. Among them, the heat source area (such as the processor) generates heat, and the heat power Q = 10W; the heat conduction channel has ethylene glycol as the internal heat conduction medium; the heat dissipation area (such as the heat sink) is responsible for conducting heat from the heat source area, and the material is aluminum; the external environment has a temperature of 25°C. Establish a heat conduction model: According to the heterogeneous system to be actually calculated, use the components trained in the one-dimensional simulation platform to build a simulation model, and the heat generated by the heat source area is conducted to the external environment through the heat conduction area; in the one-dimensional model, only the heat transfer along one direction needs to be concerned, that is, along the axis of the heat dissipation tube, and the following heat conduction equation is used to describe the heat transfer: ; Among them, T(x) represents the temperature at position x , Q is the heat generated by the heat source (unit power), k is the thermal conductivity of the material, A is the cross-sectional area of the heat conduction area, x is the axial distance along the heat conduction area; Solve the temperature distribution: Calculate the temperature distribution of each area. Assume that the heat power of the processor Q = 10W is transferred to the heat sink through the heat conduction area (heat dissipation tube); in the heat conduction area, the heat conduction equation can be integrated to obtain the temperature distribution; let x be the axial distance along the heat conduction area, and the cross-sectional area of the heat conduction tube is A , and the temperature distribution of this area can be calculated by the following equation: ; Among them, T(x) represents the temperature at position x , is the initial temperature of the heat source area (processor), Q is the heat generated by the heat source (unit power), k is the thermal conductivity of the material, A is the cross-sectional area of the heat conduction area, x is the axial distance along the heat conduction area; Set boundary conditions: At the heat source end (processor), set the temperature T to represent the processor temperature; at the end of the radiator, set the boundary temperature to the environmental temperature T = 25°C. Component calculation results: By solving the above heat conduction equation, the temperature distribution at each position of the component can be obtained; using a one-dimensional simulation platform, the above calculation process can be visualized, showing the temperature change diagram with position, and the heat dissipation effect can be optimized by adjusting the heat dissipation design (such as increasing the surface area of the heat sink or changing the material) to ensure that the device operates within the normal operating temperature range.

[0028] Furthermore, the above second heterogeneous system can be further optimized in design; specifically, according to the one-dimensional simulation results, the designer can further adjust the design of the above second heterogeneous system, such as selecting a more efficient heat dissipation solution, optimizing the component layout, etc., so as to quickly obtain the overall performance of the system without performing complex three-dimensional simulation calculations.

[0029] In a second aspect, please refer to Figure 2 , this embodiment provides a one-dimensional simulation calculation system for components based on finite element analysis simulation and neural network prediction, including: A heterogeneous system determination module, configured to: determine a first heterogeneous system; the first heterogeneous system includes a plurality of components; A geometric model establishment module, configured to: establish a three-dimensional geometric model of each component using COMSOL software; each of the components has pressure loss and heat transfer characteristics; A simulation data acquisition module, configured to: perform thermal simulation on each component using the COMSOL software to obtain three-dimensional simulation data of each component; the three-dimensional simulation data includes the working conditions of the corresponding component, as well as pressure drop and temperature rise; A neural network training module, configured to: use the three-dimensional simulation data as a data set for training a neural network to train the neural network, and use the working conditions as the input of the neural network, and the neural network outputs the pressure drop and temperature rise corresponding to the working conditions; after the training is completed, a neural network prediction model is obtained; A characteristic parameter acquisition module, configured to: input the simulation data of the component to be calculated into the neural network prediction model, and the neural network prediction model outputs the characteristic parameters of the corresponding component; the characteristic parameters include pressure drop and temperature rise; A component calculation module, configured to: input the characteristic parameters into a one-dimensional simulation platform, and the one-dimensional simulation platform performs component calculations according to the characteristic parameters.

[0030] In the description of the above embodiments, specific features, structures, materials, or characteristics may be combined in a suitable manner in any one or more embodiments or examples.

[0031] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims described above.

Claims

1. A one-dimensional simulation calculation method for components based on finite element analysis simulation and neural network prediction, characterized in that: include: determining a first heterogeneous system; The first heterogeneous system includes a plurality of components; Using COMSOL software to establish a three-dimensional geometric model of each component; each of the components has pressure loss and heat transfer characteristics; Using the COMSOL software to perform thermal simulation on each component to obtain three-dimensional simulation data of each component; the three-dimensional simulation data includes the working conditions, voltage drop and temperature rise of the corresponding component; The three-dimensional simulation data is used as a data set for training the neural network to train the neural network, and the working conditions are used as inputs of the neural network, and the neural network outputs the pressure drop and temperature rise corresponding to the working conditions; After the training is completed, the neural network prediction model is obtained; Inputting simulation data of components to be calculated into the neural network prediction model, the neural network prediction model outputs characteristic parameters of corresponding components; the characteristic parameters include voltage drop and temperature rise; The characteristic parameters are input into a one-dimensional simulation platform, and the one-dimensional simulation platform performs component calculations according to the characteristic parameters.

2. The one-dimensional simulation calculation method of components based on finite element analysis simulation and neural network prediction according to claim 1 is characterized in that: The one-dimensional simulation platform performs component calculation according to the characteristic parameters, including: The one-dimensional simulation platform performs integrated calculation on the second heterogeneous system according to the characteristic parameters, and performs thermal analysis on the second heterogeneous system; The second heterogeneous system is composed of a plurality of components to be calculated.

3. The one-dimensional simulation calculation method of components based on finite element analysis simulation and neural network prediction according to claim 2 is characterized in that: The neural network is a BP neural network.

4. The one-dimensional component simulation calculation method based on finite element analysis simulation and neural network prediction according to claim 3 is characterized in that: In the process of training the neural network by using the three-dimensional simulation data as a data set for training the neural network, the BP neural network performs layer-by-layer calculations through the input layer, hidden layer and output layer, and uses the gradient descent method to optimize the weights and biases of the network.

5. The one-dimensional component simulation calculation method based on finite element analysis simulation and neural network prediction according to claim 4 is characterized in that: Before using the three-dimensional simulation data as a data set for training the neural network to train the neural network, the method further includes: After normalizing the three-dimensional simulation data, the data are divided into a training set, a validation set, and a test set; the proportions of the training set, the validation set, and the test set in the three-dimensional simulation data are 70%, 15%, and 15%, respectively; The training set is used to train the neural network, the validation set is used to adjust the hyperparameters of the neural network, and the test set is used to evaluate the performance of the neural network prediction model obtained after training.

6. The one-dimensional component simulation calculation method based on finite element analysis simulation and neural network prediction according to claim 5 is characterized in that: The number of neurons in the input layer of the neural network is consistent with the characteristic dimension of the input data; the number of neurons in the output layer of the neural network is consistent with the dimension of the output data; and the number of neurons in the hidden layer of the neural network is 15.

7. The one-dimensional component simulation calculation method based on finite element analysis simulation and neural network prediction according to claim 6 is characterized in that: The neural network is created using MATLAB's newff function.

8. The one-dimensional component simulation calculation method based on finite element analysis simulation and neural network prediction according to claim 7 is characterized in that: After obtaining the neural network prediction model, the method further includes: The validation set is used to evaluate the performance of the neural network prediction model.

9. The one-dimensional component simulation calculation method based on finite element analysis simulation and neural network prediction according to claim 8, characterized in that: The process of using the validation set to evaluate the performance of the neural network prediction model is as follows: The evaluation index is determined to be the mean square error, and the prediction accuracy of the neural network prediction model is determined by obtaining the mean square error of the neural network prediction model; when the prediction accuracy does not meet the requirements, the three-dimensional simulation data is re-determined to input the neural network prediction model for training until the prediction accuracy meets the requirements and a new neural network prediction model is obtained.

10. A one-dimensional simulation calculation system for components based on finite element analysis simulation and neural network prediction, characterized in that: include: The heterogeneous system determination module is used to: determine a first heterogeneous system; The first heterogeneous system includes a plurality of components; A geometric model building module is used to: use COMSOL software to build a three-dimensional geometric model of each component; each of the components has pressure loss and heat transfer characteristics; A simulation data acquisition module is used to: perform thermal simulation on each component using the COMSOL software to obtain three-dimensional simulation data of each component; the three-dimensional simulation data includes the working conditions, voltage drop and temperature rise of the corresponding component; A neural network training module, used to: train the neural network using the three-dimensional simulation data as a data set for training the neural network, and use the working conditions as inputs of the neural network, and the neural network outputs the pressure drop and temperature rise corresponding to the working conditions; After the training is completed, the neural network prediction model is obtained; A characteristic parameter acquisition module, used to: input simulation data of components to be calculated into the neural network prediction model, and the neural network prediction model outputs characteristic parameters of corresponding components; the characteristic parameters include voltage drop and temperature rise; The component calculation module is used to: input the characteristic parameters into a one-dimensional simulation platform, and the one-dimensional simulation platform performs component calculation according to the characteristic parameters.

Images