A displacement prediction method and system for a glass curtain wall

Abstract

The application relates to the technical field of building engineering structures, and discloses a displacement prediction method and system for a glass curtain wall. Boundary displacement response information and boundary acceleration response information of the glass curtain wall under the condition of external force application are collected; a motion equation of the glass curtain wall is constructed based on a total stiffness matrix, a total mass matrix, the boundary displacement response information and the boundary acceleration response information; real-time sampling point parameters of the glass curtain wall are input into a preset full connection neural network model to obtain a current displacement prediction result; and the full connection neural network model is iteratively trained according to residual information between the motion equation and the current displacement prediction result. The residual information obtained by combining a structural dynamics model and a deep learning model is iteratively trained, so that the prediction model can adapt to different glass curtain walls, displacement prediction can be carried out in real time or near real time, the prediction accuracy is improved, and the calculation efficiency and real-time performance are enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of building engineering structure, and particularly relates to a displacement prediction method and system for glass curtain wall. BACKGROUND

[0002] In the current field of high-rise building design and construction, glass curtain wall is a popular external structure, which is widely welcomed due to its aesthetic appearance, light transmission and light self-weight. However, glass curtain wall may experience displacement and deformation when subjected to wind loads, which poses a potential threat to the safety of the building. In order to predict these displacements and deformations, two technical means are currently used, namely empirical formula calculation and manual finite element analysis.

[0003] In the first aspect, the empirical formula method is based on historical data and expert knowledge, and predicts displacement through a simplified mathematical model. However, the accuracy of the mathematical model depends on the quality and quantity of the empirical data, and it is difficult to adapt to complex actual engineering conditions. In addition, due to the insufficient consideration of the variability of wind loads and the complexity of curtain wall structure in the empirical formula method, the prediction accuracy is obviously limited.

[0004] On the other hand, the finite element analysis method can more accurately simulate the physical behavior of the curtain wall structure by establishing a detailed numerical model, thereby obtaining accurate displacement prediction. Although this method can more realistically reflect the stress and deformation characteristics of the structure, its calculation process is complex and requires a large amount of computing resources and time. The calculation amount is huge in the analysis of large-scale or complex structures.

[0005] Therefore, the existing technical problems are that it is difficult to adapt to the diversity and complexity of curtain wall structures, and for new materials or innovative design of curtain wall systems, the prediction model may not accurately reflect its actual performance, and for engineering practice requiring rapid response, the calculation efficiency of finite element analysis often cannot meet the needs of real-time or near real-time analysis. SUMMARY

[0006] Therefore, the present application provides a displacement prediction method and system for glass curtain wall to solve how to improve the real-time and accuracy of displacement prediction of glass curtain wall building structure.

[0007] In the first aspect, the present application provides a displacement prediction method for glass curtain wall, which comprises: calculating the overall stiffness matrix and overall mass matrix of the glass curtain wall based on the principle of structural dynamics;

[0008] Collecting boundary displacement response information and boundary acceleration response information of the glass curtain wall under the condition of external force application;

[0009] The motion equations of the glass curtain wall are constructed based on the overall stiffness matrix, overall mass matrix, boundary displacement response information, and boundary acceleration response information.

[0010] The current sampling time and sampling point location information of the glass curtain wall are used as real-time sampling point parameters and input into a preset fully connected neural network model to obtain the current displacement prediction result. Based on the residual information between the motion equation and the current displacement prediction result, the fully connected neural network model is iteratively trained. When the iteration termination condition is reached, the real-time displacement prediction model of the glass curtain wall is obtained.

[0011] The real-time displacement prediction result is obtained by inputting the next sampling time and sampling point location information of the glass curtain wall into the real-time displacement prediction model.

[0012] This invention calculates the overall stiffness and mass matrix of a glass curtain wall based on structural dynamics principles, and constructs motion equations by combining boundary displacement and acceleration response information. This more accurately describes the stress and deformation behavior of the curtain wall, enabling the prediction model to adapt to complex real-world engineering conditions. Furthermore, a fully connected neural network model is introduced, utilizing deep learning technology for displacement prediction. Iterative training is performed using the residual information from the motion equations, allowing the prediction model to adapt to different types of curtain wall materials. This ensures displacement prediction in real-time or near-real-time conditions. By combining the structural dynamics model with the deep learning model, both prediction accuracy and computational efficiency and real-time performance are improved.

[0013] In one optional implementation, the calculation of the overall stiffness matrix and overall mass matrix of the glass curtain wall based on structural dynamics principles includes:

[0014] Obtain the physical property parameters and geometric structure parameters of the glass curtain wall;

[0015] The structure of the glass curtain wall is divided into multiple planar units, each of which has a local coordinate system;

[0016] Calculate the element stiffness matrix and element mass matrix in the local coordinate system based on the physical property parameters and geometric parameters corresponding to each planar element;

[0017] The element stiffness matrix and element mass matrix of each planar element are used to form the overall stiffness matrix and overall mass matrix of the glass curtain wall.

[0018] The embodiment of the application divides the curtain wall structure into multiple planar units, calculates the stiffness matrix and the mass matrix of each planar unit in a local coordinate system, and then combines the matrices of the units into the overall stiffness matrix and the mass matrix. The overall stiffness matrix and the mass matrix calculated based on the structural dynamics principle can accurately reflect the characteristics of the glass curtain wall structure under external force, so as to construct a more accurate motion equation and further improve the prediction ability of the glass curtain wall model trained subsequently, and ensure that the prediction result is more accurate.

[0019] In an optional embodiment, the motion equation of the glass curtain wall is represented by the following formula:

[0020]

[0021] wherein M represents the overall mass matrix, K represents the overall stiffness matrix, represents the boundary acceleration response information, and v(t) represents the boundary displacement response information.

[0022] In the embodiment of the application, the mass matrix (M) and the stiffness matrix (K) can completely describe the physical characteristics of the entire glass curtain wall, and the boundary acceleration response and the boundary displacement response information are used to introduce the boundary conditions in the actual working condition into the motion equation, so as to ensure that the final motion equation more accurately simulates the real boundary conditions and improves the accuracy of the motion equation.

[0023] In an optional embodiment, the current sampling time and the sampling point position information of the glass curtain wall are taken as real-time sampling point parameters and input into a preset fully connected neural network model to obtain a current displacement prediction result, and the fully connected neural network model is iteratively trained according to the residual information between the motion equation and the current displacement prediction result, and when an iteration termination condition is reached, a real-time displacement prediction model of the glass curtain wall is obtained, and the method specifically comprises the following steps:

[0024] A fully connected neural network is constructed, and corresponding neural network parameters are set as initial values; wherein the neural network parameters include sampling point parameters, the number of different network layers, network width, the parameters of an activation function formula, the parameters of a loss function formula, and a network learning rate; the input data of the fully connected neural network includes sampling point parameters, and the output data of the fully connected neural network includes displacement responses corresponding to the sampling points inside the curtain wall;

[0025] In the iterative training process of the current round, the first sampling point parameters are input into the fully connected neural network to obtain a predicted displacement response corresponding to the first sampling point;

[0026] perform partial derivative analysis on the predicted displacement response according to an automatic differentiation principle to obtain a first partial derivative and a second partial derivative, the first partial derivative representing a second-order partial derivative of the displacement response with respect to time, and the second partial derivative representing a second-order partial derivative of the displacement response with respect to space;

[0027] calculate residual information of the motion equation based on the first partial derivative and the second partial derivative;

[0028] construct a probability density function according to the residual information;

[0029] continue sampling based on the probability density function to obtain second sampling point parameters, and simultaneously update a slope optimization parameter in an activation function formula by using a gradient descent algorithm;

[0030] update the second sampling point parameters to the first sampling point parameters, and continue iteration training in the next round; wherein the predicted displacement response in each round is also used to calculate a target loss function value in each round by bringing it into a loss function formula;

[0031] stop the iteration training when the target loss function value is lower than a preset loss function value, or when a round of iteration training reaches a preset iteration number, to obtain a glass curtain wall real-time displacement prediction model.

[0032] The embodiment of the present application models through a fully connected neural network, performs real-time prediction in combination with real-time data of sampling points, performs partial derivative analysis by using automatic differentiation, combines the analysis result of the partial derivative with a motion equation, constructs a probability density function, and adaptively updates sampling point parameters based on residual information in each iteration training, focuses on sampling a region in which a whole glass curtain wall sampling prediction result is likely to have a large error, and improves the training efficiency and accuracy of the whole model. In each round of iteration, the activation function and network parameters of the neural network are constantly optimized by using a gradient descent algorithm, the training process is continuously updated by calculating a loss value in each round, and the training of the model is guided by residual analysis of the motion equation, so that the displacement result predicted by the neural network conforms to the motion physical law of the glass curtain wall, and the glass curtain wall real-time displacement prediction model has higher interpretability.

[0033] In an alternative embodiment, the method further comprises:

[0034] perform feature expansion on the sampling point parameters based on a feature expansion layer in the fully connected neural network to obtain feature expansion parameters in the current round;

[0035] calculate gradient information of the target loss function value with respect to the feature expansion parameters by using a back propagation algorithm;

[0036] The weight parameters and bias parameters in the activation function formula for the next round are updated based on the gradient information.

[0037] This invention expands the dimension of the input data after the input layer by feature expansion, increasing the high-dimensional representation of the data. The backpropagation algorithm can further optimize the model's weight parameters and biases, ensuring that all parameters of the model can work in coordination, thereby improving the prediction accuracy of the trained model.

[0038] In one optional implementation, the residual information includes residual loss terms and the integral of all residual loss terms, and the probability density function is expressed by the following formula:

[0039]

[0040] Where p(x) represents the probability density function of the residual loss term, A represents the integral of all residual loss terms, ε(x) represents the residual loss term, T represents the residual sampling region, and the residual sampling region represents the sampling point domain where the residual is higher than the preset standard value; the k value represents the sampling concentration of the sampling points in the residual sampling region, and the c value represents the sampling uniformity of the sampling points in the residual sampling region.

[0041] In this embodiment of the invention, the probability density function is determined by the integral of all residual loss terms, residual loss terms, sampling concentration, sampling uniformity, and residual sampling region. It should be noted that the residual information refers to the residual information calculated based on the first and second partial derivatives and the equation of motion. Since the equation of motion of the glass curtain wall is incorporated, the constructed probability density function will be continuously updated in the iterative training, thereby making the prediction results of the trained model more accurate.

[0042] In one optional implementation, the loss function formula is:

[0043] L(t) = w cz loss IC +w0loss PDE +w b loss BC ;

[0044]

[0045] Among them, w cz w represents the initial value loss weighting coefficient. b w represents the boundary loss weighting coefficient. o The loss represents the weighting coefficients of the motion equation. IC The loss represents the initial value loss of the network. BC The loss represents the network boundary loss. PDErepresents the motion equation loss, L(t) represents the loss function of the whole network, N b represents the boundary loss function sampling area, N s represents the initial value loss function sampling area, N T represents the motion equation loss function sampling area, t represents a sampling point, R represents the whole sampling area, and T represents a residual sampling area, represents the displacement response corresponding to the sampling point inside the curtain wall predicted by the network at the initial sampling point parameter x c ,y c ,t c represents the displacement response corresponding to the sampling point inside the curtain wall predicted by the network at the initial sampling point parameter x c represents the actual displacement response of the initial sampling point, represents the displacement response corresponding to the sampling point inside the curtain wall predicted by the network at the initial sampling point parameter x b ,y b ,t b represents the displacement response corresponding to the sampling point inside the curtain wall predicted by the network at the initial sampling point parameter x b represents the actual displacement response of the initial sampling point, T represents the residual value of the motion equation to the network prediction.

[0046] In the embodiment of the application, the loss function includes initial value loss, boundary loss and motion equation loss, wherein the initial value loss and the boundary loss correspond to the displacement response of the curtain wall under specific conditions, and the motion equation loss corresponds to the matching degree between the network prediction and the physical equation of the glass curtain wall. Different weight coefficients are introduced in the loss function, and the three loss functions are balanced, so that the model prediction result is more in line with the physical motion phenomenon of the actual glass curtain wall, and the accuracy of the final trained model is improved.

[0047] In a second aspect, the application provides a displacement prediction system for a glass curtain wall, and the device comprises:

[0048] An overall matrix calculation module is configured to calculate the overall stiffness matrix and the overall mass matrix of the glass curtain wall based on the principle of structural dynamics.

[0049] A boundary response information acquisition module is configured to acquire boundary displacement response information and boundary acceleration response information of the glass curtain wall under the condition of external force application.

[0050] A motion equation construction module is configured to construct the motion equation of the glass curtain wall based on the overall stiffness matrix, the overall mass matrix, the boundary displacement response information and the boundary acceleration response information.

[0051] The iterative training module is configured to input the current sampling time and the sampling point position information of the glass curtain wall as real-time sampling point parameters into a preset full connection neural network model to obtain a current displacement prediction result, and iteratively train the full connection neural network model according to residual information between the motion equation and the current displacement prediction result, and obtain a real-time displacement prediction model of the glass curtain wall when an iterative termination condition is reached.

[0052] The real-time displacement prediction module is configured to input the next sampling time and the sampling point position information of the glass curtain wall into the real-time displacement prediction model to obtain a real-time displacement prediction result.

[0053] In a third aspect, the present application provides a computer device, comprising a memory and a processor, the memory and the processor are communicatively connected with each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the glass curtain wall displacement prediction method of the first aspect or any of the corresponding embodiments thereof.

[0054] In a fourth aspect, the present application provides a computer readable storage medium, which stores computer instructions, and the computer instructions are used to make a computer execute the glass curtain wall displacement prediction method of the first aspect or any of the corresponding embodiments thereof.

[0055] In a fifth aspect, the present application provides a computer program product, which comprises computer instructions, and the computer instructions are used to make a computer execute the glass curtain wall displacement prediction method of the first aspect or any of the corresponding embodiments thereof. BRIEF DESCRIPTION OF DRAWINGS

[0056] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the drawings needed in the specific embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0057] Figure 1 is a flowchart of a glass curtain wall displacement prediction method according to an embodiment of the present application;

[0058] Figure 2 is a flowchart of another glass curtain wall displacement prediction method according to an embodiment of the present application;

[0059] Figure 3 is a structure division schematic diagram of a glass curtain wall displacement prediction method according to an embodiment of the present application;

[0060] Figure 4is a flowchart of a displacement prediction method of another glass curtain wall according to an embodiment of the present application;

[0061] Figure 5 is a motion equation loss function diagram of a displacement prediction method of a glass curtain wall according to an embodiment of the present application;

[0062] Figure 6 is a total loss function diagram of a displacement prediction method of a glass curtain wall according to an embodiment of the present application;

[0063] Figure 7 is a structural block diagram of a displacement prediction system of a glass curtain wall according to an embodiment of the present application;

[0064] Figure 8 is a hardware structure diagram of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0065] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0066] In the current field of high-rise building design and construction, glass curtain walls are favored due to their aesthetic appearance, light transmission and light weight. The displacement and deformation of glass curtain walls under wind load may pose a potential threat to building safety. Existing displacement prediction methods mainly rely on two technical means: empirical formula and finite element analysis. However, each of them has certain drawbacks.

[0067] The empirical formula method is based on historical data and expert knowledge, using a simplified mathematical model to predict displacement. Although this method is simple to operate and fast to calculate, its prediction accuracy is limited by the quality and quantity of empirical data, making it difficult to adapt to complex actual engineering conditions. In addition, due to insufficient consideration of the variability of wind load and the complexity of curtain wall structure, the prediction accuracy of the empirical formula method has obvious defects.

[0068] The finite element analysis method, on the other hand, establishes a detailed numerical model to simulate the physical behavior of the curtain wall structure to obtain more accurate displacement prediction. Although this method can more realistically reflect the structural characteristics, it is complex to calculate and requires a large amount of computing resources and time, making the efficiency and practicality of finite element analysis not high.

[0069] In summary, the existing technologies mainly have the following drawbacks:

[0070] (1) Adaptability is insufficient, and existing methods are difficult to adapt to the diversity and complexity of curtain wall structures. For new materials or innovative design curtain wall systems, the prediction model may not accurately reflect its actual performance.

[0071] (2) Low computational efficiency, finite element analysis in the face of engineering practice that needs fast response, the calculation efficiency is often difficult to meet the needs of real-time or near real-time analysis.

[0072] (3) Limited generalization ability, the prediction results of empirical formula method are usually limited to specific application scenarios, lack of sufficient generalization ability, difficult to promote to more extensive engineering applications.

[0073] (4) Strong data dependence, existing methods largely depend on the quality and integrity of historical data, and data deficiencies or biases will directly affect the accuracy of the prediction results.

[0074] (5) Lack of real-time updating mechanism, as the building usage time and environmental conditions change, existing methods fail to achieve real-time updating and self-optimization of the prediction model.

[0075] The embodiment of the present application provides a displacement prediction method for glass curtain wall, which is applied to the design and construction scene of glass curtain wall of high-rise building, for example, in the scene of real-time monitoring of stress and displacement change of curtain wall structure. By integrating the element stiffness matrix and the element mass matrix in the local coordinate system, and combining the known boundary displacement response of the curtain wall, the displacement of the specific point inside the glass curtain wall can be accurately predicted, which not only improves the prediction accuracy and efficiency, but also enhances the generalization ability and adaptability of the model by embedding the physical law, and has the ability of real-time monitoring and self-optimization, effectively solving the limitations of traditional displacement prediction methods in calculation complexity, data dependence and real-time updating.

[0076] According to the embodiment of the present application, a displacement prediction method for glass curtain wall is provided. It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from here.

[0077] In this embodiment, a displacement prediction method for glass curtain wall is provided, which can be used in the above-mentioned computer, Figure 1 The flowchart of the displacement prediction method for glass curtain wall according to the embodiment of the present application is shown in Figure 1 The flowchart includes the following steps:

[0078] Step S101, calculating the overall stiffness matrix and overall mass matrix of the glass curtain wall based on the principle of structural dynamics.

[0079] It should be noted that the structural dynamics principle refers to the response law of the structure under the action of external force, including displacement, velocity, acceleration, etc., the stiffness matrix refers to the displacement deformation of the structure under the action of unit external force, and the mass matrix refers to the mass distribution of the structure.

[0080] Specifically, step S101 is implemented by a finite element method (FEM), the glass curtain wall is discretized and divided into a plurality of finite element units, the mass matrix and the stiffness matrix of each unit are calculated, and the overall mass matrix and the stiffness matrix are further calculated for subsequent construction of the motion equation.

[0081] In step S102, boundary displacement response information and boundary acceleration response information of the glass curtain wall under the action of external force are collected.

[0082] It should be noted that the boundary displacement response information refers to the displacement change of the boundary position of the glass curtain wall under the action of external force, and the boundary acceleration response information refers to the acceleration change of the boundary position of the glass curtain wall under the action of external force.

[0083] In actual operation, the displacement sensor and the acceleration sensor installed at the key positions of the glass curtain wall are used to monitor the response information in real time. Essentially, the boundary displacement response information and the boundary acceleration response information represent the actual dynamic response of the glass curtain wall under the action of external force, and provide a basis for constructing an accurate physical model.

[0084] In step S103, the motion equation of the glass curtain wall is constructed based on the overall stiffness matrix, the overall mass matrix, the boundary displacement response information, and the boundary acceleration response information.

[0085] It should be noted that the motion equation describes the dynamic behavior of the glass curtain wall under the action of external force, and the second-order partial differential equation form is adopted in this embodiment.

[0086] Specifically, according to the finite element method and the dynamics theory, the overall stiffness matrix and the mass matrix calculated in step S101, and the boundary response information collected in step S102 are used to construct the motion equation of the glass curtain wall.

[0087] In step S104, the current sampling time and the sampling point position information of the glass curtain wall are taken as real-time sampling point parameters, and are input into a preset fully connected neural network model to obtain a current displacement prediction result. The fully connected neural network model is iteratively trained according to the residual information between the motion equation and the current displacement prediction result, and the glass curtain wall real-time displacement prediction model is obtained when the iterative termination condition is reached.

[0088] It should be noted that the fully connected neural network model refers to an artificial neural network architecture, and each layer of neurons is connected to each neuron of the next layer, and the sampling point parameters include the current time and the position of the sampling point.

[0089] Specifically, the sampling time of the glass curtain wall at the current time and the position of the sampling point are input into the preset fully connected neural network model, and in this embodiment, the FEM-PINN network architecture is adopted.

[0090] It can be understood that, compared with the traditional finite element method, the FEM-PINN reduces the demand for computing resources by learning data-driven patterns, greatly improves the solving speed, and is especially suitable for large-scale or multi-parameter problems.

[0091] In step S105, the next sampling time and the sampling point position information of the glass curtain wall are input into the real-time displacement prediction model to obtain a real-time displacement prediction result.

[0092] Specifically, the trained FEM-PINN model is used to input the sampling point information of the next time into the model, and the model will predict the displacement of the glass curtain wall at this time according to the prediction rules and motion equations learned by the model.

[0093] The embodiment of the application calculates the overall stiffness matrix and mass matrix of the glass curtain wall based on the principle of structural dynamics, and constructs the motion equation combined with the boundary displacement response information and the acceleration response information.

[0094] A displacement prediction method of a glass curtain wall is provided in the present embodiment, which can be used in the above computer, Figure 2 is a flow chart of the displacement prediction method of the glass curtain wall according to an embodiment of the present application, as shown in the figure, the flow comprises the following steps: Figure 2

[0095] Step S201, calculating the overall stiffness matrix and the overall mass matrix of the glass curtain wall based on the principle of structural dynamics.

[0096] Specifically, the above step S201 comprises:

[0097] Step S2011, obtaining the physical property parameters and the geometric structure parameters of the glass curtain wall.

[0098] Step S2012, dividing the structure of the glass curtain wall into multiple planar units, wherein each planar unit has a local coordinate system.

[0099] Step S2013, calculating the unit stiffness matrix and the unit mass matrix in the local coordinate system according to the physical property parameters and the geometric structure parameters corresponding to each planar unit.

[0100] Step S2014, composing the overall stiffness matrix and the overall mass matrix of the glass curtain wall by the unit stiffness matrix and the unit mass matrix of each planar unit.

[0101] It should be noted that the physical property parameters refer to the material properties of the glass curtain wall, such as density, elastic modulus, Poisson's ratio, etc., which represent the response characteristics of the material when subjected to external force. The geometric structure parameters refer to the shape and size of the glass curtain wall, including the thickness of the panel, the boundary condition, the span, etc. The planar unit refers to the basic unit in finite element analysis, which can be referred to Figure 3 , Figure 3 The entire glass curtain wall is divided into 16 planar units. The local coordinate system refers to the independent coordinate system set with respect to each planar unit, which is used to determine the stiffness and mass matrix of each unit in its own coordinate system.

[0102] ​Specifically, the physical characteristic parameters of the glass curtain wall are obtained by consulting a material manual or performing experimental tests, and the geometric structure parameters are obtained by using design drawings or measuring tools. Through finite element analysis, the glass curtain wall structure is divided into a plurality of finite element units, and when the division is performed, a suitable unit type is selected and the density of the grid is reasonably set to ensure the calculation accuracy. The stiffness and mass matrices of each unit are first calculated in its local coordinate system. Based on the finite element theory, the expressions of the stiffness matrix and the unit mass matrix are derived based on the physical characteristic parameters and the geometric parameters. The stiffness matrix is calculated by combining the geometric parameters through the structural mechanics formula, and the mass matrix of the unit is calculated according to the mass density distribution through the mass integral formula. Finally, the stiffness matrix and the mass matrix of all units are superimposed by recombination to obtain the overall stiffness matrix and the mass matrix of the entire glass curtain wall.

[0103] For example, the unit stiffness matrix in the local coordinate system is calculated as follows:

[0104]

[0105] For example, the unit mass matrix in the local coordinate system is calculated as follows:

[0106]

[0107] The unit stiffness matrix and the unit mass matrix in the local coordinate system need to be converted into the expression in the overall coordinate system, and the conversion method is realized through the following formula:

[0108]

[0109] The conversion matrix is as follows:

[0110]

[0111] Finally, after the coordinate conversion, the stiffness matrix and the mass matrix of each unit of the curtain wall in the overall coordinate system are obtained, and the stiffness matrix K and the mass matrix M of the curtain wall are obtained by assembling.

[0112] In the embodiment of the application, the curtain wall structure is divided into a plurality of plane units, the stiffness matrix and the mass matrix of each plane unit are calculated in the local coordinate system, and then the matrices of the units are combined into the overall stiffness matrix and the mass matrix. The overall stiffness matrix and the mass matrix calculated based on the structural dynamics principle can accurately reflect the characteristics of the glass curtain wall structure under the action of external force, so as to construct a more accurate motion equation, further improve the prediction ability of the subsequent training of the glass curtain wall model, and ensure that the prediction result is more accurate.

[0113] ​​In step S202, boundary displacement response information and boundary acceleration response information of the glass curtain wall under the condition of external force are collected. For details, please refer to Figure 1 In step S102 of the embodiment shown, no further elaboration is given here.

[0114] In step S203, a motion equation of the glass curtain wall is constructed based on the overall stiffness matrix, the overall mass matrix, the boundary displacement response information and the boundary acceleration response information. For details, please refer to Figure 1 In step S103 of the embodiment shown, no further elaboration is given here.

[0115] In an alternative embodiment, the motion equation of the glass curtain wall is represented by the following formula:

[0116]

[0117] Wherein, M represents the overall mass matrix, K represents the overall stiffness matrix, represents the boundary acceleration response information, and v(t) represents the boundary displacement response information.

[0118] In the embodiment, the mass matrix (M) and the stiffness matrix (K) can completely describe the physical characteristics of the entire glass curtain wall, and the boundary conditions in the actual working condition are introduced into the motion equation by using the boundary acceleration response and the boundary displacement response information, so that the final motion equation can more accurately simulate the real boundary conditions and improve the accuracy of the motion equation.

[0119] In step S204, the current sampling time and the sampling point position information of the glass curtain wall are taken as real-time sampling point parameters, and are input into a preset fully connected neural network model to obtain a current displacement prediction result. The fully connected neural network model is iteratively trained according to residual information between the motion equation and the current displacement prediction result, and a real-time displacement prediction model of the glass curtain wall is obtained when an iteration termination condition is reached. For details, please refer to Figure 1 In step S104 of the embodiment shown, no further elaboration is given here.

[0120] In step S205, the next sampling time and the sampling point position information of the glass curtain wall are input into the real-time displacement prediction model to obtain a real-time displacement prediction result. For details, please refer to Figure 1 In step S105 of the embodiment shown, no further elaboration is given here.

[0121] In the embodiment, a displacement prediction method for a glass curtain wall is provided, which can be used in the above-mentioned computer. The flow includes the following steps:

[0122] In step S401, the overall stiffness matrix and the overall mass matrix of the glass curtain wall are calculated based on the principle of structural dynamics. For details, please refer to Figure 1Step S101 of the illustrated embodiment will not be described here again.

[0123] Step S402, the boundary displacement response information and the boundary acceleration response information of the glass curtain wall under the condition of external force are collected. For details, please refer to Figure 1 Step S102 of the illustrated embodiment will not be described here again.

[0124] Step S403, the motion equation of the glass curtain wall is constructed based on the overall stiffness matrix, the overall mass matrix, the boundary displacement response information and the boundary acceleration response information. For details, please refer to Figure 1 Step S103 of the illustrated embodiment will not be described here again.

[0125] Step S404, the current sampling time and the sampling point position information of the glass curtain wall are taken as real-time sampling point parameters and input into a preset fully connected neural network model to obtain a current displacement prediction result. The fully connected neural network model is iteratively trained according to the residual information between the motion equation and the current displacement prediction result. When an iteration termination condition is reached, a real-time displacement prediction model of the glass curtain wall is obtained.

[0126] Specifically, as Figure 4 As shown above, step S404 includes:

[0127] Step S4041, a fully connected neural network is constructed, and corresponding each neural network parameter is set as an initial value; wherein the neural network parameters include sampling point parameters, the number of different network layers, network width, the parameters of an activation function formula, the parameters of a loss function formula, and a network learning rate; the input data of the fully connected neural network includes sampling point parameters, and the output data of the fully connected neural network includes a displacement response corresponding to the sampling point inside the curtain wall.

[0128] Step S4042, in the iterative training process of the current round, the first sampling point parameter is input into the fully connected neural network to obtain a predicted displacement response corresponding to the first sampling point.

[0129] Step S4043, a first partial derivative and a second partial derivative are obtained by using the automatic differentiation principle to perform partial derivative analysis according to the predicted displacement response, the first partial derivative represents a second-order partial derivative of the displacement response with respect to time, and the second partial derivative represents a second-order partial derivative of the displacement response with respect to space.

[0130] Step S4044, residual information of the motion equation is calculated based on the first partial derivative and the second partial derivative.

[0131] Step S4045, a probability density function is constructed according to the residual information.

[0132] Step S4046, based on the probability density function, the second sampling point parameter is obtained by sampling, and the slope optimization parameter in the activation function formula is updated by using the gradient descent algorithm.

[0133] Step S4047, the second sampling point parameter is updated as the first sampling point parameter, and the next round of iterative training is continued; wherein the predicted displacement response in each round is also used to calculate the target loss function value of each round by bringing it into the loss function formula.

[0134] Step S4048, when the target loss function value is lower than the preset loss function value, or the number of iterations reaches the preset number of iterations, the iterative training is stopped, and the glass curtain wall real-time displacement prediction model is obtained.

[0135] For example, to implement the above steps S4041 to S4048, first, a fully connected neural network (FEM-PINN) needs to be constructed, and initial parameters are set for it. The input of the network is the position information (x, y) and time information t of the sampling points, and the output is the displacement response of the sampling points inside the curtain wall. Pytorch is used to implement the writing of FEM-PINN network code, and a network architecture of "1+1+8+1" is constructed, i.e. 1 input layer, 1 feature expansion layer, 8 hidden layers and 1 output layer, and each hidden layer has 50 neurons. The random initialization method is used to assign initial values to the network weights and biases. During initialization, the dimension of the input layer is 3 (corresponding to x, y, t), and the dimension of the output layer is 1 (corresponding to displacement response). The input data is linearly transformed through each layer and a nonlinear activation function to obtain the predicted value of the output layer. The final output displacement value is the prediction result of a certain sampling point at a certain time. The automatic differentiation technique is used to calculate the second-order partial derivative of the displacement response with respect to time and space. The automatic differentiation technique can calculate the derivative of the neural network output with respect to the input through the chain rule, calculate the second-order partial derivative of the displacement with respect to time (acceleration) and the second-order partial derivative of the displacement with respect to the spatial coordinates (stress), and calculate the residual based on the above calculated partial derivatives. According to the residual information, a probability density function is constructed to convert the residual into a probability distribution, which is used to guide the next sampling. The gradient descent algorithm is used to update the slope in the activation function, and the new sampling point is used as the input to repeat the previous steps. After each iteration, the loss function is calculated, and the fitting effect of the model is judged according to the value.

[0136] In a specific embodiment, the slope optimization parameter in the activation function formula is updated by using the gradient descent algorithm in the above step S4046, specifically including: updating the slope optimization parameter of the next round according to the slope optimization parameter of the current round and using the gradient descent algorithm; the gradient descent algorithm is represented by the following formula:

[0137] a k+1 = ak -η a J K (a);

[0138] wherein, η represents a network learning rate, a k represents a slope optimization parameter of the Kth round of iteration, a k+1 represents a slope optimization parameter of the K+1th round of iteration, η a J K (a) represents a gradient of the loss function along the a direction.

[0139] It can be understood that a is found by minimizing the loss function through the gradient descent method. For example, the slope transformation of the adaptive Sigmoid function under different a is as follows:

[0140]

[0141] The embodiment of the present application models through the fully connected neural network, combines the real-time data of the sampling points for real-time prediction, uses automatic differentiation for partial derivative analysis, combines the analysis result of the partial derivative with the motion equation, constructs a probability density function, and adaptively updates the sampling point parameters based on the residual information in each iteration training, focuses on sampling the area where the overall sampling prediction result of the possible glass curtain wall has a large error, improves the training efficiency and accuracy of the overall model, and in each round of iteration, the activation function and network parameters of the neural network are continuously optimized through the gradient descent algorithm, the loss value of each round is calculated to continuously update the training process, and the residual analysis of the motion equation is used to guide the training of the model, which not only makes the displacement result predicted by the neural network conform to the motion physical law of the glass curtain wall, but also obtains a higher interpretability of the real-time displacement prediction model of the glass curtain wall.

[0142] In an optional embodiment, the displacement prediction method for a glass curtain wall provided in the embodiment further comprises: performing feature expansion on the sampling point parameters based on a feature expansion layer in the fully connected neural network to obtain feature expansion parameters of the current round; calculating gradient information of the target loss function value with respect to the feature expansion parameters by using a back propagation algorithm; and updating weight parameters and bias parameters in an activation function formula in the next round according to the gradient information.

[0143] It should be noted that in the fully connected neural network, the input parameters of the sampling points are first processed by the feature expansion layer. The input spatial response variable is directly expanded into a high-dimensional form, for example, ln sin(λn x), where λn and ln are parameters that can be automatically updated during the network training process. The feature expansion method increases the dimension of the input data, so that the network can capture more complex spatial features. During the training process, the gradient information of the target loss function with respect to the feature expansion parameters (including parameters such as λn and ln) is calculated using the back propagation algorithm. According to the gradient information calculated in the previous step, the weight parameters and bias parameters in the activation function formula in the next round of training are updated, the network weights are optimized, the prediction error is minimized, and the prediction performance of the model is improved.

[0144] It can be understood that the values of λn and ln are automatically adjusted by the gradient descent algorithm to better fit the training data and capture complex patterns in the input data. The introduction of the parameter λn provides flexibility for the network, allowing the network to learn the spatial frequency characteristics in the data, and the ln parameter further enhances the sensitivity of the model to data changes.

[0145] The embodiment of the present application expands the dimension of the input data after the input layer through feature expansion, increases the high-dimensional representation of the data, and the back propagation algorithm can further optimize the weight parameters and bias of the model, ensuring that all parameters of the model work in coordination, thereby improving the prediction accuracy of the trained model.

[0146] In an optional implementation, the residual information includes a residual loss term and an integral of all residual loss terms, and the probability density function is represented by the following formula:

[0147]

[0148] Where p(x) represents the probability density function of the residual loss term, A represents the integral of all residual loss terms, ε(x) represents the residual loss term, T represents the residual sampling region, the residual sampling region represents the sampling point domain where the residual is higher than the preset standard value, k represents the sampling concentration of the sampling point in the residual sampling region, and c represents the sampling uniformity of the sampling point in the residual sampling region.

[0149] It should be noted that the residual information includes a residual loss term and an integral of all residual loss terms. The residual information is calculated based on the first derivative and the second derivative, and the residual sampling region T is a specific region representing the sampling point domain where the residual is higher than the preset standard value. In this region, the position of the sampling point is determined by the probability density function to ensure that the region with larger residual has more sampling points. Two important hyperparameters are introduced in the sampling process to adjust the sampling density and uniformity.

[0150] wherein k value represents the concentration degree of the sampling point in the residual sampling area, and a larger k value means that more intensive sampling will be performed in the area with larger residual; c value represents the uniformity of the sampling point, and a larger c value means that the sampling points will be more evenly distributed in the sampling area.

[0151] It can be understood that with the training of the model, the probability density function will be updated in each iteration, thereby continuously optimizing the distribution of the sampling points and making the model training more efficient. The sampling points of the residual are used to train the FEM-PINN model. Through this residual-based sampling method, i.e., the RAD method, the network weights can be optimized, the residual loss can be minimized, and finally the prediction accuracy of the model can be improved. The combination of the RAD method and the FEM-PINN improves the efficiency of data sampling points and ensures the physical consistency of the model output, so that the network can not only converge faster, but also better comply with the physical law.

[0152] The embodiment of the application determines the probability density function through the integral of all residual loss terms, residual loss terms, sampling concentration, sampling uniformity and residual sampling area. It should be noted that the residual information refers to the residual information calculated based on the first derivative and the second derivative and the motion equation. Since the motion equation of the glass curtain wall is combined, the constructed probability density function will be continuously updated in the iterative training, so that the prediction result of the trained model is more accurate.

[0153] Step S405, input the next sampling time of the glass curtain wall and the sampling point position information into the real-time displacement prediction model to obtain the real-time displacement prediction result. For details, please refer to Figure 1 The step S105 of the embodiment shown will not be described here.

[0154] In an alternative embodiment, the loss function formula is:

[0155] L(t)=w cz loss IC +w0loss PDE +w b loss BC ;

[0156]

[0157] wherein w cz represents the initial value loss weight coefficient, w b represents the boundary loss weight coefficient, w o represents the motion equation loss weight coefficient, loss IC represents the network initial value loss, loss BC represents the network boundary loss, loss PDErepresents the motion equation loss, L(t) represents the loss function of the whole network, N b represents the boundary loss function sampling region, N s represents the initial value loss function sampling region, N T represents the motion equation loss function sampling region, t represents the sampling point, r represents the whole sampling region, T represents the residual sampling region, represents the displacement response predicted by the network at the initial sampling point parameter x c ,y c ,t c inside the curtain wall corresponding to the sampling point, u c represents the actual displacement response at the initial sampling point, represents the displacement response predicted by the network at any sampling point parameter x b ,y b ,t b inside the curtain wall corresponding to the sampling point, u b represents the actual displacement response at any sampling point during the network training process, f(t T ) represents the residual value of the motion equation for the network prediction.

[0158] In the actual training process, the network architecture of the FEM-PINN algorithm for solving PDE (motion equation loss) considering data characteristics is as follows: the input of the network is the coordinate information sampled from the residual region T, that is, (x, y, t), where x and y are spatial coordinates and t is a time coordinate. The network processes the input x and y through a feature expansion layer to expand them into a feature form suitable for the network. The number of network layers is L, the number of neurons in the hidden layer is m, the activation function is selected as the Tanh function, and the learning rate is a. The output of the network is the approximate prediction value of the displacement response of some specific points inside the curtain wall.

[0159] The steps of the FEM-PINN algorithm for solving PDE considering data characteristics are as follows: Step 1: build a fully connected neural network and initialize the parameters. Step 2: define the loss function, which considers the residual of the partial differential equation, the initial condition and the boundary condition. Step 3: start training the model. Step 4: perform a certain number of iterations (K increases by 1 each time) until the training requirements of the network are met.

[0160] After training by the above algorithm, the motion equation loss function of FEM-PINN is as shown in Figure 5 , and the total loss function of the training process is as shown in Figure 6 . With the decrease of the number of training rounds, the loss training threshold decreases by about 10 6 times and then becomes flat. At this time, both the motion equation and the overall loss function have decreased below a certain threshold, and the model iteration training is completed.

[0161] In the embodiment of the present application, the loss function contains information of initial value loss, boundary loss and motion equation loss, wherein the initial value loss and the boundary loss correspond to displacement responses of the curtain wall under specific conditions, and the motion equation loss corresponds to matching degrees between network prediction and physical equations of the glass curtain wall. Different weight coefficients are introduced in the loss function, and the three loss functions are balanced, so that the model prediction result is more in line with the physical motion phenomenon of the actual glass curtain wall, and the accuracy of the final trained model is improved.

[0162] In the embodiment, a displacement prediction system for a glass curtain wall is also provided, which is used to implement the above-mentioned embodiments and preferred embodiments, and will not be described again. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware or a combination of software and hardware is also possible and is contemplated.

[0163] The embodiment provides a displacement prediction system for a glass curtain wall, as shown in Figure 7 The system comprises:

[0164] A general matrix calculation module 701 is configured to calculate a general stiffness matrix and a general mass matrix of the glass curtain wall based on structural dynamics principles.

[0165] A boundary response information acquisition module 702 is configured to acquire boundary displacement response information and boundary acceleration response information of the glass curtain wall under the condition of external force application.

[0166] A motion equation construction module 703 is configured to construct a motion equation of the glass curtain wall based on the general stiffness matrix, the general mass matrix, the boundary displacement response information and the boundary acceleration response information.

[0167] An iterative training module 704 is configured to input current sampling time and sampling point position information of the glass curtain wall as real-time sampling point parameters into a pre-set fully connected neural network model to obtain a current displacement prediction result, and iteratively train the fully connected neural network model according to residual information between the motion equation and the current displacement prediction result, so as to obtain a real-time displacement prediction model of the glass curtain wall when an iterative termination condition is reached.

[0168] A real-time displacement prediction module 705 is configured to input next sampling time and sampling point position information of the glass curtain wall into the real-time displacement prediction model to obtain a real-time displacement prediction result.

[0169] In some optional embodiments, the general matrix calculation module 701 comprises:

[0170] a physical property parameter unit configured to obtain physical property parameters and geometric structure parameters of the glass curtain wall;

[0171] a planar unit division unit configured to divide the structure of the glass curtain wall into a plurality of planar units, each of which has a local coordinate system;

[0172] a local unit stiffness matrix calculation unit configured to calculate a unit stiffness matrix and a unit mass matrix in the local coordinate system according to the physical property parameters and the geometric structure parameters corresponding to each planar unit;

[0173] a total stiffness and mass matrix assembly unit configured to assemble the unit stiffness matrix and the unit mass matrix of each planar unit into a total stiffness matrix and a total mass matrix of the glass curtain wall.

[0174] In some optional embodiments, the motion equation of the glass curtain wall is represented by the following formula:

[0175]

[0176] wherein M represents the total mass matrix, K represents the total stiffness matrix, represents boundary acceleration response information, and v(t) represents boundary displacement response information.

[0177] In some optional embodiments, the iterative training module 704 specifically includes:

[0178] a neural network construction unit configured to construct a fully connected neural network and set corresponding various neural network parameters as initial values; wherein the neural network parameters include sampling point parameters, the number of different network layers, network width, the parameters of an activation function formula, the parameters of a loss function formula, and a network learning rate; the input data of the fully connected neural network includes the sampling point parameters, and the output data of the fully connected neural network includes a displacement response corresponding to a sampling point inside the curtain wall;

[0179] a predicted displacement response unit configured to input first sampling point parameters into the fully connected neural network to obtain a predicted displacement response corresponding to the first sampling point during the iterative training process of the current round;

[0180] a partial derivative calculation unit configured to perform partial derivative analysis according to the predicted displacement response by using an automatic differentiation principle to obtain a first partial derivative and a second partial derivative, the first partial derivative representing a second-order partial derivative of the displacement response with respect to time, and the second partial derivative representing a second-order partial derivative of the displacement response with respect to space;

[0181] a residual calculation unit configured to calculate residual information of the motion equation based on the first partial derivative and the second partial derivative;

[0182] a probability density function construction unit configured to construct a probability density function according to the residual information;

[0183] a sampling parameter updating unit configured to continue sampling to obtain second sampling point parameters and update a slope optimization parameter in an activation function formula by using a gradient descent algorithm;

[0184] an iterative training unit configured to update the second sampling point parameters to the first sampling point parameters and continue the next round of iterative training; wherein the predicted displacement response in each round is further used to calculate a target loss function value of each round by using the loss function formula;

[0185] a loss function calculation unit configured to stop the iterative training when the target loss function value is lower than a preset loss function value or the number of rounds of the iterative training reaches a preset number of iterations, and obtain a glass curtain wall real-time displacement prediction model.

[0186] In some optional embodiments, the device further includes:

[0187] a feature expansion module configured to perform feature expansion on the sampling point parameters based on a feature expansion layer in the fully connected neural network to obtain feature expansion parameters of the current round;

[0188] a gradient calculation module configured to calculate gradient information of the target loss function value with respect to the feature expansion parameters by using a back propagation algorithm;

[0189] a parameter updating module configured to update a weight parameter and a bias parameter in an activation function formula in the next round according to the gradient information.

[0190] In some optional embodiments, the residual information includes a residual loss term and an integral of all residual loss terms, and the probability density function is represented by the following formula:

[0191]

[0192] wherein p(x) represents a probability density function of the residual loss term, A represents an integral of all residual loss terms, ε(x) represents the residual loss term, T represents a residual sampling region, the residual sampling region represents a sampling point domain in which the residual is higher than a preset standard value, k represents a sampling concentration of the sampling point in the residual sampling region, and c represents a sampling uniformity of the sampling point in the residual sampling region.

[0193] In some optional embodiments, the loss function formula is:

[0194] L(t) = w cz loss IC +w0loss PDE+w b loss BC ;

[0195]

[0196] wherein, w cz represents the initial value loss weight coefficient, w b represents the boundary loss weight coefficient, w o represents the motion equation loss weight coefficient, loss IC represents the network initial value loss, loss BC represents the network boundary loss, loss PDE represents the motion equation loss, L(t) represents the loss function of the whole network, N b represents the boundary loss function sampling region, N s represents the initial value loss function sampling region, N T represents the motion equation loss function sampling region, t represents the sampling point, R represents the whole sampling region, and T represents the residual sampling region, represents the displacement response corresponding to the sampling point inside the curtain wall predicted by the network when the initial sampling point parameters x c , y c , and t c are taken, u c represents the actual displacement response of the initial sampling point, represents the displacement response corresponding to the sampling point inside the curtain wall predicted by the network when any sampling point parameters x b , y b , and t b are taken during the network training process, u b represents the actual displacement response of any sampling point during the network training process, and f(t T ) represents the residual value of the motion equation to the network prediction.

[0197] The further function description of each module and unit is the same as that of the above-mentioned corresponding embodiment, and will not be repeated here.

[0198] The displacement prediction system of the glass curtain wall in the embodiment is presented in the form of a functional unit, wherein the unit refers to an ASIC (Application Specific Integrated Circuit, Application Specific Integrated Circuit) circuit, a processor and a memory executing one or more software or fixed programs, and / or other devices that can provide the above-mentioned functions.

[0199] The embodiment of the application further provides a computer device with the above-mentioned Figure 7 displacement prediction system of the glass curtain wall.

[0200] Please refer to Figure 8, Figure 8 is a structural schematic diagram of a computer device provided by an optional embodiment of the present application. As shown in Figure 8 the computer device includes one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components communicate with each other using different buses, and can be mounted on a common mainboard or mounted in other manners as needed. The processor can process instructions executed within the computer device, including instructions stored in the memory or on the memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, multiple processors and / or multiple buses can be used with multiple memories and multiple memory, if needed. Also, multiple computer devices can be connected, each providing part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 8 The processor 10 is taken as an example in the

[0201] The processor 10 can be a central processor, a network processor, or a combination thereof. The processor 10 can further include a hardware chip. The hardware chip can be an application specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device can be a complex programmable logic device, a field programmable logic device, a general array logic, or any combination thereof.

[0202] The memory 20 stores instructions executable by the at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiments.

[0203] The memory 20 can include a program storage area and a data storage area. The program storage area can store an operating system and application programs required by at least one function; the data storage area can store data created according to the use of the computer device, etc. In addition, the memory 20 can include a high-speed random access memory, and can also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some optional embodiments, the memory 20 can optionally include a memory remotely arranged with respect to the processor 10, and these remote memories can be connected to the computer device through a network. Examples of the network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0204] The memory 20 can include a volatile memory, such as a random access memory; the memory can also include a non-volatile memory, such as a flash memory, a hard disk, or a solid state disk; the memory 20 can also include a combination of the above kinds of memories.

[0205] The computer device also comprises a communication interface 30 for communication of the computer device with other devices or communication networks.

[0206] The embodiments of the present application also provide a computer readable storage medium, the method according to the embodiments of the present application can be implemented in hardware, firmware, or recorded in a storage medium, or stored in a remote storage medium or a non-transitory machine readable storage medium and downloaded from a network and stored in a local storage medium, so that the method described herein can be processed by such software on a storage medium using a general purpose computer, a special purpose processor or programmable or special purpose hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk or a solid state disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that the computer, processor, microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code, when the software or computer code is accessed and executed by the computer, processor or hardware, the method shown in the above embodiments is implemented.

[0207] Part of the present application can be applied as a computer program product, for example, computer program instructions, when executed by a computer, through the operation of the computer, the method and / or technical solutions according to the present application can be called or provided. Those skilled in the art should understand that the form of computer program instructions in computer readable medium includes but is not limited to source file, executable file, installation package file, etc., accordingly, the way of computer program instructions executed by computer includes but is not limited to: the computer directly executes the instructions, or the computer compiles the instructions and then executes the corresponding compiled program, or the computer reads and executes the instructions, or the computer reads and installs the instructions and then executes the corresponding installed program. Here, the computer readable medium can be any available computer readable storage medium or communication medium accessible to the computer.

[0208] Although the embodiments of the present application are described in conjunction with the accompanying drawings, various modifications and changes can be made by those skilled in the art without departing from the spirit and scope of the present application, and such modifications and changes fall within the defined scope.

Claims

1. A displacement prediction method of a glass curtain wall, characterized by, The method comprises: calculating a total stiffness matrix and a total mass matrix of the glass curtain wall based on structural dynamics principles; collecting boundary displacement response information and boundary acceleration response information of the glass curtain wall under the action of external force; constructing a motion equation of the glass curtain wall based on the total stiffness matrix, the total mass matrix, the boundary displacement response information, and the boundary acceleration response information, the motion equation of the glass curtain wall being represented by the following formula: ； wherein, M denotes the global mass matrix, K denotes the global stiffness matrix, denotes the boundary acceleration response information, denotes the boundary displacement response information; inputting current sampling time and sampling point position information of the glass curtain wall as real-time sampling point parameters into a preset fully connected neural network model to obtain a current displacement prediction result, and iteratively training the fully connected neural network model according to residual information between the motion equation and the current displacement prediction result, and obtaining a real-time displacement prediction model of the glass curtain wall when an iterative termination condition is reached, specifically comprising: constructing a fully connected neural network, and setting corresponding neural network parameters as initial values; wherein the neural network parameters comprise sampling point parameters, the number of different network layers, network width, the parameters of an activation function formula, the parameters of a loss function formula, and a network learning rate; the input data of the fully connected neural network comprises sampling point parameters, and the output data of the fully connected neural network comprises displacement responses corresponding to sampling points inside the curtain wall; in the iterative training process of the current round, inputting first sampling point parameters into the fully connected neural network to obtain a predicted displacement response corresponding to the first sampling point; performing partial derivative analysis according to the predicted displacement response by using an automatic differentiation principle to obtain a first partial derivative and a second partial derivative, the first partial derivative representing a second-order partial derivative of the displacement response with respect to time, and the second partial derivative representing a second-order partial derivative of the displacement response with respect to space; calculating residual information of the motion equation based on the first partial derivative and the second partial derivative; constructing a probability density function according to the residual information; continuing sampling based on the probability density function to obtain second sampling point parameters, and simultaneously updating a slope optimization parameter in the activation function formula by using a gradient descent algorithm; updating the second sampling point parameters to the first sampling point parameters, and continuing the iterative training of the next round; wherein the predicted displacement response in each round is also used to calculate a target loss function value of each round by bringing it into a loss function formula; stopping the iterative training when the target loss function value is lower than a preset loss function value, or when the number of rounds of the iterative training reaches a preset iteration number, and obtaining the real-time displacement prediction model of the glass curtain wall; inputting next sampling time and sampling point position information of the glass curtain wall into the real-time displacement prediction model to obtain a real-time displacement prediction result.

2. The method of claim 1, wherein, The method comprises: obtaining physical property parameters and geometric structure parameters of the glass curtain wall; dividing the structure of the glass curtain wall into a plurality of planar units, wherein each planar unit has a local coordinate system; calculating an element stiffness matrix and an element mass matrix in the local coordinate system according to the physical property parameters and the geometric structure parameters corresponding to each planar unit; Assemble the element stiffness matrix and the element mass matrix of each planar element into the overall stiffness matrix and the overall mass matrix of the glass curtain wall.

3. The method of claim 1, wherein, The method further comprises: performing feature expansion on the sampling point parameters based on a feature expansion layer in the fully connected neural network to obtain feature expansion parameters of the current round; calculating gradient information of the target loss function value with respect to the feature expansion parameters using a back propagation algorithm; updating weight parameters and bias parameters in an activation function formula in the next round according to the gradient information.

4. The method of claim 1, wherein, The residual information includes a residual loss term and an integral of all residual loss terms, and the probability density function is represented by the following formula: ； ； wherein, a probability density function representing the residual loss term, an integral representing all residual loss terms, a residual loss term, T a residual sampling region, the residual sampling region representing a sampling point domain in which the residual is higher than a preset standard value; k a value representing a sampling concentration of the sampling point in the residual sampling region, c a value representing a sampling uniformity of the sampling point in the residual sampling region.

5. The method of claim 1, wherein, The loss function formula is: ； ； ； ； wherein, represents the initial value loss weight coefficient, represents the boundary loss weight coefficient, represents the motion equation loss weight coefficient, represents the network initial value loss, represents the network boundary loss, represents the motion equation loss, represents the loss function of the network as a whole, represents the sampling region of the boundary loss function, represents the sampling region of the initial value loss function, represents the sampling region of the motion equation loss function, t represents the sampling point, represents the overall sampling region, represents the residual sampling region, represents the displacement response corresponding to the sampling point inside the curtain wall predicted by the network at the initial sampling point parameter , represents the actual displacement response of the initial sampling point, represents the displacement response corresponding to the sampling point inside the curtain wall predicted by the network at any sampling point parameter during the network training process, represents the actual displacement response of any sampling point during the network training process, represents the residual value of the motion equation to the network prediction.

6. A displacement prediction system for a glass curtain wall, the system comprising: The system is applied to the displacement prediction method of the glass curtain wall in any one of claims 1-5, comprising: an overall matrix calculation module configured to calculate the overall stiffness matrix and the overall mass matrix of the glass curtain wall based on structural dynamics principles; a boundary response information acquisition module configured to acquire boundary displacement response information and boundary acceleration response information of the glass curtain wall under the condition of external force application; a motion equation construction module configured to construct a motion equation of the glass curtain wall based on the overall stiffness matrix, the overall mass matrix, the boundary displacement response information, and the boundary acceleration response information; an iterative training module configured to input current sampling time and sampling point position information of the glass curtain wall as real-time sampling point parameters into a preset fully connected neural network model to obtain a current displacement prediction result, and iteratively train the fully connected neural network model according to residual information between the motion equation and the current displacement prediction result, and obtain a real-time displacement prediction model of the glass curtain wall when an iterative termination condition is reached; a real-time displacement prediction module configured to input next sampling time and sampling point position information of the glass curtain wall into the real-time displacement prediction model to obtain a real-time displacement prediction result.

7. A computer device, characterized by comprise: a memory and a processor, which are communicatively connected, and the memory stores computer instructions, and the processor executes the computer instructions to perform the displacement prediction method of the glass curtain wall in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions for causing a computer to perform the displacement prediction method of the glass curtain wall in any one of claims 1-5.

Images