Mechanical response and collapse assessment method and system of frame structures adapting to component removal

Abstract

The present invention discloses a method and system for evaluating the mechanical response and collapse of a frame structure that is adaptable to the removal of components. The method comprises: using a pre-trained and optimized digital prediction model for the mechanical response of a frame structure to predict the structural mechanical response of a field frame after the removal of components, and then obtaining a collapse evaluation result of the field frame structure according to the frame structure collapse judgment standard; the digital prediction model for the mechanical response of the frame structure comprises a cascaded spatial domain graph convolution layer, a depth graph convolution layer, and an expansion graph convolution layer. The evaluation method of the present invention has strong generalization ability and high computational efficiency.

Description

Technical Field

[0001] The invention belongs to the technical field of data processing, and in particular to a method and system for evaluating the mechanical response and collapse of a frame structure adaptable to component removal. Background Art

[0002] Frame structure collapse assessment refers to the prediction and analysis of the process in which the components of the frame structure are gradually destroyed due to material yielding, fracture, connection failure or loss of stability under the action of unexpected loads, which in turn leads to the overall or partial functional failure of the structure.

[0003] Events such as accidental explosions and natural disasters are usually the main causes of frame building collapse. These extreme events and their consequences seriously threaten the safety of buildings, property and lives. In order to cope with these disaster events, it is of great significance to quickly and accurately evaluate the collapse behavior of buildings under strong dynamic loads. However, the building collapse assessment process involves complex nonlinear and dynamic effects. Considering the diversity of building types and the uncertainty of disasters, it brings great limitations and challenges to the test and evaluation of the building collapse process and the real-time determination of the collapse area.

[0004] At present, the methods used to study the process or effect of continuous collapse of buildings mainly include theoretical research and numerical analysis. Among them, theoretical research mainly describes the physical behavior of building structures by establishing mathematical models, and usually appropriately simplifies the building structure. Some basic assumptions are also made according to the characteristics of the building during the research process. For example, in the Chinese invention patent application "A method for evaluating and predicting the collapse of steel structure buildings under explosion scenarios" (application number: CN202111344617.1, publication date: 2022-02-18), when determining the collapse analysis of steel structures under explosions, the scope of damaged beams and plate components is considered. According to the theoretical calculation formula of the reduction coefficient of the anti-collapse capacity of beam and plate structures under explosions, combined with the principle of resistance equivalence, a method for simulating the damage of steel structure beams and plates under explosion scenarios is obtained. Finally, a continuous collapse analysis model of steel structures is established to evaluate the anti-continuous collapse performance of the structure. It is not difficult to find from this study that theoretical methods are very suitable for understanding the basic mechanism of continuous collapse of buildings. The method is relatively economical and can provide guidance in the early stages of building design. However, since theoretical models often require simplification and assumptions as a premise, they cannot be easily used to evaluate real scenarios.

[0005] Numerical analysis mainly includes finite element method, discrete element method and applied element method, etc. Different methods have their corresponding applicable conditions and application scope. With the rapid development of computer technology, software iteration and algorithm development have continuously improved the computational basis of numerical analysis, enabling these methods to handle more and more complex problems. For example, in the Chinese invention patent application "A method and device for simulating post-earthquake fire spread considering building collapse" (application number: CN202410077172.2, publication date: 2024-05-28), a geometric model of the target building complex is established, and the geometric model is simulated for earthquake collapse to obtain the earthquake damage morphological data of the building complex, which is imported into the fire numerical simulation platform. After obtaining the post-earthquake building complex numerical model, sparse grid division considering height is performed, and sparse grid acceleration calculation of the post-earthquake fire spread simulation of the building complex is performed based on large eddy simulation and parallel computing theory to obtain the simulation results of the post-earthquake fire spread of the building complex. However, the modeling and computational efficiency of numerical methods cannot meet the needs of rapid and accurate assessment of building collapse, especially in the face of collapse assessment of large scenes and large-scale buildings, which will bring huge time costs and computing resource consumption.

[0006] In summary, the problems with the existing frame structure collapse assessment methods are: the accuracy of the collapse assessment of frame structures under multiple scenarios and multiple working conditions is not high enough, the versatility is poor, and the effectiveness is relatively low. Summary of the invention

[0007] In view of this, the purpose of the present invention is to provide a method and system for mechanical response and collapse assessment of frame structures that are adaptable to component removal, which uses a graph convolutional neural network to accurately predict the overall structural response of a three-dimensional frame structure after its components are removed, and combines the collapse criterion of the frame structure with strong generalization ability and high computational efficiency.

[0008] The technical solution to achieve the purpose of the present invention is:

[0009] In the first aspect, a method for evaluating the mechanical response of a frame structure that is adaptable to the removal of components is provided, wherein a pre-trained and optimized digital intelligent prediction model for the mechanical response of a frame structure is used to predict the structural mechanical response of a field frame after the removal of components, thereby realizing the evaluation of the mechanical response of the frame structure; the digital intelligent prediction model for the mechanical response of the frame structure comprises a cascaded spatial domain graph convolution layer, a depth graph convolution layer, and a dilated graph convolution layer;

[0010] The spatial domain graph convolution layer includes a first graph pooling module, a first graph convolution module, a first adaptive activation function AAF, a second graph convolution module, a second adaptive activation function AAF, a second graph pooling module, a third graph convolution module and a third adaptive activation function AAF. The input data enters the first graph pooling module and the first graph convolution module respectively and then enters the first adaptive activation function AAF together. The output of the first adaptive activation function AAF enters the second graph convolution module and the second adaptive activation function AAF in sequence. The output of the second adaptive activation function AAF enters the second graph pooling module and the third graph convolution module respectively and then enters the third adaptive activation function AAF together. The output of the third adaptive activation function AFF is the potential graph data G0 of the input data.

[0011] The deep graph convolution layer includes stacking n ResGCN blocks, whose input is G0 and output is the abstract feature G n ;

[0012] The dilated graph convolution layer includes a heterogeneous feature aggregation layer and a dilated pooling layer. The heterogeneous feature aggregation layer includes three parallel graph convolution modules of different sizes. The input of each graph convolution module is G n , the outputs are respectively passed through a heterogeneous cache module and then enter the aggregation operation module; the dilated pooling layer includes a cascaded first upsampling graph transposed convolution module, a first pooling module, a second upsampling graph transposed convolution module, a second pooling module and a third upsampling graph transposed convolution module. The input of the first upsampling graph transposed convolution module is the output of the aggregation operation module, and the output of the third upsampling graph transposed convolution module is the structural mechanical response data G of the frame. m .

[0013] As a further optimization scheme of the present invention, the on-site input data is input into a pre-trained digital prediction model of the mechanical response of the frame structure to obtain the corresponding mechanical response of the frame structure; the on-site input data includes the component dismantling conditions and the corresponding on-site frame structure parameters, and the on-site frame structure parameters include the geometric characteristics of the frame structure, material properties, dismantling location, and dismantling quantity.

[0014] As a further optimization scheme of the present invention, a multi-scenario framework structure mechanical response database including a training data set and a verification data set is used to complete the pre-training and optimization of the framework structure mechanical response digital intelligent prediction model;

[0015] The steps for constructing the multi-scenario frame structure mechanical response database are as follows:

[0016] According to the task requirements, the characteristics of the frame structure are designed, and the working conditions of removing beams and columns in the frame structure at different positions are set. The calculation domain of the frame structure after the discrete removal of components is divided by meshing, and the mechanical response of the frame structure under different working conditions is calculated. Combined with batch processing technology, the mechanical response data of the frame structure is batch generated;

[0017] The demolition conditions of the target frame structure in different scenarios are associated with its corresponding mechanical response data of the frame structure to obtain a multi-scenario frame structure mechanical response characteristic data set. Data features are extracted and fused for the multi-scenario frame structure mechanical response characteristic data set, and a multi-scenario frame structure mechanical response database including a training data set and a verification data set is constructed.

[0018] As a further optimization scheme of the present invention, the step of batch generating the mechanical response data of the frame structure comprises:

[0019] According to the task requirements, set the characteristics of the frame structure; according to the design technical standards, set the characteristics of the beams and columns; the characteristics of the beams and columns include the size of the beams and columns, and the cross-sectional shape; the characteristics of the frame structure include the length, width, and height of the frame structure, the number of floors, the layout of the rooms on each floor, and the material properties; the material properties include the elastic modulus, density, and Poisson's ratio;

[0020] Generate multiple frame structure feature models covering possible frame design types according to frame structure features and beam-column features;

[0021] Traverse all components of the frame structure and set the demolition plan for beams and columns at different positions of the frame structure; for beam components, select beams of different floors and different spans as demolition objects; for column components, select different side columns, middle columns, and corner columns as demolition objects;

[0022] Combined with the characteristics of the frame structure, single component dismantling conditions and multi-component combined dismantling conditions are formulated, and a unique condition identification code is assigned to each component dismantling condition;

[0023] Use the meshing tool to discretize the frame structure after the components are removed, and locally encrypt the mesh at the removed parts to obtain a frame structure with locally encrypted mesh;

[0024] The mesh of the frame structure with local mesh encryption is imported into the simulation software, and the nonlinear static analysis method is used to calculate the mechanical response of the frame structure after the components are removed;

[0025] According to the mechanical response of the frame structure after the components are removed, the mechanical response data of all nodes of the frame structure are extracted, and the mechanical response data includes displacement, stress and strain information;

[0026] Batch process multiple demolition conditions, extract the mechanical response of the frame structure under each condition, and batch generate mechanical response data of the frame structure.

[0027] As a further optimization scheme of the present invention, the step of constructing the multi-scenario frame structure mechanical response database includes:

[0028] Based on the working condition identification code, the multi-source information of the frame structure's geometric characteristics, material properties, demolition location, and demolition quantity in each demolition working condition is associated with the corresponding frame structure mechanical response data, and the distance radiation characteristics between the demolished components and the frame structure are calculated to obtain a multi-scenario frame structure mechanical response characteristic data set.

[0029] The demolition conditions and distance radiation features in the multi-scenario frame structure mechanical response feature data set are trimmed, normalized, feature extracted and multi-source data fused to obtain the demolition condition feature vector of multi-source information fusion;

[0030] Obtain demolition condition feature vectors under different scenarios, stack all demolition condition feature vectors and corresponding frame structure mechanical response data to form a set with demolition condition vectors as input and frame structure mechanical response data as output;

[0031] The collection is divided to obtain a multi-scenario frame structure mechanical response database including a training data set and a validation data set.

[0032] As a further optimization scheme of the present invention, the calculation expression of the distance radiation characteristic is:

[0033] ,

[0034] In the formula, n represents the number of nodes on the removed component, N represents the number of nodes on the frame structure, It represents the distance radiation characteristic from the i-th node on the removed component to the j-th node on the frame structure. , , They represent the horizontal coordinate, vertical coordinate and vertical coordinate of the jth node on the frame structure respectively. , , They respectively represent the horizontal coordinate, vertical coordinate, and vertical coordinate of the i-th node on the removed component.

[0035] On the second aspect, a frame structure collapse assessment method that adapts to component removal is also provided, which uses the structural mechanical response assessment method as described above to assess the structural mechanical response of the on-site frame after component removal, and then obtains the collapse assessment result of the on-site frame structure based on the frame structure collapse judgment standard.

[0036] In a third aspect, a frame structure collapse assessment system adapted to component removal is also provided, comprising:

[0037] Model unit, used to pre-train and optimize the digital intelligent prediction model of the mechanical response of the frame structure;

[0038] A prediction unit, used to obtain prediction results of the structural mechanical response of the on-site frame after the components are removed;

[0039] The evaluation unit is used to obtain the collapse evaluation result of the on-site frame structure according to the frame structure collapse judgment standard.

[0040] In a fourth aspect, a computer-readable storage medium storing one or more programs is also provided, wherein the one or more programs include instructions, which, when executed by a computing device, cause the computing device to perform the collapse assessment method as described above.

[0041] In a fifth aspect, an electronic device is also provided, comprising one or more processors, one or more memories and one or more programs, wherein the one or more programs are stored in the one or more memories and are configured to be executed by the one or more processors, and the one or more programs include instructions for executing the collapse assessment method as described above.

[0042] Compared with the prior art, the present invention has the following significant advantages:

[0043] 1. Strong flexibility and good adaptability: The present invention adopts multi-source feature fusion and graph convolutional neural network algorithm, which can predict and evaluate the collapse effect of frame structure in real time according to different demolition conditions. Compared with traditional evaluation methods, the present invention improves the flexibility and adaptability of evaluation technology, and can more conveniently carry out damage and protection scheme design.

[0044] 2. More accurate prediction: The present invention forms a high-fidelity database through finite element calculation, and uses graph convolutional neural network to learn the potential invariant law of frame structure response data. The formed frame structure response prediction model can make accurate predictions of frame structure responses under different demolition conditions;

[0045] 3. High efficiency: The frame structure response prediction model established by the present invention omits a large number of intermediate processes in traditional simulations through an end-to-end prediction mode, and can directly output the structural response results with high efficiency. At the same time, the well-trained model can be quickly applied to different working conditions, and has the ability to efficiently evaluate the collapse of frame structures in multiple scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is the main flow chart of the present invention.

[0047] Figure 2 It is a flow chart of the steps for generating the framework structure response data.

[0048] Figure 3It is a flowchart of the construction of a multi-scenario framework structure response database.

[0049] Figure 4 It is a flow chart of the steps for establishing a digital prediction model for the mechanical response of frame structures.

[0050] Figure 5A It is a schematic diagram of the spatial domain graph convolution layer structure in the digital intelligent prediction model of the mechanical response of the frame structure.

[0051] Figure 5B is a depth graph convolutional layer.

[0052] Figure 5C It is a dilated graph convolutional layer.

[0053] Figure 6 It is a flow chart of the optimization steps of the digital prediction model for the mechanical response of frame structures.

[0054] Figure 7 It is a flow chart of the steps for evaluating the response and collapse of frame structures. DETAILED DESCRIPTION

[0055] like Figure 1 As shown, the method for evaluating the mechanical response and collapse of a frame structure adapted to component removal of the present invention comprises the following steps:

[0056] S10, frame structure response data generation: Design the target frame structure characteristics, set the working conditions of removing beams and columns in the frame structure at different positions, divide the calculation domain of the frame structure after discrete removal of components through meshing, calculate and analyze the mechanical response of the frame structure under different working conditions, and combine batch processing technology to batch generate frame structure mechanical response data;

[0057] like Figure 2 As shown, the step S10, generating framework structure response data, includes:

[0058] S11. Design of target frame structure characteristics: according to the design technical standards, set the characteristics of beams and columns, including the size of beams and columns, and the cross-sectional shape; according to the target task requirements, set the geometric characteristics of the frame structure, including the basic structural characteristics of the frame structure, the length, width, and height of the structure, the number of floors, the layout of rooms on each floor, and material properties, including elastic modulus, density, and Poisson's ratio;

[0059] S12, generating multiple frame structures: generating multiple frame structure feature models covering possible frame design types according to the frame structure features and beam and column features;

[0060] S13. Design of beam and column demolition objects: traverse all components of the frame structure and set demolition plans for beams and columns at different positions of the frame structure; for beam components, select beams of different floors and different spans as demolition objects; for column components, select different side columns, middle columns, and corner columns as demolition objects;

[0061] S14. Demolition working condition planning: Based on the characteristics of the framework structure, formulate single component demolition working conditions and multi-component combined demolition working conditions, and assign a unique working condition identification code to each component demolition working condition;

[0062] S15. Frame structure mesh division: Use mesh division tools to discretize the frame structure after the components are removed, and locally encrypt the mesh at the removed component location to obtain a frame structure with locally encrypted mesh;

[0063] S16, numerical simulation calculation: importing the mesh of the frame structure with the locally encrypted mesh into the simulation software, and using a nonlinear static analysis method to simulate and calculate the mechanical response of the frame structure after the components are removed;

[0064] S17, simulation data collection: according to the mechanical response of the frame structure after the components are removed, extract the mechanical response data of key nodes and cross-sections, including displacement, stress, and strain information;

[0065] S18. Simulation and batch data processing: Write automated scripts to batch process various demolition conditions, automatically generate input files and call simulation software for simulation calculations, extract the mechanical response data of the frame structure under each condition, and obtain batch-generated mechanical response data of the frame structure.

[0066] S20, constructing a multi-scenario frame structure response database: using the frame structure mechanical response data, associating the component dismantling conditions of the frame structure in different scenarios with their corresponding structural mechanical responses, and calculating the distance radiation characteristics to obtain a multi-scenario frame structure mechanical response feature data set, extracting and fusing data features of the multi-scenario frame structure response feature data set, and constructing a multi-scenario frame structure mechanical response database including a training data set and a verification data set;

[0067] like Figure 3 As shown, the step S20, building a multi-scenario framework structure response database includes:

[0068] S21. Correlation between response data and demolition conditions: Based on the condition identification code, the multi-source information of the frame structure's geometric features, material features, demolition location, and demolition quantity in each demolition condition is correlated with the corresponding frame structure mechanical response data, and the distance radiation features are calculated to obtain a multi-scenario frame structure mechanical response feature data set;

[0069] Step S21 completes the preliminary association of data features, forming a corresponding relationship between the target frame demolition conditions and the target frame response data, ensuring that subsequent data preprocessing can accurately match the demolition conditions with the response data one by one.

[0070] S22. Generation of distance radiation characteristics: Based on the position and quantity information of the removed components and the characteristics of the frame structure, the distance radiation information between the removed components and the overall frame structure is calculated to generate the distance radiation characteristics. The calculation expression is:

[0071] ,

[0072] In the formula, n represents the number of nodes on the removed component, N represents the number of nodes on the frame structure, It represents the distance radiation from the i-th node of the demolished component to the j-th node of the frame structure. represents the horizontal coordinate of the frame structure node j, represents the corresponding ordinate, represents the corresponding vertical coordinate, , , They respectively represent the horizontal, vertical and vertical coordinates of the node i of the demolished component.

[0073] S23, multi-source data fusion: the multi-source demolition condition information and distance radiation characteristics in the scene framework structure response feature data set are subjected to data pruning, normalization, feature extraction and multi-source data fusion to obtain a demolition condition feature vector of multi-source information fusion.

[0074] The step of S23, extracting and fusing multi-source data features, includes:

[0075] S231, data pruning: pruning the geometric features of the frame structure, material properties, component removal positions, component removal quantities, and distance radiation feature information, and integrating heterogeneous features of different dimensions and magnitudes into the same dimension and magnitude;

[0076] S232, data normalization: After data pruning is completed, the data needs to be normalized to the minimum and maximum values. This method scales all features to the range of 0, 1. The calculation formula can be expressed as:

[0077] ,

[0078] In the formula Represents the original data, represents the normalized data;

[0079] S233, feature extraction: use a multi-layer convolutional neural network to downsample and encode the normalized data, remove redundant information, and extract the potential invariant features of the data;

[0080] S234, multi-source data fusion process: connect multiple vectors of the encoded frame structure geometric features, material properties, component removal positions, and component removal quantity information along one dimension to obtain a demolition condition feature vector fused with multi-source information.

[0081] The fused multi-source information removal condition feature vector is as follows:

[0082] ,

[0083] Among them, M represents the fusion data, Z k Normalized data characterizing the multi-source removal condition (e.g., frame length, height, width, number of removed components, location coordinates, etc.);

[0084] S24. Data entry: For multi-source demolition conditions in different scenarios, each scenario is fused in turn through a multi-source data fusion method to obtain the demolition condition feature vector of the corresponding scenario. All demolition condition feature vectors and the corresponding frame structure response results are stacked to form a multi-scenario multi-source information fused demolition condition feature vector as input feature, and the corresponding frame structure mechanical response set is entered into the library system as the output result. The feature vector set is divided to obtain a multi-scenario frame structure mechanical response database including a training data set and a verification data set.

[0085] S30. Establishment of a digital prediction model: construct a graph convolutional neural network, use the training data set to train the graph convolutional neural network, and obtain a digital prediction model of the mechanical response of the frame structure based on the graph convolutional neural network.

[0086] like Figure 4 As shown, the step of establishing a digital prediction model in S30 includes:

[0087] S31. Graph convolutional neural network construction: By constructing spatial domain graph convolutional layer, depth graph convolutional layer and dilated graph convolutional layer, a graph convolutional neural network is obtained;

[0088] The step of constructing the graph convolutional neural network in S31 includes:

[0089] S311, spatial domain graph convolution layer construction: Figure 5A As shown in the figure, the graph convolution module and the graph pooling module are introduced to construct the spatial domain graph convolution layer, which is used to identify the spatial features of the input data. Specifically, the three graph convolution modules are connected in sequence with the adaptive activation function (AAF) to form the spatial feature learning ability; the two graph pooling modules are jump-connected to form the edge feature learning ability.

[0090] Among them, the graph pooling module uses the maximum pooling method to extract edge features. The extracted edge features are spliced ​​with the results calculated by the graph convolution module to obtain a spatial feature recognition layer that can enhance edge information. The potential graph data G0 of the input data can be calculated through the spatial domain graph convolution layer.

[0091] S312, depth graph convolution layer construction: Figure 5B As shown in the figure, ResGCN is used as the backbone block of the graph convolutional network, where each ResGCN is composed of two graph convolution modules and an activation function. By stacking n ResGCN blocks, each ResGCN is jump-connected by a graph convolution module to form a deep structure. The message passing mechanism and information mean aggregation method are used to make each ResGCN block communicate with each other, thereby constructing a deep graph convolution layer.

[0092] In the deep graph convolution layer, the potential graph data G0 obtained by the spatial domain graph convolution layer is iteratively processed through n ResGCN structures with the same structure to achieve in-depth analysis and feature extraction of the data; each ResGCN structure iteration includes the update of two key information, namely the edge information and the point information, and finally the abstract feature G is obtained. n ;

[0093] S313, dilated graph convolution layer construction: Figure 5C As shown in the figure, it is composed of a heterogeneous feature aggregation layer and a dilated pooling layer connected in sequence, wherein the heterogeneous feature aggregation layer includes graph convolution modules of three sizes, a heterogeneous cache module and an aggregation operation; the dilated pooling layer includes three upsampling graph transposed convolution modules and two pooling modules.

[0094] The input of the dilated graph convolution layer is the abstract feature G output by the deep graph convolution layer. m , after expansion calculation and restoration, the structural response data with frame node unit displacement can be obtained.

[0095] S314. Graph convolutional neural network construction: A graph convolutional neural network is formed by splicing spatial domain graph convolutional layers, depth graph convolutional layers and dilated graph convolutional layers.

[0096] S32. Establishment of a digital prediction model: Use the training data set to train the graph convolutional neural network to obtain a digital prediction model of the mechanical response of the frame structure based on the graph convolutional neural network.

[0097] S40, optimization of the digital intelligent prediction model: using the verification data set to test the digital intelligent prediction model of the mechanical response of the frame structure based on the graph convolutional neural network, performing error evaluation on the test, and performing structural optimization on the digital intelligent prediction model of the mechanical response of the frame structure based on the graph convolutional neural network according to the error evaluation result to obtain an optimized digital intelligent prediction model of the mechanical response of the frame structure based on the graph convolutional neural network.

[0098] like Figure 6 As shown, the step S40, the digital prediction model optimization step includes:

[0099] S41. Error assessment: Use the verification data set to verify the mathematical prediction model of the mechanical response of the frame structure based on the graph convolutional neural network, and use the error metric to perform error assessment by comparing the mechanical response results of the frame structure predicted by the prediction model with the simulation data or experimentally measured data.

[0100] The S41 error evaluation step is specifically as follows:

[0101] The mean square error (MSE) is obtained by calculating the average of the squared differences between the frame structure response results predicted by the calculation model and the simulation data or experimental data, which measures the overall error. To measure the prediction of the model and evaluate the overall fitting effect of the model.

[0102] S42. Optimization of digital prediction model: According to the error evaluation results, an optimized digital prediction model of the mechanical response of the frame structure based on graph convolutional neural network is obtained through parameter tuning, structural optimization, data structure enhancement and balance.

[0103] The step S42, the step of optimizing the digital prediction model, includes:

[0104] S421, Hyperparameter tuning: Perform parameter tuning based on error evaluation results;

[0105] S422, structural optimization: by adjusting the number of layers, node aggregation method and convolution method of each level in the digital intelligent prediction model of mechanical response of the frame structure, the model structure is optimized, and the prediction results of the digital intelligent prediction model of mechanical response of the frame structure based on the graph convolutional neural network after structural optimization are evaluated for error, and the structure is tuned;

[0106] S423, data structure enhancement and balance: balance the proportion of different scene data under specific conditions through undersampling and oversampling methods, and achieve data enhancement and balance through scene data expansion, data perturbation, and changing graph structure;

[0107] S424, model optimization: by repeating steps S421 to S423 until the error evaluation standard is met, an optimized digital prediction model of the mechanical response of the frame structure based on the graph convolutional neural network is obtained.

[0108] S50. Frame structure response and collapse assessment: Based on the on-site frame structure parameters and component removal information, an on-site input data set is constructed, and the data of the on-site input data set is input into the optimized digital prediction model of the mechanical response of the frame structure based on the graph convolutional neural network to obtain the mechanical response of the frame structure after the component removal, and according to the frame structure collapse judgment standard, the collapse assessment result of the on-site frame structure is obtained.

[0109] like Figure 7 As shown, the step S50, frame structure response and collapse assessment, includes:

[0110] S51. Determination of target frame structure parameters: Combine the construction design drawings and on-site measurements to obtain the geometric parameters and material properties of the on-site frame structure.

[0111] For example: The measured frame structure has 5 layers from bottom to top, the steel frame is fixedly connected to the ground, and the whole structure is added with self-contact conditions. The number of columns in the horizontal direction of each layer of the frame is 11, and the number of columns in the vertical direction is 5. The reinforcement ratio of the beam component is 0.3, the reinforcement ratio of the column component is 0.5, the strength grade of the steel bar is 500Mpa, and the maximum force total elongation limit is 5.0;

[0112] S52, constructing a prediction input data set: extracting and fusing multi-source data features on the geometric features and material properties of the on-site frame structure, beam and column objects, demolition locations, and demolition quantity information to obtain an on-site prediction input data set;

[0113] S53, frame structure response prediction: mathematically inputting the field prediction input data set into the optimized frame structure mechanical response mathematical intelligence prediction model based on graph convolutional neural network to obtain the frame structure mechanical response after the components are removed;

[0114] S54. Frame structure collapse assessment: According to the frame structure collapse judgment standard, based on the mechanical response of the frame structure after the components are removed, the collapse type of the target frame structure is assessed.

[0115] The ratio of the relative displacement at both ends of the frame structure beams and columns to the span is used as the judgment standard, and the standards are as follows: a ratio of 0-0.01 means no risk of collapse; a ratio of 0.01-0.05 means slight displacement, which has exceeded the normal range but has not yet reached the critical state of collapse; a ratio of 0.05-0.10 means moderate displacement, and the structure may have been damaged to a certain extent and partially collapsed; a ratio of 0.10-0.20 means severe displacement, and the structure may have been severely damaged and there is a risk of continuous collapse; a displacement greater than 0.20 will result in direct collapse.

[0116] Through the above steps, the collapse performance of the frame building structure after the beam and column components are removed can be systematically evaluated, which can provide effective data support for the subsequent anti-collapse protection design of the frame structure and post-collapse recovery suggestions.

[0117] The core advantages of the present invention lie in its flexibility and adaptability. Through multi-source feature fusion and graph convolutional neural network algorithms, it can predict and evaluate the collapse effect of the frame structure in real time for different demolition conditions. In terms of technical implementation, the present invention first establishes a multi-scenario frame structure response database through a high-precision numerical method, then constructs a digital prediction model, and trains it based on a graph convolutional neural network to obtain a frame structure response prediction model based on a graph convolutional neural network with fast learning ability and strong generalization ability, and performs error evaluation and parameter tuning on the model to improve the prediction ability of the digital prediction model. On this basis, combined with the frame structure collapse judgment criteria, the collapse results of the frame structure are further evaluated. The implementation of the present invention can realize real-time and accurate evaluation of the collapse of the frame structure under the destruction of beam-column components in multiple scenarios and at multiple scales.

[0118] The present invention also designs a frame structure collapse assessment system that is adaptable to component removal, comprising:

[0119] Model unit, used to pre-train and optimize the digital intelligent prediction model of the mechanical response of the frame structure;

[0120] A prediction unit, used to obtain prediction results of the structural mechanical response of the on-site frame after the components are removed;

[0121] The evaluation unit is used to obtain the collapse evaluation result of the on-site frame structure according to the frame structure collapse judgment standard.

[0122] The technical solution of the above-mentioned frame structure collapse assessment system is similar to the above-mentioned frame structure collapse assessment method, which will not be repeated here.

[0123] Based on the same technical solution, the present invention also discloses a computer-readable storage medium storing one or more programs, wherein the one or more programs include instructions, which, when executed by a computing device, enable the computing device to execute the above-mentioned frame structure collapse assessment method that adapts to component removal.

[0124] Based on the same technical solution, the present invention also discloses a computing device, including one or more processors, one or more memories and one or more programs, wherein the one or more programs are stored in the one or more memories and are configured to be executed by the one or more processors, and the one or more programs include instructions for executing the above-mentioned frame structure collapse assessment method for adaptive component dismantling.

[0125] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0126] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0127] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0128] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0129] The above embodiments are only for illustrating the technical idea of ​​the present invention, and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for evaluating the mechanical response of a frame structure adapted to component removal, characterized in that: The pre-trained and optimized digital prediction model for mechanical response of frame structures is used to predict the structural mechanical response of the on-site frame after the components are removed, so as to evaluate the mechanical response of the frame structure. The digital prediction model for mechanical response of frame structures includes a cascaded spatial domain graph convolution layer, a depth graph convolution layer, and a dilated graph convolution layer. The spatial domain graph convolution layer includes a first graph pooling module, a first graph convolution module, a first adaptive activation function AAF, a second graph convolution module, a second adaptive activation function AAF, a second graph pooling module, a third graph convolution module and a third adaptive activation function AAF. The input data enters the first graph pooling module and the first graph convolution module respectively and then enters the first adaptive activation function AAF together. The output of the first adaptive activation function AAF enters the second graph convolution module and the second adaptive activation function AAF in sequence. The output of the second adaptive activation function AAF enters the second graph pooling module and the third graph convolution module respectively and then enters the third adaptive activation function AAF together. The output of the third adaptive activation function AFF is the potential graph data G0 of the input data. The deep graph convolution layer consists of stacking n ResGCN blocks, whose input is G0 and output is the abstracted feature G n ; The dilated graph convolution layer includes a heterogeneous feature aggregation layer and a dilated pooling layer. The heterogeneous feature aggregation layer includes three parallel graph convolution modules of different sizes. The input of each graph convolution module is G n , the outputs are respectively passed through a heterogeneous cache module and then enter the aggregation operation module; the dilated pooling layer includes a cascaded first upsampling graph transposed convolution module, a first pooling module, a second upsampling graph transposed convolution module, a second pooling module and a third upsampling graph transposed convolution module. The input of the first upsampling graph transposed convolution module is the output of the aggregation operation module, and the output of the third upsampling graph transposed convolution module is the structural mechanical response data G of the frame. m .

2. The frame structure mechanical response evaluation method according to claim 1, characterized in that: The on-site input data is input into the pre-trained digital prediction model of the mechanical response of the frame structure to obtain the corresponding mechanical response of the frame structure; the on-site input data includes the component dismantling conditions and the corresponding on-site frame structure parameters. The on-site frame structure parameters include the geometric characteristics of the frame structure, material properties, dismantling location, and dismantling quantity.

3. The frame structure mechanical response evaluation method according to claim 1, characterized in that: The multi-scenario frame structure mechanical response database containing training data sets and verification data sets is used to complete the pre-training and optimization of the digital intelligent prediction model of the frame structure mechanical response; The steps for constructing the multi-scenario frame structure mechanical response database are as follows: According to the task requirements, the characteristics of the frame structure are designed, and the working conditions of removing beams and columns in the frame structure at different positions are set. The calculation domain of the frame structure after the discrete removal of components is divided by meshing, and the mechanical response of the frame structure under different working conditions is calculated. Combined with batch processing technology, the mechanical response data of the frame structure is batch generated; The demolition conditions of the target frame structure in different scenarios are associated with its corresponding mechanical response data of the frame structure to obtain a multi-scenario frame structure mechanical response characteristic data set. Data features are extracted and fused for the multi-scenario frame structure mechanical response characteristic data set, and a multi-scenario frame structure mechanical response database including a training data set and a verification data set is constructed.

4. The frame structure mechanical response evaluation method according to claim 3 is characterized in that: The step of batch generating the mechanical response data of the frame structure comprises: According to the task requirements, set the characteristics of the frame structure; according to the design technical standards, set the characteristics of the beams and columns; the characteristics of the beams and columns include the size of the beams and columns, and the cross-sectional shape; the characteristics of the frame structure include the length, width, and height of the frame structure, the number of floors, the layout of the rooms on each floor, and the material properties; the material properties include the elastic modulus, density, and Poisson's ratio; Generate multiple frame structure feature models covering several frame design types according to frame structure features and beam-column features; Traverse all components of the frame structure and set the demolition plan for beams and columns at different positions of the frame structure; for beam components, select beams of different floors and different spans as demolition objects; for column components, select different side columns, middle columns, and corner columns as demolition objects; Combined with the characteristics of the frame structure, single component dismantling conditions and multi-component combined dismantling conditions are formulated, and a unique condition identification code is assigned to each component dismantling condition; Use the meshing tool to discretize the frame structure after the components are removed, and locally encrypt the mesh at the removed parts to obtain a frame structure with locally encrypted mesh; The mesh of the frame structure with local mesh encryption is imported into the simulation software, and the nonlinear static analysis method is used to calculate the mechanical response of the frame structure after the components are removed; According to the mechanical response of the frame structure after the components are removed, the mechanical response data of all nodes of the frame structure are extracted, and the mechanical response data includes displacement, stress and strain information; Batch process multiple demolition conditions, extract the mechanical response of the frame structure under each condition, and batch generate mechanical response data of the frame structure.

5. The frame structure mechanical response evaluation method according to claim 4, characterized in that: The multi-scenario frame structure mechanical response database construction step includes: Based on the working condition identification code, the multi-source information of the frame structure's geometric characteristics, material properties, demolition location, and demolition quantity in each demolition working condition is associated with the corresponding frame structure mechanical response data, and the distance radiation characteristics between the demolished components and the frame structure are calculated to obtain a multi-scenario frame structure mechanical response characteristic data set. The demolition conditions and distance radiation features in the multi-scenario frame structure mechanical response feature data set are trimmed, normalized, feature extracted and multi-source data fused to obtain the demolition condition feature vector of multi-source information fusion; Obtain demolition condition feature vectors under different scenarios, stack all demolition condition feature vectors and corresponding frame structure mechanical response data to form a set with demolition condition vectors as input and frame structure mechanical response data as output; The collection is divided to obtain a multi-scenario framework structure mechanical response database including a training data set and a validation data set.

6. The frame structure mechanical response evaluation method according to claim 5, characterized in that: The calculation expression of distance radiation characteristics is: ， In the formula, n represents the number of nodes on the removed component, N represents the number of nodes on the frame structure, It represents the distance radiation characteristic from the i-th node on the removed component to the j-th node on the frame structure. , , They represent the horizontal, vertical and vertical coordinates of the jth node on the frame structure, respectively. , , They respectively represent the horizontal coordinate, vertical coordinate, and elevation coordinate of the i-th node on the removed component.

7. A method for assessing the collapse of a frame structure adapted to component removal, characterized in that: The structural mechanical response of the on-site frame after the components are removed is evaluated using the evaluation method as described in any one of claims 1 to 6, and then the collapse evaluation result of the on-site frame structure is obtained according to the frame structure collapse judgment standard.

8. A frame structure collapse assessment system based on the collapse assessment method according to claim 7, characterized in that: include: Model unit, used to pre-train and optimize the digital intelligent prediction model of the mechanical response of the frame structure; A prediction unit, used to obtain prediction results of the structural mechanical response of the on-site frame after the components are removed; The evaluation unit is used to obtain the collapse evaluation result of the on-site frame structure according to the frame structure collapse judgment standard.

9. A computer-readable storage medium storing one or more programs, wherein the one or more programs include instructions, characterized in that: When the instructions are executed by a computing device, the computing device is caused to perform the collapse assessment method of claim 7 .

10. An electronic device, characterized in that: The method comprises one or more processors, one or more memories and one or more programs, wherein the one or more programs are stored in the one or more memories and are configured to be executed by the one or more processors, and the one or more programs include instructions for executing the collapse assessment method according to claim 7.

Images