A cable net structure initial prestress distribution calculation method and system

Abstract

The application provides a cable net structure initial prestress distribution calculation method and system, and the method comprises the following steps: obtaining each boundary node coordinate and each free node coordinate of the cable net structure; inputting each boundary node coordinate into a deep learning force finding model to output a predicted value of each free node coordinate; calculating the sum of the mean square residual of all free node physical balance equations and the mean square error of the predicted value and the true value of the coordinate, which is the loss function value; adjusting the parameters of the deep learning force finding model and the cable prestress according to the loss function value, updating the deep learning force finding model until the calculated loss function is less than the set error, and obtaining the cable prestress, that is, the initial prestress distribution of the cable net structure. In the application, the boundary node coordinates of the cable net structure are taken as the input of the deep learning force finding model, and the free node coordinates are taken as the output, the initial prestress distribution of the cable net structure is found by the deep learning force finding model, and compared with the existing artificial force finding, the calculation efficiency is high, and the method is fast and accurate.

Description

Technical Field

[0001] This invention relates to the field of construction, and more specifically, to a method and system for calculating the initial prestress distribution of a cable net structure. Background Technology

[0002] Cable net structures are novel, aesthetically pleasing, and lightweight, making them widely used in large-scale spatial structures. However, due to geometric nonlinearity, the geometry of this structure changes significantly from a zero-stress state to an initial prestressed state or a loaded state, affecting its load-bearing performance. Therefore, the initial prestressed state plays a crucial role in the structural performance, influencing the rationality of subsequent design and analysis.

[0003] When designing cable net structures, the geometric location of the initial prestress state is generally determined by the architectural form designed by the architect, i.e., the initial state of the cable net structure is determined. However, the initial prestress distribution is unknown and needs to be determined through force analysis. Currently, commonly used force analysis methods include the force density method, dynamic relaxation method, nonlinear finite element method, and analytical method, which have been well applied in practical engineering.

[0004] However, the above methods are either mathematically complex and difficult for structural engineers to master, or the stiffness matrix is ​​prone to singular convergence difficulties or requires repeated debugging to determine a suitable initial value. The debugging process is cumbersome and time-consuming, resulting in low efficiency in form finding of cable net structures. Summary of the Invention

[0005] For cable net structures, it is necessary to propose a quick and easy-to-use force-finding method. This invention provides a method and system for calculating the initial prestress distribution of cable net structures.

[0006] According to a first aspect of the present invention, a method for calculating the initial prestress distribution of a cable-net structure is provided, comprising:

[0007] Step 1: Given the initial geometry of the cable net structure, obtain the coordinates of each boundary node and each free node of the cable net structure.

[0008] Step 2: Set the initial parameters and initial cable prestress of the deep learning force-finding model;

[0009] Step 3: Input the coordinates of each boundary node into the deep learning force-finding model, and output the predicted coordinates of each free node.

[0010] Step 4: Calculate the mean square residual of the physical equilibrium equation for all free nodes and the mean square error between the predicted and actual coordinates of all free nodes. Based on the mean square residual of the physical equilibrium equation and the mean square error, calculate the loss function value of the deep learning force-finding model.

[0011] Step 5: Adjust the parameters of the deep learning force-finding model and the cable prestress according to the loss function value, update the deep learning force-finding model, and repeat steps 3 to 5 until the calculated loss function is less than the set error, and obtain the cable prestress, that is, the initial prestress distribution of the cable net structure.

[0012] According to a second aspect of the present invention, a system for calculating the initial prestress distribution of a cable net structure is provided, comprising:

[0013] The acquisition module is used to acquire the coordinates of each boundary node and each free node of the cable net structure when the initial geometry of the cable net structure is known.

[0014] The configuration module is used to set the initial parameters and initial cable prestress of the deep learning force-finding model;

[0015] The prediction module is used to input the coordinates of each boundary node into the deep learning force-finding model and output the predicted coordinates of each free node.

[0016] The calculation module is used to calculate the mean square residual of the physical equilibrium equation for all free nodes and the mean square error between the predicted and actual coordinates of all free nodes, and to calculate the loss function value of the deep learning force-finding model based on the mean square residual of the physical equilibrium equation and the mean square error.

[0017] The adjustment module is used to adjust the parameters and cable prestress of the deep learning force-finding model according to the loss function value, update the deep learning force-finding model, and repeatedly call the prediction module, the calculation module and the adjustment module until the calculated loss function is less than the set error, and obtain the cable prestress, that is, the initial prestress distribution of the cable net structure.

[0018] This invention provides a method and system for calculating the initial prestress distribution of a cable net structure. The method involves obtaining the coordinates of each boundary node and each free node of the cable net structure; inputting the coordinates of each boundary node into a deep learning force-finding model, and outputting the predicted coordinates of each free node; calculating the mean square residual of the physical equilibrium equations for all free nodes and the mean square error between the predicted and actual coordinates of all free nodes; calculating the loss function value of the deep learning force-finding model based on the mean square residual and mean square error of the physical equilibrium equations; adjusting the parameters of the deep learning force-finding model and the cable prestress based on the loss function value, and updating the deep learning force-finding model until the calculated loss function is less than a set error, thus obtaining the cable prestress, i.e., the initial prestress distribution of the cable net structure. This invention uses the boundary node coordinates of the cable net structure as input to the deep learning force-finding model and the free node coordinates as output, utilizing the deep learning force-finding model to find the initial prestress distribution of the cable net structure. Compared with existing manual force-finding methods, this method is more efficient, faster, and more accurate. Attached Figure Description

[0019] Figure 1 A flowchart of a method for calculating the initial prestress distribution of a cable net structure provided by the present invention;

[0020] Figure 2 A schematic diagram of the structure for finding a force model for deep learning;

[0021] Figure 3 This is a schematic diagram of the chain rule for differentiation.

[0022] Figure 4 A schematic diagram showing the node numbering, boundary nodes, and free nodes of an orthogonal cable net structure;

[0023] Figure 5 This is a schematic diagram of the initial state of an orthogonal cable net structure.

[0024] Figure 6 A schematic diagram of the force finding results for an orthogonal cable net;

[0025] Figure 7 A schematic diagram showing the node numbering, boundary nodes, and free nodes of an oblique cable net structure;

[0026] Figure 8 This is a schematic diagram of the initial state of the oblique cable net structure;

[0027] Figure 9 A schematic diagram of the force finding results for the oblique cable net;

[0028] Figure 10 A schematic diagram showing the node numbering, boundary nodes, and free nodes of an irregular cable net structure;

[0029] Figure 11 This is a schematic diagram of the initial state of an irregular cable net structure.

[0030] Figure 12 A schematic diagram of the force finding results for an irregular cable net structure;

[0031] Figure 13 This is a schematic diagram of a system for calculating the initial prestress distribution of a cable net structure, provided by the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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 embodiments of the present invention, not all embodiments. 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. In addition, the technical features of the various embodiments or individual embodiments provided by the present invention can be arbitrarily combined with each other to form feasible technical solutions. Such combinations are not constrained by the order of steps and / or structural composition patterns, but must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0033] In view of the deficiencies in the background technology, this invention proposes a deep learning force-finding model for cable net structures driven by physical laws and data coupling. This deep learning force-finding model solves for the initial prestress distribution of the cable net structure when the initial geometry is known: the boundary node coordinates are used as the input layer, the free node coordinates are used as the output layer, and the cable prestress is used as the unknown variable. F A physical equilibrium equation is established and incorporated into the loss function of the deep learning force-finding model. At the same time, the mean square error between the output value of the free node and the true value is introduced into the loss function, driving the deep learning model to reduce the loss function value through semi-supervised training and gradually approach zero, thereby achieving the force-finding purpose of the cable net structure.

[0034] Figure 1 This invention provides a flowchart of a method for calculating the initial prestress distribution of a cable net structure. The flowchart is written in Python and developed using the PyTorch library. Figure 1 As shown, the method includes:

[0035] Step 1: Given the initial geometry of the cable net structure, obtain the coordinates of each boundary node and each free node of the cable net structure.

[0036] Understandably, given that the initial geometry of the cable net structure is known, the coordinates of each boundary node and each free node of the cable net structure can be obtained separately.

[0037] Step 2: Set the initial parameters and initial cable prestress of the deep learning force-finding model.

[0038] Understandably, in a cable net structure, two nodes are connected by cable elements. To obtain the coordinates of each boundary node and each free node in the cable net structure, it is necessary to solve for the prestress of the cable elements. This invention uses a deep learning force-finding model to obtain the cable prestress of the cable net structure.

[0039] Initially, set the initial parameters and initial cable prestress of the deep learning force-finding model.

[0040] Step 3: Input the coordinates of each boundary node into the deep learning force-finding model, and output the predicted coordinates of each free node.

[0041] Among them, see Figure 2 The deep learning force-finding model includes an input layer, multiple hidden layers, and an output layer. Therefore, in step 2, when setting the initial parameters and initial cable prestress of the deep learning force-finding model, the weight coefficients and bias coefficients of each layer of the deep learning force-finding model, as well as the initial cable prestress, are set. F .

[0042] The input layer transmits the coordinates of each boundary node to multiple hidden layers. Each hidden layer receives data from the neurons of the previous hidden layer, performs weighted summation and activation processing, and then transmits the data to the neurons of the next hidden layer. After passing through multiple hidden layers, the output layer outputs the predicted coordinates of each free node.

[0043] The weighted summation and activation process for each of the hidden layers includes:

[0044] ,

[0045]

[0046] In the formula, i For the first i There are n+1 hidden layers and n+1 output layers. y i For the first i The output value of the hidden layer, z To find the output value of the force model for deep learning, For activation function, w i and b i For the first i The weight and bias coefficients of the hidden layers are parameters for finding the force model in deep learning. These are the weight coefficients of the output layer. This represents the bias coefficient of the output layer.

[0047] Step 4: Calculate the mean square residuals of the physical equilibrium equations for all free nodes and the mean square error between the predicted and actual coordinates of all free nodes. Based on the mean square residuals of the physical equilibrium equations and the mean square error, calculate the loss function value of the deep learning force-finding model.

[0048] The calculation of the mean square residuals of the physical equilibrium equations for all the free nodes includes:

[0049]

[0050]

[0051]

[0052] Calculate the mean square error between the predicted and actual values ​​of all free node coordinates, including:

[0053]

[0054] Based on the mean square residuals and mean square errors of the physical equilibrium equations, the loss function value of the deep learning force-finding model is calculated, including:

[0055] L total = MSE x + MSE y + MSE z + MSE m

[0056] In the formula, MSE x , MSE y and MSE z The mean square error of the physical equilibrium equations in the x, y, and z directions at the free nodes, respectively. MSE m The mean square error between the predicted and actual free node coordinates is given. N s The number of free nodes, m For free nodes j The number of connected cable units, L jk and F jk Each is a free node j With nodes k Between cable element length and cable force, ( x j , y j , z j ) is a free node j coordinates, ( x k , y k , zk ) is a node k coordinates, ( x j , y j , z j ) pred and u true ( x j , y j , z j ) true These are the predicted and actual values ​​of the free node coordinates, respectively.

[0057] Step 5: Adjust the parameters of the deep learning force-finding model and the cable prestress according to the loss function value, update the deep learning force-finding model, and repeat steps 3 to 5 until the calculated loss function is less than the set error, and obtain the cable prestress, that is, the initial prestress distribution of the cable net structure.

[0058] If the loss function value is greater than a set error range, then the gradient of the loss function with respect to the weight coefficients and bias coefficients of each hidden layer, as well as the cable prestress, are calculated in the backpropagation of the deep learning force-finding model using the chain rule. F The derivative is used to update the weight coefficients, bias coefficients, and cable prestress of the deep learning force-finding model through an optimization algorithm. F To update the deep learning force-finding model. Among them, such as... Figure 3 As shown, the gradient of the loss function with respect to the weights and biases of each layer is calculated by automatic differentiation through error backpropagation.

[0059] Based on the updated deep learning force-finding model, steps 2 to 5 are repeated to continuously calculate the loss function until the loss function is less than the set error, thereby obtaining the cable prestress, i.e. the initial prestress distribution of the cable-net structure.

[0060] The following describes the method for calculating the initial prestress distribution of the cable net structure provided by the present invention through several embodiments.

[0061] Example 1: Orthogonal cable net structure, node numbering and initial state of the structure are as follows: Figure 4 and Figure 5 As shown, where, as Figure 4 As shown, nodes numbered 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, and 12 are boundary nodes, and nodes numbered 13, 14, 15, 16, 17, 18, 19, 20, and 21 are free nodes. (Transverse and longitudinal cable forces) F 1 and F2. The boundary nodes are fixed on the saddle surface, and their equation is Z=(X 2 -Y 2 ) / 50, by Figure 2 Determine the coordinates of the boundary nodes. x f , y f , z f The coordinates of the free nodes are shown in Table 1.

[0062] Table 1 Initial state free node coordinates of orthogonal cable net structure

[0063]

[0064] For this orthogonal cable net structure, the calculation steps of the model in this invention are as follows:

[0065] (1) Input the coordinates of the boundary nodes ( x f , y f , z f The optimizer Adamax was selected with a learning rate of 0.01. The coordinates of the input boundary nodes were passed to the hidden layers (6 layers) and the output layer through the input layer. Each layer of neurons (32 neurons) received the data passed from the neurons in the previous layer, and after weighted summation and activation (Tanh function), it was passed to the next layer until the output data was obtained.

[0066] (2) After the data is output, the loss function is calculated. If the loss function value is greater than the set error range, the gradient of the loss function with respect to the weights and biases of each layer in the backpropagation of the deep learning model is calculated using the chain rule. F The derivative of the value is used to update the model's weights, biases, and variables through an optimization algorithm. F Recalculate.

[0067] (3) The deep learning model automatically repeats steps 1 and 2 until the loss function is less than the set error, then the output variable is changed. F This is the desired initial prestress distribution.

[0068] Figure 6 The force analysis results for the orthogonal cable net structure. Figure 6 As can be seen from this, after n=3000 training steps, F 2 / F 1 gradually approaches 1, compared to the theoretical value. F 2 / F A 1=1 match is relatively good.

[0069] Example 2: Oblique cable net structure, node numbering and initial state of the structure are as follows. Figure 7 and Figure 8 As shown, where, Figure 7 In the diagram, nodes numbered 1, 3, 4, and 5 are boundary nodes, and nodes numbered 6, 7, 8, 9, and 10 are free nodes. The cable forces are as follows: F 1. F 2. F 3. F 4 and F 5. The boundary nodes are fixed on the saddle surface, and their equation is Z=(X 2 -Y 2 ) / 50, by Figure 6 Determine the coordinates of the boundary nodes. x f , y f , z f The coordinates of the free nodes are shown in Table 2.

[0070] Table 2 Initial state free node coordinates of the skew cable net structure

[0071]

[0072] The calculation steps are the same as in Example 1.

[0073] Figure 9 The force analysis results for the oblique cable net structure. Figure 9 As can be seen from this, after n=4000 training steps, F 2 / F 1. F 3 / F 1. F 4 / F 1. F 5 / F 1 gradually approaches 1, compared to the theoretical value. F 1: F 2: F 3: F 4: F The ratio 5 = 1:1:1:1:1 matches well.

[0074] Example 3: Irregular cable net structure, node numbering and initial state of the structure are as follows: Figure 10 and Figure 11 As shown, where, Figure 10 In the diagram, nodes numbered 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 are boundary nodes, and nodes numbered 11, 12, 13, 14, 15, and 16 are free nodes. The cable forces are as follows: F 1. F 2. F 3. F 4 and F 5. By Figure 10 The coordinates of the boundary nodes are determined in Table 3. x f , y f , z f The coordinates of the free nodes are shown in Table 4.

[0075] Table 3 Z-coordinates of boundary nodes of irregular cable net structures

[0076]

[0077] Table 4 Initial state free node coordinates of irregular cable net structure

[0078]

[0079] The calculation steps are the same as in Example 1.

[0080] Figure 12 Force analysis results for irregular cable net structures. Figure 12 As can be seen from this, after n=4000 training steps, F 2 / F 1. F 3 / F 1. F 4 / F 1. F 5 / F 1 gradually approaches 1, compared to the theoretical value. F 1: F 2: F 3: F 4: F The ratio 5 = 1:1:1:1:1 matches well.

[0081] The above three embodiments all show that the force-finding results of the model of the present invention are in good agreement with the theoretical solution, with high accuracy, and the time taken is no more than 1 minute, which shows high computational efficiency.

[0082] See Figure 13 The present invention provides a calculation system for the initial prestress distribution of a cable net structure, the system comprising:

[0083] The acquisition module 1301 is used to acquire the coordinates of each boundary node and each free node of the cable net structure when the initial geometry of the cable net structure is known.

[0084] The configuration module 1302 is used to set the initial parameters and initial cable prestress of the deep learning force-finding model;

[0085] The prediction module 1303 is used to input the coordinates of each boundary node into the deep learning force-finding model and output the predicted coordinates of each free node.

[0086] The calculation module 1304 is used to calculate the mean square residual of the physical equilibrium equation of all the free nodes and the mean square error between the predicted and actual coordinates of all the free nodes, and to calculate the loss function value of the deep learning force-finding model based on the mean square residual of the physical equilibrium equation and the mean square error.

[0087] The adjustment module 1305 is used to adjust the parameters and cable prestress of the deep learning force-finding model according to the loss function value, update the deep learning force-finding model, and repeatedly call the prediction module, the calculation module and the adjustment module until the calculated loss function is less than the set error, and obtain the cable prestress, that is, the initial prestress distribution of the cable net structure.

[0088] It is understood that the initial prestress distribution calculation system for cable net structures provided by the present invention corresponds to the initial prestress distribution calculation method for cable net structures provided in the foregoing embodiments. The relevant technical features of the initial prestress distribution calculation system for cable net structures can be referred to the relevant technical features of the initial prestress distribution calculation method for cable net structures, and will not be repeated here.

[0089] This invention provides a method for calculating the initial prestress distribution of a cable net structure. The method involves obtaining the coordinates of each boundary node and each free node of the cable net structure; inputting the coordinates of each boundary node into a deep learning force-finding model, and outputting the predicted coordinates of each free node; calculating the mean square residual of the physical equilibrium equations for all free nodes and the mean square error between the predicted and actual coordinates of all free nodes; calculating the loss function value of the deep learning force-finding model based on the mean square residual and the mean square error; adjusting the parameters and cable prestress of the deep learning force-finding model based on the loss function value, and updating the deep learning force-finding model until the calculated loss function is less than a set error, thus obtaining the cable prestress, i.e., the initial prestress distribution of the cable net structure. This invention uses the boundary node coordinates of the cable net structure as input to the deep learning force-finding model and the free node coordinates as output, utilizing the deep learning force-finding model to find the initial prestress distribution of the cable net structure. Compared with existing manual force-finding methods, this method is more efficient, faster, and more accurate.

[0090] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0091] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0092] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0093] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0094] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0095] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0096] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for calculating the initial prestress distribution of a cable net structure, characterized in that, include: Step 1: Given the initial geometry of the cable net structure, obtain the coordinates of each boundary node and each free node of the cable net structure. Step 2: Set the initial parameters and initial cable prestress of the deep learning force-finding model; Step 3: Input the coordinates of each boundary node into the deep learning force-finding model, and output the predicted coordinates of each free node. Step 4: Calculate the mean square residual of the physical equilibrium equation for all free nodes and the mean square error between the predicted and actual coordinates of all free nodes. Based on the mean square residual of the physical equilibrium equation and the mean square error, calculate the loss function value of the deep learning force-finding model. Step 5: Adjust the parameters of the deep learning force-finding model and the cable prestress according to the loss function value, update the deep learning force-finding model, and repeat steps 3 to 5 until the calculated loss function value is less than the set error, and obtain the cable prestress, i.e. the initial prestress distribution of the cable net structure. The deep learning force-finding model includes an input layer, multiple hidden layers, and an output layer. Step 3 involves inputting the coordinates of each boundary node into the deep learning force-finding model and outputting the predicted coordinates of each free node, including: The input layer passes the coordinates of each boundary node to multiple hidden layers. The neurons in each hidden layer receive the data from the neurons in the previous hidden layer, and after weighted summation and activation processing, the data is passed to the neurons in the next hidden layer. After passing through multiple hidden layers, the output layer outputs the predicted value of the coordinates of each free node. Step 4, which calculates the mean square residuals of the physical equilibrium equations for all free nodes and the mean square error between the predicted and actual coordinates of all free nodes, includes: Based on the mean square residual of the physical equilibrium equation and the mean square error between the predicted and actual free node coordinates, the loss function value of the deep learning force-finding model is calculated, including: L total = MSE x + MSE y + MSE z + MSE m In the formula, MSE x , MSE y and MSE z The mean square residuals of the physical equilibrium equations in the x, y, and z directions at the free nodes, respectively. MSE m Let $\mathbf{ ... N s The number of free nodes, m For free nodes j The number of connected cable units, L jk and F jk Each is a free node j With nodes k Between cable element length and cable force, ( x j , y j , z j ) is a free node j coordinates, ( x k , y k , z k ) is a node k coordinates, ( x j , y j , z j ) pred and u true ( x j , y j , z j ) true These are the predicted and actual values ​​of the free node coordinates, respectively.

2. The method for calculating the initial prestress distribution of a cable net structure according to claim 1, characterized in that, Step 2, setting the initial parameters and initial cable prestress of the deep learning force-finding model, includes: Set the weight coefficients and bias coefficients for each layer of the deep learning force-finding model, as well as the initial cable prestress. F .

3. The method for calculating the initial prestress distribution of a cable net structure according to claim 1, characterized in that, The weighted summation and activation process for each of the hidden layers includes: ， In the formula, i For the first i There are n+1 hidden layers and n+1 output layers. y i For the first i The output value of the hidden layer, z To find the output value of the force model for deep learning, For activation function, w i and b i For the first i The weight and bias coefficients of the hidden layers are parameters for finding the force model in deep learning. These are the weight coefficients of the output layer. This represents the bias coefficient of the output layer.

4. The method for calculating the initial prestress distribution of a cable net structure according to claim 1, characterized in that, Step 5, adjusting the parameters and cable prestress of the deep learning force-finding model based on the loss function value, and updating the deep learning force-finding model, includes: If the loss function value is greater than the set error range, then the gradient of the loss function with respect to the weight coefficients and bias coefficients of each layer, as well as the cable prestress, are calculated in the backpropagation of the deep learning force-finding model using the chain rule. F The derivative is used to update the weight coefficients, bias coefficients, and cable prestress of the deep learning force-finding model through an optimization algorithm. F .

5. A calculation system for the initial prestress distribution of a cable net structure, characterized in that, include: The acquisition module is used to acquire the coordinates of each boundary node and each free node of the cable net structure when the initial geometry of the cable net structure is known. The configuration module is used to set the initial parameters and initial cable prestress of the deep learning force-finding model; The prediction module is used to input the coordinates of each boundary node into the deep learning force-finding model and output the predicted coordinates of each free node. The calculation module is used to calculate the mean square residual of the physical equilibrium equation for all free nodes and the mean square error between the predicted and actual coordinates of all free nodes, and to calculate the loss function value of the deep learning force-finding model based on the mean square residual of the physical equilibrium equation and the mean square error. The adjustment module is used to adjust the parameters and cable prestress of the deep learning force-finding model according to the loss function value, update the deep learning force-finding model, and repeatedly call the prediction module, the calculation module and the adjustment module until the calculated loss function is less than the set error, and obtain the cable prestress, that is, the initial prestress distribution of the cable net structure. The deep learning force-finding model includes an input layer, multiple hidden layers, and an output layer. The prediction module is used to input the coordinates of each boundary node into the deep learning force-finding model and output the predicted coordinates of each free node, including: The input layer passes the coordinates of each boundary node to multiple hidden layers. The neurons in each hidden layer receive the data from the neurons in the previous hidden layer, and after weighted summation and activation processing, the data is passed to the neurons in the next hidden layer. After passing through multiple hidden layers, the output layer outputs the predicted value of the coordinates of each free node. The calculation module is used to calculate the mean square residuals of the physical equilibrium equations for all free nodes and the mean square error between the predicted and actual coordinates of all free nodes, including: Based on the mean square residual of the physical equilibrium equation and the mean square error between the predicted and actual free node coordinates, the loss function value of the deep learning force-finding model is calculated, including: L total = MSE x + MSE y + MSE z + MSE m In the formula, MSE x , MSE y and MSE z The mean square residuals of the physical equilibrium equations in the x, y, and z directions at the free nodes, respectively. MSE m Let $\mathbf{ ... N s The number of free nodes, m For free nodes j The number of connected cable units, L jk and F jk Each is a free node j With nodes k Between cable element length and cable force, ( x j , y j , z j ) is a free node j coordinates, ( x k , y k , z k ) is a node k coordinates, ( x j , y j , z j ) pred and u true ( x j , y j , z j ) true These are the predicted and actual values ​​of the free node coordinates, respectively.

Images