A method and system for analyzing special-shaped structures based on deep neural network

Abstract

The present invention discloses a method and system for analyzing special-shaped structures based on a deep neural network, and relates to the technical field of wind tunnel design and analysis. The method comprises: obtaining an initial wind pressure load of each load area based on a maximum airflow velocity and a corresponding equivalent wind pressure in the special-shaped structure, obtaining a compensation coefficient for compensating the initial wind pressure load using each equivalent wind pressure, and compensating the initial wind pressure load using the compensation coefficient; inputting the corrected wind pressure load into a temperature compensation model based on a deep neural network and trained for processing, obtaining a target wind pressure load after temperature compensation, obtaining the relationship characteristics between the peak frequency of pulsating pressure and the magnitude of pulsating pressure at wind speed in the special-shaped structure, and performing a steady-state analysis on the special-shaped structure based on the relationship characteristics and the target wind pressure load; providing a reference and a foundation for the construction of large-span and ultra-long structures, and providing a strong data support for the selection of construction equipment, connection methods, etc.

Description

Technical Field

[0001] The present invention relates to the technical field of wind tunnel design and analysis, and more specifically, to a method and system for analyzing special-shaped structures based on a deep neural network. Background Art

[0002] With the continuous improvement of requirements for architectural space and shape, complex analysis methods for large-span spatial structures are also constantly developing; however, large-scale analysis is extremely difficult, and design is even more difficult. A large wind tunnel is a very complex large-scale scientific and technological infrastructure, and its own design and construction require a lot of basic research work as support. At the same time, due to the difference in ambient temperature in different seasons, and the internal airflow temperature will also rise during operation, and it is much higher than the ambient temperature change, the cave structure needs to at least consider the effects of airflow alternating temperature, external ambient temperature, airflow internal pressure and earthquakes; for such a scientific test site, in addition to the dead load, live load, wind load, temperature effect and earthquake effect that general building structures usually bear, it will also face many special loads and effects; therefore, for the establishment of a large wind tunnel, it is first necessary to understand the loads it needs to bear.

[0003] From the perspective of structural design, the integration of large-span, ultra-long structures and high-precision special-shaped surfaces has greatly increased the complexity of the structure. Traditional structural design methods require a lot of manual operations when establishing mechanical models, with low calculation efficiency and poor accuracy. In the analysis involving a large number of software combinations and iterations, the defects are particularly prominent, and there are deviations caused by human factors, making it difficult to accurately consider the impact of various load conditions on such complex structures; the tunnel structure of large wind tunnels presents characteristics such as large spans, spaciousness, special shapes, and ultra-long lengths, which makes the structure bear wind loads, temperature loads, seismic loads, and internal airflow alternating temperature loads. When the load is complex, the refinement of the analysis process, seamless software connection, and zero error have become rigid needs. The existing mechanical models cannot fully and accurately simulate its mechanical behavior. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for analyzing special-shaped structures based on deep neural networks to solve the problems existing in the above-mentioned background technology.

[0005] The above technical objectives of the present invention are achieved through the following technical solutions:

[0006] In a first aspect, the present application provides a method for analyzing a special-shaped structure based on a deep neural network, comprising the following specific steps:

[0007] The shape structure parameters are obtained according to the preset three-dimensional model of the special-shaped structure, and the shape of the special-shaped structure is divided into regions based on the shape structure parameters to obtain multiple load regions affecting the special-shaped structure;

[0008] Based on the maximum airflow velocity and the corresponding equivalent wind pressure in the special-shaped structure, the initial wind pressure load of each load area is obtained, and the compensation coefficient for compensating the initial wind pressure load is obtained by using each equivalent wind pressure. The initial wind pressure load is compensated by using the compensation coefficient to obtain the corrected wind pressure load of each load area;

[0009] Based on the maximum operating temperature of each load area during use, the corrected wind pressure load is input into a temperature compensation model based on a deep neural network and trained for processing, and a target wind pressure load corresponding to the corrected wind pressure load of each load area after temperature compensation is obtained;

[0010] MIDAS GEN software was used to simulate the relationship between the peak frequency and pulsating pressure of the special-shaped structure and the wind speed, and the corresponding relationship characteristics were obtained. Based on the relationship characteristics and the target wind pressure load, the steady-state analysis of the special-shaped structure was carried out.

[0011] The beneficial effects of the present invention are as follows: in the present solution, firstly, the area division is carried out according to the specific external structural parameters of the special-shaped structure, and the special-shaped structure can be divided into a plurality of load areas. The area division can be carried out according to the change trend of the structure, that is, if the maximum bending or change angle of each load area obtained by division does not exceed the preset angle value, it can be considered that the wind pressure, vibration and other influences on the load area are consistent; secondly, the initial wind pressure load is calculated by the maximum airflow velocity of the special-shaped structure after it is put into use and the equivalent wind pressure corresponding to each load area. For example, if the special-shaped structure is a large wind tunnel with a large span and an ultra-long structure, the maximum airflow velocity of this large wind tunnel when it is put into use is used to calculate the initial wind pressure load; and each The compensation coefficient is obtained by replacing wind pressure to compensate each initial wind pressure load, and finally the compensated corrected wind pressure load is obtained; finally, since temperature will affect wind pressure, and the temperature in most special-shaped structures will change with the change of ambient temperature on the one hand, and the temperature of special-shaped structures during operation also has a large span, it is necessary to consider the influence of airflow alternating temperature, and then the trained temperature compensation model is used to compensate the corrected wind pressure load, so as to accurately obtain the wind pressure load of each load area considering the temperature influence, that is, the final target wind pressure load; of course, large-span and ultra-long structures also need to consider the influence of other multi-factor loads, such as earthquake resistance and vibration resistance, etc. These aspects can be inverted and simulated through MIDAS GEN software, and finally the target load of each load area can be obtained.

[0012] In this solution, by predicting and judging the loads required for large-span and ultra-long structures, the impact of various load conditions on such complex structures is accurately considered, thereby improving the accuracy and efficiency of structural analysis and design, reducing deviations caused by human factors, and saving a lot of time and resources. This provides a reference and preparation for the construction of special-shaped structures, and provides strong data support for the selection of equipment (such as damper parameters, etc.) and connection methods for the construction of special-shaped structures.

[0013] Based on the above technical solution, the present invention can also be improved as follows.

[0014] Furthermore, the above simulation of the relationship between the peak frequency of the pulsating pressure and the pulsating pressure in the special-shaped structure and the wind speed is as follows:

[0015] The wind pressure time history data in the fixed wind speed test is obtained based on the measured data of the small wind tunnel, and the wind pressure time history data is converted from the time domain to the frequency domain through Fourier transform to obtain the pulsating wind pressure frequency domain data corresponding to the wind pressure time history data;

[0016] The frequency domain data of pulsating wind pressure under different test wind speeds are summarized to obtain the relationship characteristics between the peak frequency and pulsating pressure in the special-shaped structure and the wind speed, and the steady-state analysis of the special-shaped structure is carried out based on the relationship characteristics and the target wind pressure load to obtain the response of the special-shaped structure under pulsating wind pressure.

[0017] Furthermore, the above initial wind pressure load is specifically:

[0018] ; In the formula, Indicates the load area i The initial wind pressure load is represents the average wind pressure in each load area, which is directly obtained through simulation experiments. g represents the peak factor. The correlation coefficient matrix is i List, A diagonal matrix representing the standard deviation of the equivalent wind pressure for each load area.

[0019] Furthermore, the above equivalent wind pressure represents the wind pressure time history obtained by dividing the algebraic sum of the elastic restoring forces of all nodes in each load area of ​​the special-shaped structure in a certain direction by the projection area of ​​the corresponding load area in this direction. The equivalent wind pressure of each load area is specifically:

[0020] ,in:

[0021] ; , ;

[0022] In the formula, Indicates the load area i The equivalent wind pressure, represents the modal coordinate function, represents the mass matrix, Represents the modal characteristic matrix of the heterogeneous structure, represents the modal eigenvalue matrix, represents the influence coefficient matrix, represents the elastic restoring force of node j on the load area i The contribution of the equivalent wind pressure is Indicates the load area i The projected area of They respectively represent the angles between the projection direction and the X, Y, and Z axes in the preset spatial rectangular coordinate system.

[0023] The beneficial effect of adopting the above further scheme is that the equivalent wind pressure is different from the actual wind pressure acting on the structure. It is a wind pressure derived from the wind vibration response, which includes the influence of background and resonance components. Under reasonable zoning, the regional equivalent wind pressure calculated by the elastic restoring force of the large-span roof structure can reflect the wind-induced response characteristics of this type of structure.

[0024] Furthermore, the above compensation coefficient is obtained by the following method:

[0025] The modal response of each load area is obtained based on each equivalent wind pressure, and the predicted displacement response of each load area is obtained based on the modal response;

[0026] Based on the predicted displacement response and the actual displacement response of each load area, the first coefficient and the second coefficient are obtained by fitting through the least square method, and the compensation coefficient includes the first coefficient and the second coefficient.

[0027] The beneficial effect of adopting the above further scheme is: in order to reduce the error caused by the zoned wind pressure, the first coefficient and the second coefficient obtained by the correlation between the displacement response and the equivalent wind pressure are used to compensate for the initial wind pressure load, thereby improving the calculation accuracy of the wind pressure in each load area.

[0028] Furthermore, the above predicted displacement response is specifically:

[0029] ,in, ;

[0030] In the formula, Indicates the load area i The predicted displacement response, The modal characteristic matrix of the special-shaped structure, Indicates the load area i The modal response of represents the average response of each mode, g represents the peak factor, represents the diagonal matrix of the modal response standard deviations, The first column represents the correlation coefficient matrix between the modal response and the equivalent wind pressure. i List.

[0031] Furthermore, the predicted displacement response and the actual displacement response based on each load area are fitted by the least square method to obtain the first coefficient and the second coefficient, which are specifically:

[0032] , ;

[0033] In the formula, denote the mean and pulsating parts of the predicted displacement response, respectively. They represent the mean and pulsating parts of the actual displacement response, Represent the first coefficient and the second coefficient respectively. The first coefficient and the second coefficient are determined when the value is the minimum; the corrected wind pressure load is specifically:

[0034] ; In the formula, Indicates the load area i Corrected wind pressure load, represents the average wind pressure in each load area, which is directly obtained through simulation experiments. g represents the peak factor. The correlation coefficient matrix is i List, Represents the diagonal matrix consisting of the standard deviation of the equivalent wind pressure in each load area, denote the first coefficient and the second coefficient respectively.

[0035] Furthermore, the above temperature compensation model is obtained by the following method:

[0036] Calculate the corrected wind pressure loads at different airflow speeds, and simulate the real wind pressure loads of the corrected wind pressure loads at different temperatures and airflow speeds based on the MIDAS GEN software, and construct training samples through each corrected wind pressure load and the corresponding real wind pressure load;

[0037] The preset initial model is trained using the training samples until the preset training end condition is reached, and the preset initial model that reaches the preset training end condition is determined as the temperature compensation model. The preset training end condition is that the loss function does not exceed the threshold, and the loss function is specifically:

[0038] ,in:

[0039] , ;

[0040] In the formula, represents the loss function, represents the first function, represents the second function, Indicates the training sample i The actual wind pressure load of the samples, Indicates the training sample i The predicted wind pressure load of samples, n represents the total number of samples in the training sample, Indicates the training sample i The absolute value of the difference between the true wind pressure load of a sample and the corresponding corrected wind pressure load, represent the weight of the first function and the weight of the second function respectively.

[0041] In a second aspect, the present application provides a special-shaped structure analysis system based on a deep neural network, which is applied to a special-shaped structure analysis method based on a deep neural network in any one of the first aspects, including:

[0042] A region division module is used to obtain the shape structure parameters according to the preset three-dimensional model of the special-shaped structure, divide the shape of the special-shaped structure into regions based on the shape structure parameters, and obtain multiple load regions that affect the special-shaped structure;

[0043] A load correction module is used to obtain the initial wind pressure load of each load area based on the maximum airflow velocity and the corresponding equivalent wind pressure in the special-shaped structure, obtain the compensation coefficient for compensating the initial wind pressure load using each equivalent wind pressure, and compensate the initial wind pressure load using the compensation coefficient to obtain the corrected wind pressure load of each load area;

[0044] The load compensation module is used to input the corrected wind pressure load into the temperature compensation model trained based on the deep neural network based on the maximum working temperature of each load area when in use, and obtain the target wind pressure load that has been temperature compensated and corresponds to the corrected wind pressure load of each load area;

[0045] The steady-state analysis module is used to simulate the relationship between the peak frequency of the pulsating pressure and the pulsating pressure in the special-shaped structure and the wind speed using the MIDAS GEN software, and obtain the corresponding relationship characteristics, and perform a steady-state analysis on the special-shaped structure based on the relationship characteristics and the target wind pressure load.

[0046] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any one of the methods in the first aspect when executing the computer program.

[0047] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions enable a computer to execute any one of the methods in the first aspect.

[0048] Compared with the prior art, the present invention has at least the following beneficial effects:

[0049] In the present application, firstly, the area division is carried out according to the specific external structural parameters of the special-shaped structure. The special-shaped structure can be divided into multiple load areas. The area division can be carried out according to the change trend of the structure, that is, if the maximum bending or change angle of each load area obtained by division does not exceed the preset angle value, it can be considered that the wind pressure, vibration and other influences on the load area are consistent; secondly, the initial wind pressure load is calculated by the maximum airflow velocity of the special-shaped structure after it is put into use and the equivalent wind pressure corresponding to each load area. For example, if the special-shaped structure is a large wind tunnel with a large span and an extra-long structure, the maximum airflow velocity of this large wind tunnel when it is put into use is used to calculate the initial wind pressure load; and the wind pressure values ​​obtained by each equivalent wind pressure are used to obtain the initial wind pressure load. The compensation coefficient is taken to compensate each initial wind pressure load, and finally the compensated corrected wind pressure load is obtained; finally, since temperature will affect wind pressure, and the temperature in most special-shaped structures will change with the change of ambient temperature on the one hand, and the temperature of special-shaped structures during operation also has a large span, it is necessary to consider the influence of airflow alternating temperature, and then the trained temperature compensation model is used to compensate the corrected wind pressure load, so as to accurately obtain the wind pressure load of each load area considering the temperature influence, that is, the final target wind pressure load; of course, large-span and ultra-long structures also need to consider the influence of other multi-factor loads, such as earthquake resistance and vibration resistance, these aspects can be inverted and simulated through MIDAS GEN software, and finally the target load of each load area can be obtained.

[0050] In the present application, by predicting and judging the loads required for large-span and ultra-long structures, the influence of various load conditions on such complex structures is accurately considered, thereby improving the accuracy and efficiency of structural analysis and design, reducing deviations caused by human factors, and saving a lot of time and resources. This provides a reference and a foundation for the construction of special-shaped structures, and provides strong data support for the selection of equipment (such as damper parameters, etc.) and connection methods for the construction of special-shaped structures. The equivalent wind pressure is different from the actual wind pressure acting on the structure. It is a wind pressure derived from the wind vibration response, which includes the influence of background and resonance components. Under reasonable zoning, the regional equivalent wind pressure calculated by the elastic restoring force of the large-span roof structure can reflect the wind-induced response characteristics of this type of structure. At the same time, in order to reduce the error caused by the partitioned wind pressure, the first coefficient and the second coefficient obtained by the correlation between the displacement response and the equivalent wind pressure are used to compensate for the initial wind pressure load, thereby improving the calculation accuracy of the wind pressure in each load area. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0052] Figure 1 A method flow chart of the analysis method in an embodiment of the present invention;

[0053] Figure 2 A connection diagram of an analysis system in an embodiment of the present invention;

[0054] Figure 3 A schematic diagram of the connection of electronic equipment in an embodiment of the present invention;

[0055] Figure 4 Schematic diagram of the process of performing structural analysis using MIDAS GEN software or the like in an embodiment of the present invention;

[0056] Figure 5 It is a schematic diagram of a flow chart for performing earthquake resistance and earthquake resistance analysis in an embodiment of the present invention. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0058] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0060] In the description of the embodiments of the present invention, "plurality" means at least 2.

[0061] Example 1: In order to predict and judge the load required for large-span and ultra-long structures, the influence of various load conditions on such complex structures is accurately considered, and a reference and preparation is provided for the construction of special-shaped structures, and a strong data support is provided for the equipment selection (such as damper parameters, etc.) and connection methods for the construction of special-shaped structures. This embodiment provides a special-shaped structure analysis method based on a deep neural network, such as Figure 1 As shown, the specific steps include:

[0062] S1, obtaining shape structural parameters according to a preset three-dimensional model of the special-shaped structure, dividing the shape of the special-shaped structure into regions based on the shape structural parameters, and obtaining multiple load regions that affect the special-shaped structure.

[0063] Among them, the area division of the special-shaped structure can be based on the change trend of the structure, that is, if the maximum bending or change angle of each load area obtained by division does not exceed the preset angle value, it can be considered that the wind pressure, vibration and other influences on the load area are consistent.

[0064] S2, based on the maximum airflow velocity and the corresponding equivalent wind pressure in the special-shaped structure, the initial wind pressure load of each load area is obtained, and the compensation coefficient for compensating the initial wind pressure load is obtained by using each equivalent wind pressure, and the initial wind pressure load is compensated by using the compensation coefficient to obtain the corrected wind pressure load of each load area.

[0065] Among them, since the movement of the large-span roof structure under the action of random wind load can be described as the random wind load generated by the joint action of the mass matrix, the damping matrix and the stiffness matrix, the equivalent wind pressure can be derived based on the random wind load.

[0066] Optionally, the above equivalent wind pressure represents the wind pressure time history obtained by dividing the algebraic sum of the elastic restoring forces of all nodes in each load area of ​​the special-shaped structure projected in a certain direction by the projection area of ​​the corresponding load area in the direction. The equivalent wind pressure of each load area is specifically:

[0067] ,in:

[0068] ; , ;

[0069] In the formula, Indicates the load area i The equivalent wind pressure, represents the modal coordinate function, represents the mass matrix, represents the modal characteristic matrix obtained through the heterogeneous structure, represents the modal eigenvalue matrix, represents the influence coefficient matrix, represents the elastic restoring force of node j on the load area i The contribution of the equivalent wind pressure is Indicates the load area i The projected area of They respectively represent the angles between the projection direction and the X, Y, and Z axes in the preset spatial rectangular coordinate system.

[0070] Among them, the equivalent wind pressure is different from the actual wind pressure acting on the structure. It is a wind pressure derived from the wind vibration response, which includes the influence of background and resonance components. Under reasonable zoning, the regional equivalent wind pressure calculated by the elastic restoring force of the large-span roof structure can reflect the wind-induced response characteristics of this type of structure.

[0071] Optionally, the above initial wind pressure load is specifically:

[0072] ; In the formula, Indicates the load area i The initial wind pressure load is represents the average wind pressure in each load area, which is directly obtained through simulation experiments. g represents the peak factor. The correlation coefficient matrix is i List, A diagonal matrix representing the standard deviation of the equivalent wind pressure for each load area.

[0073] Specifically, since zoning will cause errors in the calculation of wind pressure, the correlation between the displacement response and the equivalent wind pressure can be used to compensate for the initial wind pressure load, thereby improving the calculation accuracy of the wind pressure in each load area.

[0074] Optionally, the above compensation coefficient is obtained by:

[0075] S21, obtaining a modal response of each load area based on each equivalent wind pressure, and obtaining a predicted displacement response of each load area based on the modal response.

[0076] The above predicted displacement response is specifically:

[0077] ,in, ;

[0078] In the formula, Indicates the load area i The predicted displacement response, The modal characteristic matrix of the special-shaped structure, Indicates the load area i The modal response of represents the average response of each mode, g represents the peak factor, represents the diagonal matrix of the modal response standard deviations, The first column represents the correlation coefficient matrix between the modal response and the equivalent wind pressure. i List.

[0079] S22, based on the predicted displacement response and the actual displacement response of each load area, a first coefficient and a second coefficient are obtained by fitting through a least square method, so that the compensated response is the same as the actual response, and the compensation coefficient includes the first coefficient and the second coefficient.

[0080] The predicted displacement response and the actual displacement response based on each load area are fitted by the least square method to obtain the first coefficient and the second coefficient, which are specifically:

[0081] , ;

[0082] In the formula, denote the mean and pulsating parts of the predicted displacement response, respectively. They represent the mean and pulsating parts of the actual displacement response, Represent the first coefficient and the second coefficient respectively. The first coefficient and the second coefficient are determined when the values ​​are minimum.

[0083] Furthermore, the above-mentioned corrected wind pressure load is specifically:

[0084] ; In the formula, Indicates the load area i Corrected wind pressure load, represents the average wind pressure in each load area, which is directly obtained through simulation experiments. g represents the peak factor. The correlation coefficient matrix is i List, Represents the diagonal matrix consisting of the standard deviation of the equivalent wind pressure in each load area, denote the first coefficient and the second coefficient respectively.

[0085] S3, based on the maximum operating temperature of each load area when in use, the corrected wind pressure load is input into a temperature compensation model based on a deep neural network and trained for processing, so as to obtain a target wind pressure load that has been temperature compensated and corresponds to the corrected wind pressure load of each load area.

[0086] Optionally, the temperature compensation model is obtained by:

[0087] S31, calculate the corrected wind pressure loads at different airflow velocities, and simulate the real wind pressure loads of the corrected wind pressure loads at different temperatures and different airflow velocities based on MIDAS GEN software, and construct training samples through each corrected wind pressure load and the corresponding real wind pressure load.

[0088] S32, training the preset initial model using the training samples until a preset training end condition is reached, and determining the preset initial model that reaches the preset training end condition as the temperature compensation model, wherein the preset training end condition is that the loss function does not exceed a threshold.

[0089] Specifically, the loss function is:

[0090] ,in:

[0091] , ;

[0092] In the formula, represents the loss function, represents the first function, represents the second function, Indicates the training sample i The actual wind pressure load of the samples, Indicates the training sample i The predicted wind pressure load of samples, n represents the total number of samples in the training sample, Indicates the training sample i The absolute value of the difference between the true wind pressure load of a sample and the corresponding corrected wind pressure load, represent the weight of the first function and the weight of the second function respectively.

[0093] S4, using MIDAS GEN software to simulate the relationship between the peak frequency and pulsating pressure of the special-shaped structure and the wind speed, and obtain the corresponding relationship characteristics, and perform a steady-state analysis of the special-shaped structure based on the relationship characteristics and the target wind pressure load.

[0094] Among them, the process simulated by MIDAS GEN software is as follows Figure 4 As shown, see Figure 4First, a calculation model is established according to the complex structural composition, complex boundary conditions and complex load conditions of the special-shaped structure. Then, the functions in the software are used to simulate such as grid analysis and rod selection, stress ratio control, surface accuracy control, slenderness ratio requirements, fatigue stress amplitude control, etc. The internal force is reviewed and ball node design is performed by using 3D3S, and the non-intersection of welding ball welds is strictly controlled for adjustment; secondly, SAP2000 is used for seismic elastoplastic analysis to complete the structural design of the roof panel and hanger system, and then the overall model is analyzed, including the grid structure, upper roof panel and lower hanger, lower concrete structure, etc., until all structures meet the conditions; finally, the construction process analysis model is used to perform construction simulation analysis to determine the final structure.

[0095] Optionally, the relationship between the peak frequency of the pulsating pressure and the pulsating pressure in the special-shaped structure and the wind speed is simulated, specifically:

[0096] S41, obtaining wind pressure time history data during the fixed wind speed test according to the measured data of the small wind tunnel, and converting the wind pressure time history data from the time domain to the frequency domain through Fourier transform to obtain pulsating wind pressure frequency domain data corresponding to the wind pressure time history data.

[0097] S42, the pulsating wind pressure frequency domain data under different test wind speeds are summarized to obtain the relationship characteristics between the pulsating pressure peak frequency and the pulsating pressure in the special-shaped structure and the wind speed, and the special-shaped structure is subjected to a steady-state analysis based on the relationship characteristics and the target wind pressure load to obtain the response of the special-shaped structure under the pulsating wind pressure.

[0098] Among them, the steady-state analysis of the special-shaped structure is performed by using the relationship characteristics and target wind pressure load obtained in this embodiment, which can be as follows: Figure 5 For the process shown, see Figure 5 , firstly, the relationship characteristics and the target wind pressure loads of each area are obtained through the scheme in this embodiment; in the vibration control with low-frequency pulsating airflow vibration, the relationship characteristics and the target wind pressure loads of each area are first obtained to carry out pulsating load characteristic research, dynamic analysis method research and low-frequency pulsation analysis of large-span structures under complex alternating loads, and the design is optimized through various analysis results to finally obtain the structural scheme; for example, in vibration control, the seismic design parameters and performance targets are first determined, and then the seismic design of the maintenance structure in the cave body, the seismic analysis research of the roof structure and the seismic analysis research of the overall structure are carried out, and based on the research results, the structure is improved through strengthening measures until the final structural model meets the requirements of low-frequency pulsating airflow vibration and seismic fortification.

[0099] Specifically, large-span and ultra-long structures also need to consider the influence of other multi-factor loads, such as earthquake resistance and vibration resistance. These aspects can be inverted and simulated through MIDAS GEN software, and finally the target load for each load area can be obtained. Among them, by predicting and judging the loads required for large-span and ultra-long structures, the influence of various load conditions on such complex structures is accurately considered, which improves the accuracy and efficiency of structural analysis and design, reduces deviations caused by human factors, and saves a lot of time and resources. This provides a reference and preparation for the construction of special-shaped structures, and provides strong data support for the selection of equipment and connection methods for the construction of special-shaped structures.

[0100] Embodiment 2: This embodiment of the present application provides a special-shaped structure analysis system based on a deep neural network, such as Figure 2 As shown, including:

[0101] The area division module is used to obtain the shape structure parameters according to the preset three-dimensional model of the special-shaped structure, divide the shape of the special-shaped structure into regions based on the shape structure parameters, and obtain multiple load areas that affect the special-shaped structure.

[0102] The load correction module is used to obtain the initial wind pressure load of each load area based on the maximum airflow velocity and the corresponding equivalent wind pressure in the special-shaped structure, obtain the compensation coefficient for compensating the initial wind pressure load using each equivalent wind pressure, and compensate the initial wind pressure load using the compensation coefficient to obtain the corrected wind pressure load of each load area.

[0103] The load compensation module is used to input the corrected wind pressure load into the temperature compensation model trained based on the deep neural network for processing based on the maximum operating temperature of each load area when in use, so as to obtain the target wind pressure load that has been temperature compensated and corresponds to the corrected wind pressure load of each load area.

[0104] The steady-state analysis module is used to simulate the relationship between the peak frequency of the pulsating pressure and the pulsating pressure in the special-shaped structure and the wind speed using the MIDAS GEN software, and obtain the corresponding relationship characteristics, and perform a steady-state analysis on the special-shaped structure based on the relationship characteristics and the target wind pressure load.

[0105] Embodiment 3: The embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the method in Embodiment 1 is implemented when the processor executes the computer program.

[0106] Example 4: The embodiment of the present application provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions enable the computer to execute the method in Example 1.

[0107] The above specific implementation methods further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific implementation methods of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for analyzing special-shaped structures based on deep neural networks, characterized in that: The specific steps include: Acquire shape structural parameters according to a preset three-dimensional model of the special-shaped structure, divide the shape of the special-shaped structure into regions based on the shape structural parameters, and obtain multiple load regions that affect the special-shaped structure; Based on the maximum airflow velocity and the corresponding equivalent wind pressure in the special-shaped structure, the initial wind pressure load of each load area is obtained, and the compensation coefficient for compensating the initial wind pressure load is obtained by using each equivalent wind pressure, and the initial wind pressure load is compensated by using the compensation coefficient to obtain the corrected wind pressure load of each load area; Based on the maximum operating temperature of each load area when in use, the corrected wind pressure load is input into a temperature compensation model based on a deep neural network and trained for processing, so as to obtain a target wind pressure load that has been temperature compensated and corresponds to the corrected wind pressure load of each load area; MIDAS GEN software is used to simulate the relationship between the peak frequency of pulsating pressure and the pulsating pressure in the special-shaped structure and the wind speed, and the corresponding relationship characteristics are obtained. Based on the relationship characteristics and the target wind pressure load, a steady-state analysis of the special-shaped structure is performed.

2. The method for analyzing special-shaped structures based on deep neural networks according to claim 1, characterized in that: The relationship between the peak frequency of pulsating pressure and the pulsating pressure in the special-shaped structure and the wind speed is simulated, specifically: The wind pressure time history data in the fixed wind speed test is obtained according to the measured data of the small wind tunnel, and the wind pressure time history data is converted from the time domain to the frequency domain by Fourier transform to obtain the pulsating wind pressure frequency domain data corresponding to the wind pressure time history data; The pulsating wind pressure frequency domain data under different test wind speeds are summarized to obtain the relationship characteristics between the pulsating pressure peak frequency and the pulsating pressure in the special-shaped structure and the wind speed, and the special-shaped structure is subjected to a steady-state analysis based on the relationship characteristics and the target wind pressure load to obtain the response of the special-shaped structure under pulsating wind pressure.

3. The method for analyzing special-shaped structures based on deep neural networks according to claim 1, characterized in that: The initial wind pressure load is specifically: ; In the formula, Indicates the load area i The initial wind pressure load is represents the average wind pressure in each load area, which is directly obtained through simulation experiments. g represents the peak factor. The correlation coefficient matrix is i List, A diagonal matrix representing the standard deviation of the equivalent wind pressure for each load area.

4. The method for analyzing special-shaped structures based on deep neural networks according to claim 1, characterized in that: The equivalent wind pressure represents the wind pressure time history obtained by dividing the algebraic sum of the elastic restoring forces of all nodes in each load area of ​​the special-shaped structure in a certain direction by the projection area of ​​the corresponding load area in this direction. The equivalent wind pressure of each load area is specifically: ,in: ； ， ； In the formula, Indicates the load area i The equivalent wind pressure, represents the modal coordinate function, represents the mass matrix, The modal characteristic matrix of the special-shaped structure, represents the modal eigenvalue matrix, represents the influence coefficient matrix, represents the elastic restoring force of node j on the load area i The contribution of the equivalent wind pressure is Indicates the load area i The projected area of They respectively represent the angles between the projection direction and the X, Y, and Z axes in the preset spatial rectangular coordinate system.

5. The method for analyzing special-shaped structures based on deep neural networks according to claim 1, characterized in that: The compensation coefficient is obtained by: Based on each of the equivalent wind pressures, a modal response of each load area is obtained, and based on the modal response, a predicted displacement response of each load area is obtained. The predicted displacement response is specifically: ,in, ; In the formula, Indicates the load area i The predicted displacement response, The modal characteristic matrix of the special-shaped structure, Indicates the load area i The modal response of represents the average response of each mode, g represents the peak factor, represents the diagonal matrix of the standard deviations of the modal responses, The first column represents the correlation coefficient matrix between the modal response and the equivalent wind pressure. i List; Based on the predicted displacement response and the actual displacement response of each load area, the first coefficient and the second coefficient are obtained by fitting through the least square method, and the compensation coefficient includes the first coefficient and the second coefficient.

6. The method for analyzing special-shaped structures based on deep neural networks according to claim 5, characterized in that: The predicted displacement response and the actual displacement response based on each load area are fitted by the least square method to obtain the first coefficient and the second coefficient, which are specifically: ， ； In the formula, denote the mean and pulsating parts of the predicted displacement response, respectively. They represent the mean and pulsating parts of the actual displacement response, Represent the first coefficient and the second coefficient respectively. The first coefficient and the second coefficient are determined when the value is the minimum; the corrected wind pressure load is specifically: ; In the formula, Indicates the load area i Corrected wind pressure load, represents the average wind pressure in each load area, which is directly obtained through simulation experiments. g represents the peak factor. The correlation coefficient matrix is i List, Represents the diagonal matrix consisting of the standard deviation of the equivalent wind pressure in each load area, denote the first coefficient and the second coefficient respectively.

7. The method for analyzing special-shaped structures based on deep neural networks according to claim 1, characterized in that: The temperature compensation model is obtained by: Calculate the corrected wind pressure loads at different airflow speeds, and simulate the real wind pressure loads of the corrected wind pressure loads at different temperatures and airflow speeds based on the MIDAS GEN software, and construct training samples through each corrected wind pressure load and the corresponding real wind pressure load; The preset initial model is trained using the training samples until a preset training end condition is reached, and the preset initial model that reaches the preset training end condition is determined as a temperature compensation model, and the preset training end condition is that the loss function does not exceed a threshold value, and the loss function is specifically: ,in: ， ； In the formula, represents the loss function, represents the first function, represents the second function, Indicates the training sample i The actual wind pressure load of the samples, Indicates the training sample i The predicted wind pressure load of samples, n represents the total number of samples in the training sample, Indicates the training sample i The absolute value of the difference between the true wind pressure load of a sample and the corresponding corrected wind pressure load, represent the weight of the first function and the weight of the second function respectively.

8. A special-shaped structure analysis system based on a deep neural network, applied to a special-shaped structure analysis method based on a deep neural network as claimed in any one of claims 1 to 7, characterized in that: include: A region division module is used to obtain the shape structure parameters according to the preset three-dimensional model of the special-shaped structure, and divide the shape of the special-shaped structure into regions based on the shape structure parameters to obtain multiple load regions affecting the special-shaped structure; A load correction module, for obtaining an initial wind pressure load of each load area based on the maximum airflow velocity and the corresponding equivalent wind pressure in the special-shaped structure, obtaining a compensation coefficient for compensating the initial wind pressure load using each equivalent wind pressure, and compensating the initial wind pressure load using the compensation coefficient to obtain a corrected wind pressure load of each load area; A load compensation module, which is used to input the corrected wind pressure load into a temperature compensation model trained based on a deep neural network for processing based on the maximum operating temperature of each load area when in use, so as to obtain a target wind pressure load that has been temperature compensated and corresponds to the corrected wind pressure load of each load area; The steady-state analysis module is used to simulate the relationship between the peak frequency of the pulsating pressure and the pulsating pressure in the special-shaped structure and the wind speed using the MIDAS GEN software, and obtain the corresponding relationship characteristics, and perform a steady-state analysis on the special-shaped structure based on the relationship characteristics and the target wind pressure load.

9. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the method according to any one of claims 1 to 7 is implemented when the processor executes the computer program.

10. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium stores computer instructions, which enable a computer to execute the method of any one of claims 1-7.

Images