Building structure material usage prediction method and device, electronic equipment and storage medium

Abstract

The application provides a building structure material quantity prediction method and device based on a knowledge embedding neural network, electronic equipment and a storage medium, wherein the method comprises the following steps: obtaining design conditions of a target building structure, the design conditions comprising seismic design conditions, wind resistance design conditions and overall geometric design conditions of the structure; obtaining a structure design scheme of the target building structure, the structure design scheme comprising the spatial positions and geometric sizes of walls, columns, beams, plates and supports; inputting the design conditions and the structure design scheme into a pre-trained material quantity prediction model to obtain the material quantities of various materials in the target building structure. The method embeds prior knowledge into the material quantity prediction model, overcomes the defects of the prior art that the material quantity prediction value violates the basic common sense of engineering and has low prediction accuracy due to the inability to consider the prior knowledge in the field related to building structures, and realizes accurate and efficient prediction of the material quantity of the target building structure.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and building structure design technology, and in particular to a method, device, electronic device and storage medium for predicting the amount of building structure materials used. Background Technology

[0002] During the design and bidding process of building structures, the project cost is a core indicator of concern to both engineers and owners, and material usage is a significant component of this cost. Therefore, it is essential to efficiently and accurately assess the material usage based on the design scheme of the target building structure.

[0003] Traditional material usage assessments primarily rely on engineers' personal experience or calculations from structural design software. The former is unreliable, while the latter requires lengthy calculations, failing to meet the efficiency requirements of various computer-aided design methods.

[0004] To address this, existing technologies have proposed a material usage prediction method based on neural networks. However, this method cannot take into account prior knowledge in the field of building structures, resulting in predicted material usage that violates engineers' basic common sense and has low prediction accuracy.

[0005] Therefore, it is essential to address the problem that existing technologies cannot accurately and efficiently predict the material usage of target building structures. Summary of the Invention

[0006] This invention provides a method, apparatus, electronic device, and storage medium for predicting the amount of building structural materials used, which overcomes the shortcomings of existing technologies in accurately and efficiently predicting the amount of target building structural materials used, and achieves accurate and efficient prediction of the amount of target building structural materials used, while conforming to prior knowledge related to building structures.

[0007] On one hand, the present invention provides a method for predicting the material usage of building structures, comprising: obtaining design conditions for a target building structure, the design conditions including seismic design conditions, wind resistance design conditions, and overall structural geometric design conditions; obtaining a structural design scheme for the target building structure, the structural design scheme including the spatial location and geometric dimensions of walls, columns, beams, slabs, and supports; inputting the design conditions and the structural design scheme into a pre-trained material usage prediction model to obtain the material usage of various materials in the target building structure; wherein, the material usage prediction model is obtained by embedding prior knowledge, using a material usage dataset as the training set, and optimizing the training with the goal of minimizing the sum of material usage loss and prior knowledge loss.

[0008] Further, the step of inputting the design conditions and the structural design scheme into a pre-trained material usage prediction model to obtain the material usage of various materials in the target building structure includes: extracting a first high-dimensional feature from the structural design scheme; extracting a second high-dimensional feature from the design conditions; concatenating the first high-dimensional feature and the second high-dimensional feature to obtain a fused high-dimensional feature; and mapping the fused high-dimensional feature to material usage to obtain the material usage of various materials in the target building structure.

[0009] Furthermore, the material usage prediction model includes a first sub-model, a second sub-model, and a third sub-model; the extraction of the first high-dimensional feature from the structural design scheme includes: representing the structural design scheme as a structural design graph, wherein the structural design graph uses the nodes of the structural components as graph nodes, the node coordinates of the structural components as node features, the structural components as graph edges, and the geometric dimensions of the structural components as edge features; and inputting the structural design graph into the first sub-model to obtain the first high-dimensional feature.

[0010] Furthermore, the material usage prediction model includes a first sub-model, a second sub-model, and a third sub-model; the extraction of the second high-dimensional feature from the design conditions includes: representing the design conditions as a design feature vector, the design feature vector including seismic load, wind load, building height, and building width; inputting the design feature vector into the second sub-model to obtain the second high-dimensional feature.

[0011] Furthermore, the material usage prediction model includes a first sub-model, a second sub-model, and a third sub-model; the step of mapping the fused high-dimensional features to material usage to obtain the material usage of various materials in the target building structure includes: inputting the fused high-dimensional features into the third sub-model to obtain the material usage of various materials in the target building structure.

[0012] Furthermore, the prior knowledge includes: the amount of concrete used in the target building structure is less than the sum of the volumes of all concrete components; the amount of steel used in any steel-reinforced component in the target building structure is greater than the minimum amount of steel required for the structure; the amount of masonry blocks used in the target building structure is less than the sum of the volumes of all masonry components; and the amount of composite material used in any composite material-reinforced component in the target building structure is greater than the minimum amount of composite material required for the structure.

[0013] Furthermore, the specific steps for training the material usage prediction model include: using the material usage dataset as a training set, and based on prior knowledge, calculating the material usage loss using a first preset formula and calculating the prior knowledge loss using a second preset formula, with the training objective being to minimize the sum of the material usage loss and the prior knowledge loss, thereby completing the training of the material usage prediction model.

[0014] Further, the specific steps for constructing the material usage dataset include: acquiring relevant historical design conditions, historical building design schemes, and historical structural design schemes based on historical engineering projects; constructing a historical structural scheme dataset using the historical design conditions and historical building design schemes as input data and the historical structural design schemes as real label data; training a pre-constructed structural scheme design model using the historical structural scheme dataset as a training set, with the training objective being to minimize the difference between the structural design schemes output by the structural scheme design model and their corresponding real label data; inputting the historical building design schemes and a predetermined number of randomly generated augmented design conditions into the trained structural scheme design model to obtain the predetermined number of augmented structural design schemes; calculating the corresponding material usage data based on the constructed structural parametric modeling and analysis process, according to the augmented design conditions and the augmented structural design schemes; and constructing the material usage dataset using the augmented structural design schemes and augmented design conditions as input data and the material usage data as real label data. The structural parametric modeling and analysis process includes: reading modeling information, creating a structural model, conducting structural design, and outputting material usage.

[0015] Furthermore, the first preset formula is as follows:

[0016] L M =MSELoss(Q adjust Q real )

[0017] Q adjust =Q orig -λ·ReLU(Q orig -Q max )+λ·ReLU(Q min -Q orig )

[0018] Among them, L M Q represents the material usage loss. adjust To adjust the material usage, Q real Q represents the actual material usage. orig Q represents the initial material usage. max Q represents the maximum expected material usage. minλ represents the expected minimum material usage, λ is the material loss correction weight, ReLU(·) is the activation function, and MSELoss(·) is the mean squared error loss function.

[0019] Furthermore, the second preset formula is as follows:

[0020] L K =μ·Sum(ReLU(Q) orig -Q max ))+μ·Sum(ReLU(Q min -Q orig ))

[0021] Among them, L K For prior knowledge loss, Q orig Q represents the initial material usage. max Q represents the maximum expected material usage. min denoted as the expected minimum material usage, μ is the prior loss correction weight, ReLU(·) is the activation function, and Sum(·) is the summation function.

[0022] Secondly, the present invention also provides a building structure material usage prediction device, comprising: a design condition acquisition module for acquiring design conditions of a target building structure, the design conditions including seismic design conditions, wind resistance design conditions, and overall structural geometric design conditions; a structural design scheme acquisition module for acquiring structural design schemes of the target building structure, the structural design schemes including the spatial positions and geometric dimensions of walls, columns, beams, slabs, and supports; and a material usage prediction module for inputting the design conditions and the structural design schemes into a pre-trained material usage prediction model to obtain the material usage of various materials in the target building structure; wherein the material usage prediction model is obtained by embedding prior knowledge, using a material usage dataset as a training set, and optimizing the training with the goal of minimizing the sum of material usage loss and prior knowledge loss.

[0023] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the building structure material usage prediction method as described above.

[0024] Fourthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for predicting the amount of building structural materials as described above.

[0025] The method for predicting the material usage of building structures provided by this invention obtains the design conditions and structural design scheme of the target building structure, and inputs these conditions and schemes into a material usage prediction model to obtain the material usage of various materials in the target building structure. This method overcomes the shortcomings of existing technologies, which suffer from low accuracy due to the inability to consider prior knowledge in building structure-related fields, leading to predicted material usage values ​​that contradict basic engineering principles. This method achieves accurate and efficient prediction of the material usage of the target building structure. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0027] Figure 1 A flowchart illustrating the method for predicting the amount of building structural materials provided by this invention;

[0028] Figure 2 A schematic diagram of the material usage prediction process provided by this invention;

[0029] Figure 3 A schematic diagram illustrating the reasoning process of the material usage prediction model provided by this invention;

[0030] Figure 4 A schematic diagram illustrating the construction process of the material usage dataset provided by this invention;

[0031] Figure 5 A schematic diagram of the overall process for predicting the amount of building structural materials used according to the present invention;

[0032] Figure 6 A schematic diagram of the building structure material usage prediction device provided by the present invention;

[0033] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0035] Figure 1 A flowchart illustrating the method for predicting the amount of building structural materials provided by this invention is shown. Figure 1 As shown, the method includes:

[0036] S110, Obtain the design conditions of the target building structure, which include seismic design conditions, wind resistance design conditions, and overall structural geometric design conditions;

[0037] S120, Obtain the structural design scheme of the target building structure, which includes the spatial location and geometric dimensions of walls, columns, beams, slabs, and supports;

[0038] S130: Input the structural design scheme and design conditions into the pre-trained material usage prediction model to obtain the material usage of various materials in the target building structure.

[0039] Understandably, to predict the material usage of various materials in a target building structure, it is first necessary to obtain the structural design scheme and design conditions of the target building structure. The target building structure is the building structure for which the usage of various materials needs to be predicted.

[0040] In step S110, the design conditions of the target building structure are obtained, including but not limited to the seismic design conditions, wind resistance design conditions, and overall geometric design conditions of the target building structure.

[0041] In step S120, based on the design conditions of the target building structure obtained in step S110, the structural design scheme of the target building structure is obtained. The structural design scheme includes, but is not limited to, the spatial positions of walls, columns, beams, slabs and supports, as well as the geometric dimensions of walls, columns, beams, slabs and supports, such as the length, width and height of the corresponding structures.

[0042] Based on obtaining the design conditions and structural design scheme of the target building structure in steps S110 and S120 respectively, step S130 is further executed to input the structural design scheme and design conditions as input data into the pre-trained material usage prediction model to obtain the corresponding output, namely the material usage of various materials in the target building structure.

[0043] It should be noted that existing methods for predicting material usage cannot take into account prior knowledge in the field of building structure, which leads to predicted material usage that contradicts basic common sense among engineers and results in low prediction accuracy.

[0044] In view of this, the material usage prediction model provided in this embodiment will embed prior knowledge to overcome the above-mentioned defects, and will be optimized and trained using the material usage dataset as the training set with the goal of minimizing the sum of material usage loss and prior knowledge loss, so as to ensure the prediction accuracy of the usage of various materials in the target building structure.

[0045] The prior knowledge includes: 1) The amount of concrete used in the target building structure is less than the sum of the volumes of all concrete components; 2) The amount of steel used in any steel-reinforced component in the target building structure is greater than the minimum amount of steel required for the structure; 3) The amount of masonry blocks used in the target building structure is less than the sum of the volumes of all masonry components; 4) The amount of composite material used in any composite material-reinforced component in the target building structure is greater than the minimum amount of composite material required for the structure.

[0046] The material usage of various materials in the target building structure includes, but is not limited to, the usage of concrete, steel, masonry blocks, and composite materials.

[0047] In this embodiment, the design conditions and structural design scheme of the target building structure are obtained, and these conditions and schemes are used as input data into a material usage prediction model to obtain the material usage of various materials in the target building structure. This method overcomes the shortcomings of existing technologies, which suffer from low accuracy due to the inability to consider prior knowledge in building structure-related fields, leading to predicted material usage values ​​that contradict basic engineering principles. This approach achieves accurate and efficient prediction of material usage for the target building structure.

[0048] Based on the above embodiments, further, Figure 2 A schematic diagram of the material usage prediction process provided by the present invention is shown to further explain step S130.

[0049] like Figure 2 As shown, the design conditions and structural design scheme are input into a pre-trained material usage prediction model to obtain the material usage of various materials in the target building structure, including:

[0050] S131, Extract the first high-dimensional feature from the structural design scheme;

[0051] S132, Extract the second high-dimensional feature from the design conditions;

[0052] S133, the first high-dimensional feature and the second high-dimensional feature are spliced ​​together to obtain the fused high-dimensional feature;

[0053] S134 maps the fused high-dimensional features to material usage, thus obtaining the material usage of various materials in the target building structure.

[0054] It is understandable that the design conditions and structural design schemes of the target building structure are processed by a pre-trained material usage prediction model. Specifically, the first high-dimensional features in the structural design scheme and the second high-dimensional features in the design conditions are extracted respectively.

[0055] Furthermore, the extracted first and second high-dimensional features are spliced ​​together to obtain fused high-dimensional features, which are then mapped to material usage to obtain the material usage of various materials in the target building structure.

[0056] Preferably, in one embodiment, the material usage prediction model includes three sub-models: a first sub-model, a second sub-model, and a third sub-model.

[0057] Based on the above, in a specific embodiment, the first high-dimensional feature in the structural design scheme is extracted. Specifically, the structural design scheme is represented as a structural design graph, which uses the nodes of the structural components as graph nodes, the node coordinates of the structural components as node features, the structural components as graph edges, and the geometric dimensions of the structural components as edge features. The structural design graph is input into the first sub-model to obtain the first high-dimensional feature.

[0058] Among these are the geometric dimensions of structural components, such as their length, height, and width.

[0059] In detail, after the structural design graph is input into the first sub-model, the first sub-model aggregates the node features and edge features of the structural design graph through a graph neural network layer to obtain updated edge features.

[0060] Furthermore, the first sub-model maps the updated edge features through a fully connected neural network layer and sums all the updated edge features to obtain the corresponding high-dimensional feature vector, i.e., the first high-dimensional feature.

[0061] In another specific embodiment, the second high-dimensional feature in the design conditions is extracted. Specifically, the design conditions are represented as design feature vectors, which include seismic load, wind load, building height, and building width. The design feature vectors are then input into the second sub-model to obtain the second high-dimensional feature.

[0062] In detail, the design conditions include seismic fortification intensity, site conditions, and structural height. The seismic fortification intensity is represented by the maximum value of the seismic influence coefficient α. max The site conditions are characterized as the site characteristic period T. g The structural height is represented by the number of building floors n. s This forms a global feature vector of length 3, which is the design feature vector.

[0063] Furthermore, the design feature vector is input into the second sub-model, which maps the design feature vector into a corresponding high-dimensional feature vector, i.e., the second high-dimensional feature, through a fully connected neural network.

[0064] In another specific embodiment, the fused high-dimensional features are mapped to material usage to obtain the material usage of various materials in the target building structure. Specifically, the fused high-dimensional features are input into the third sub-model to obtain the material usage of various materials in the target building structure.

[0065] In this way, the fused high-dimensional features are input into the third sub-model, which maps the fused high-dimensional features into a material usage vector through a fully connected neural network, that is, the material usage of various materials in the target building structure.

[0066] Among them, the material usage vector, for example, has a length of 6, which represents the concrete and steel reinforcement usage of the shear wall, the concrete and steel reinforcement usage of the beam, and the concrete and steel reinforcement usage of the floor slab.

[0067] Other, Figure 3 A schematic diagram illustrating the reasoning of the material usage prediction model provided by this invention is shown.

[0068] like Figure 3 As shown, the material usage prediction model in this embodiment includes three sub-neural network models, namely the first sub-model, the second sub-model, and the third sub-model mentioned above.

[0069] In the first sub-model, the node features and edge features of the structural design graph are aggregated through a graph neural network layer to obtain updated edge features. The updated edge features are then mapped through a fully connected neural network layer, and all updated edge features are summed to obtain the first high-dimensional feature.

[0070] In the second sub-model, the design feature vector formed by the maximum value of the seismic influence coefficient, the site characteristic period, and the number of building stories is input into the second sub-model to obtain the corresponding second high-dimensional feature.

[0071] In the third sub-model, the fused high-dimensional feature obtained by fusing the first and second high-dimensional features is mapped to a material usage vector, which represents the material usage of various materials in the target building structure.

[0072] In this embodiment, by extracting the first high-dimensional features from the structural design scheme and the second high-dimensional features from the design conditions, and then splicing the first and second high-dimensional features to obtain the fused high-dimensional features, the fused high-dimensional features are then mapped to material usage to obtain the material usage of various materials in the target building structure. This method can achieve accurate and efficient prediction of the material usage of the target building structure.

[0073] Based on the above embodiments, the method further includes training the material usage prediction model. Specifically, the material usage dataset is used as the training set, and based on prior knowledge, the material usage loss is calculated using a first preset formula, and the prior knowledge loss is calculated using a second preset formula. The training objective is to minimize the sum of the material usage loss and the prior knowledge loss, thereby completing the training of the material usage prediction model.

[0074] Understandably, a knowledge-embedded neural network, namely a material usage prediction model, was pre-constructed, which includes three sub-neural network models: the first sub-model, the second sub-model, and the third sub-model mentioned above.

[0075] Before training the material usage prediction model, it is first necessary to construct a material usage dataset. Specifically, Figure 4 A schematic diagram illustrating the construction process of the material usage dataset provided by this invention is shown.

[0076] like Figure 4 As shown, the specific steps for constructing the material usage dataset include:

[0077] S410, based on historical engineering projects, obtain the corresponding historical design conditions, historical building design schemes and historical structural design schemes;

[0078] S420 uses historical design conditions and historical building design schemes as input data and historical structural design schemes as real label data to construct a historical structural scheme dataset.

[0079] It is understandable that we collect historical engineering projects with different building structures and obtain the corresponding historical design conditions, historical building design schemes, and historical structural design schemes.

[0080] For example, in one specific embodiment, 430 actual engineering cases of reinforced concrete shear wall structures were collected, including historical design conditions, historical building design schemes and corresponding historical structural design schemes.

[0081] Among them, historical design conditions such as seismic fortification intensity, site conditions and structural height, historical building design schemes such as the plan layout of partition walls and doors and windows, and historical structural design schemes such as the plan layout of shear walls and beams.

[0082] Historical design conditions, historical building design schemes, and historical structural design schemes constitute the structural design scheme dataset.

[0083] S430 uses the historical structural scheme dataset as the training set to train a pre-built structural scheme design model, with the training objective being to minimize the difference between the structural design schemes output by the structural scheme design model and their corresponding real label data.

[0084] Based on the historical structural scheme dataset constructed in step S420, the structural scheme design model is further trained using the historical structural scheme dataset. During the training process, the historical design conditions and historical building design schemes are used as input data and fed into the pre-constructed adversarial generative network, i.e., the structural scheme design model. The historical structural design schemes are used as real label data, and the structural scheme design model is trained according to the preset loss function.

[0085] The preset loss function is the difference between the structural design scheme output by the structural scheme design model and the corresponding real label data. After training, the model parameters of the current structural scheme design model are saved for subsequent augmentation of historical data, thereby obtaining a larger material usage dataset and better completing the training of the material usage prediction model.

[0086] S440: Input the historical building design scheme and a preset number of augmented design conditions into the trained structural design model to obtain a preset number of augmented structural design schemes.

[0087] Understandably, for each historical building design scheme, a combination of design conditions with a set threshold can be randomly generated. Regarding specific combinations, for example, the seismic fortification intensity could be 6 degrees, 7 degrees, or 8 degrees. Another example is the structural height, which could range from 15m to 100m. Yet another example is the site conditions, which could be classified as Class I-IV.

[0088] The threshold can be set according to the actual situation. For example, in a specific embodiment, for each historical building design scheme, 10 combinations of design conditions can be randomly generated to obtain the set threshold and augmented design conditions.

[0089] The preset number can also be set according to the actual situation. For example, in another specific embodiment, 4,300 sets of historical building design schemes and randomly generated augmented design conditions are input into a pre-trained structural scheme design model to obtain the corresponding 4,300 sets of historical structural design schemes, that is, 4,300 sets of augmented structural design schemes.

[0090] S450, based on the constructed structural parametric modeling and analysis process, calculates the corresponding material usage data according to the augmented design conditions and augmented structural design scheme.

[0091] It is easy to understand that, based on obtaining the augmented design conditions and augmented structural design schemes, the next step is to establish a structural parametric modeling and analysis process in the structural design software. This structural parametric modeling and analysis process mainly includes: reading modeling information, creating a structural model, carrying out structural design, and outputting material usage.

[0092] Based on the aforementioned structural parametric modeling and analysis process, after inputting the augmented design conditions and historical building design schemes into the structural scheme design model, the augmented structural design scheme output by the structural scheme design model is read, and the corresponding structural model is established, thereby calculating the corresponding material usage.

[0093] S460 uses augmented structural design schemes and augmented design conditions as input data and material usage data as real label data to construct a material usage dataset.

[0094] It is easy to understand that, based on obtaining the augmented structure design scheme and augmented design conditions in step S440 and calculating the corresponding material usage data in step S450, a material usage dataset can be constructed from the data obtained in the aforementioned steps.

[0095] In the constructed material usage dataset, augmented structural design schemes and augmented design conditions serve as input data, while material usage data represents the true label data. Based on this dataset, 80% (3440 samples) can be used as the training set to train the material usage prediction model, and 20% (860 samples) can be used as the test set to test the model's prediction performance.

[0096] During the training process of the material usage prediction model, the first high-dimensional feature in the historical structural design scheme is extracted using the first sub-model. Specifically, the historical structural design scheme is represented as a structural design map. The structural design map uses the nodes of the structural components as map nodes, the node coordinates of the structural components as node features, the structural components as map edges, and the geometric dimensions of the structural components as edge features. By inputting the structural design map into the first sub-model, the first high-dimensional feature can be obtained.

[0097] The second high-dimensional features in the historical design conditions are extracted using the second sub-model. Specifically, the historical design conditions are represented as design feature vectors, which include seismic loads, wind loads, building height, and building width. The design feature vectors are then input into the second sub-model to obtain the second high-dimensional features.

[0098] The first high-dimensional feature and the second high-dimensional feature are concatenated to form a fused feature vector, which is then input into the third sub-model to output the corresponding material usage vector.

[0099] Finally, based on prior knowledge, the material usage loss is calculated using the first preset formula, and the prior knowledge loss is calculated using the second preset formula. The training objective is to minimize the sum of the material usage loss and the prior knowledge loss, thereby completing the training of the material usage prediction model.

[0100] The first preset formula is as follows:

[0101] L M =MSELoss(Q adjust Q real )

[0102] Q adjust =Q orig -λ·ReLU(Q orig -Q max )+λ·ReLU(Q min -Q orig )

[0103] Among them, L M Q represents the material usage loss. adjust To adjust the material usage, Q real Q represents the actual material usage. orig Q represents the initial material usage. max Q represents the maximum expected material usage. min Let λ represent the expected minimum material usage, 0 ≤ λ ≤ 1, ReLU(·) be the activation function, and MSELoss(·) be the mean squared error loss function. In a specific embodiment, taking the material usage loss of concrete and steel reinforcement during training as an example, the correction formula is as follows:

[0104]

[0105]

[0106] in, and These are the revised concrete and steel reinforcement quantities, respectively. and These are the concrete and steel reinforcement quantities output by the material usage prediction model, respectively. and These are the maximum concrete usage and minimum steel reinforcement usage calculated based on prior knowledge, respectively, and ReLU(·) is the activation function.

[0107] The second preset formula is as follows:

[0108] L K =μ·Sum(ReLU(Q) orig -Q max ))+μ·Sum(ReLU(Qmin -Q orig ))

[0109] Among them, L K For prior knowledge loss, Q orig Q represents the initial material usage. max Q represents the maximum expected material usage. min The expected minimum material usage is given by μ, which is the prior loss correction weight, 0 ≤ μ ≤ 1, ReLU(·) is the activation function, and Sum(·) is the summation function.

[0110] In the above preset formula, and The calculation formulas are as follows:

[0111]

[0112]

[0113] Among them, V wall V beam V slab These are the volumes of individual shear walls, beams, and floor slabs, respectively. These are the steel reinforcement quantities for a single shear wall, beam, and floor slab, calculated according to the minimum reinforcement ratio required by structural requirements.

[0114] Furthermore, regarding The calculation method is as follows: for a single shear wall member, the minimum amount of distributed reinforcement for its edge restraint members and wall distribution reinforcement is shown in Table 1 and Table 2 respectively.

[0115] Table 1 Minimum reinforcement requirements for edge-constrained members

[0116]

[0117]

[0118] Note: A c This represents the cross-sectional area of ​​the edge member.

[0119] Table 2 Minimum Reinforcement Ratio of Distributed Bars in Walls

[0120]

[0121] about The calculation formulas are as follows: for beam members, the minimum reinforcement amounts of longitudinal tensile reinforcement, stirrups and gussets are shown in Tables 3 to 6.

[0122] Table 3 Minimum reinforcement ratio (%) of longitudinal tensile reinforcement in frame beams (span-to-depth ratio greater than 1.5)

[0123]

[0124] Note: f t f is the tensile strength of concrete. y This represents the yield strength of the steel reinforcement.

[0125] Table 4 Minimum reinforcement ratio (%) of longitudinal tensile reinforcement in coupling beams (span-to-depth ratio not greater than 1.5)

[0126]

[0127] Note: f t f is the tensile strength of concrete. y This represents the yield strength of the steel reinforcement.

[0128] Table 5 Minimum Reinforcement Quantity for Stirrups

[0129]

[0130] Table 6 Minimum reinforcement requirements for stirrups

[0131]

[0132] Note: h b d is the beam section height, and d is the diameter of the longitudinal reinforcement.

[0133] about The calculation method, for slab members, specifies the minimum reinforcement ratio (%) of longitudinal tensile reinforcement as 0.20 and 45f. t / f y The larger value in the range.

[0134] In addition, this embodiment also provides test results of testing the material usage prediction model using a 20% material usage dataset.

[0135] Specifically, the test results are shown in Table 7 below:

[0136] Table 7. Accuracy of the Material Usage Prediction Model

[0137]

[0138] In Table 7, APE represents the absolute percentage error.

[0139] In this embodiment, the constructed material usage dataset is used as the training set. Based on prior knowledge, the material usage loss is calculated using a first preset formula, and the prior knowledge loss is calculated using a second preset formula. The training objective is to minimize the sum of the material usage loss and the prior knowledge loss, thereby completing the training of the material usage prediction model. This enables accurate and efficient prediction of the material usage of the target building structure during actual inference.

[0140] in addition, Figure 5 A schematic diagram of the overall process of the method for predicting the amount of building structural materials provided by the present invention is shown.

[0141] like Figure 5 As shown, firstly, a structural design model was constructed. When training the model, historical design conditions and architectural design schemes were used as input data, and historical structural design schemes were used as label data. After training until convergence, the corresponding model parameters were saved.

[0142] Then, based on historical architectural design schemes and randomly combined design conditions, a large number of corresponding structural design schemes can be generated through the structural scheme design model. Subsequently, based on the constructed structural parametric modeling and analysis process, the corresponding material usage list can be calculated, thereby constructing a material usage dataset.

[0143] Next, using historical design conditions and structural design schemes as input data, and material usage as the real-world label data, the material usage prediction model is trained. Based on... Figure 5 It can be seen that knowledge embedding has been carried out, that is, prior knowledge has been embedded to overcome the defects existing in the current technology.

[0144] After training the material usage prediction model to convergence, save the model parameters.

[0145] Finally, the trained material usage prediction model is used for practical reasoning. That is, using the design conditions and structural design scheme of the target building structure as input, the corresponding output is obtained, namely the material usage of various materials in the target building structure, for example... Figure 5 The amount of concrete used was 2739.3 cubic meters, and the amount of steel reinforcement was 271.0 tons.

[0146] Figure 6 A schematic diagram of the building structure material usage prediction device provided by the present invention is shown. Figure 6 As shown, the device includes:

[0147] The design condition acquisition module 610 is used to acquire the design conditions of the target building structure, including seismic design conditions, wind resistance design conditions, and overall structural geometric design conditions. The structural design scheme acquisition module 620 is used to acquire the structural design scheme of the target building structure, including the spatial location and geometric dimensions of walls, columns, beams, slabs, and supports. The material usage prediction module 630 is used to input the design conditions and the structural design scheme into a pre-trained material usage prediction model to obtain the material usage of various materials in the target building structure. The material usage prediction model is obtained by embedding prior knowledge, using the material usage dataset as the training set, and optimizing the training with the goal of minimizing the sum of material usage loss and prior knowledge loss.

[0148] In this embodiment, the design conditions acquisition module 610 and the structural design scheme acquisition module 620 acquire the design conditions and structural design scheme of the target building structure, respectively. The material usage prediction module 630 inputs the acquired design conditions and structural design scheme as input data into the material usage prediction model to obtain the material usage of various materials in the target building structure. By embedding prior knowledge into the material usage prediction model, this device overcomes the shortcomings of existing technologies, which cannot consider prior knowledge in the field of building structure, resulting in material usage prediction values ​​that violate basic engineering common sense and have low prediction accuracy. This achieves accurate and efficient prediction of the material usage of the target building structure.

[0149] It should be noted that the building structure material usage prediction device provided in this embodiment can be referred to in correspondence with the building structure material usage prediction method described above, and will not be repeated here.

[0150] Figure 7 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 7 As shown, the electronic device may include a processor 710, a communication interface 720, a memory 730, and a communication bus 740, wherein the processor 710, the communication interface 720, and the memory 730 communicate with each other through the communication bus 740. The processor 710 can call logical instructions in the memory 730 to execute a method for predicting the material usage of a building structure. This method includes: obtaining the design conditions of the target building structure, including seismic design conditions, wind resistance design conditions, and overall structural geometric design conditions; obtaining the structural design scheme of the target building structure, including the spatial location and geometric dimensions of walls, columns, beams, slabs, and supports; inputting the design conditions and the structural design scheme into a pre-trained material usage prediction model to obtain the material usage of various materials in the target building structure; wherein the material usage prediction model is obtained by embedding prior knowledge, using a material usage dataset as the training set, and optimizing the training to minimize the sum of material usage loss and prior knowledge loss.

[0151] Furthermore, the logical instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0152] On the other hand, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the method for predicting the material usage of building structures provided by the above methods. The method includes: obtaining design conditions for a target building structure, the design conditions including seismic design conditions, wind resistance design conditions, and overall structural geometric design conditions; obtaining a structural design scheme for the target building structure, the structural design scheme including the spatial positions and geometric dimensions of walls, columns, beams, slabs, and supports; inputting the design conditions and the structural design scheme into a pre-trained material usage prediction model to obtain the material usage of various materials in the target building structure; wherein the material usage prediction model is obtained by embedding prior knowledge, using a material usage dataset as a training set, and optimizing the training with the goal of minimizing the sum of material usage loss and prior knowledge loss.

[0153] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0154] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0155] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the amount of building structural materials used, characterized in that, include: Obtain the design conditions for the target building structure, including seismic design conditions, wind resistance design conditions, and overall structural geometric design conditions; Obtain the structural design scheme of the target building structure, wherein the structural design scheme includes the spatial location and geometric dimensions of walls, columns, beams, slabs, and supports; The design conditions and the structural design scheme are input into a pre-trained material usage prediction model to obtain the material usage of various materials in the target building structure. The material usage prediction model is obtained by embedding prior knowledge, using the material usage dataset as the training set, and optimizing the training with the goal of minimizing the sum of material usage loss and prior knowledge loss. The prior knowledge includes: The amount of concrete used in the target building structure is less than the sum of the volumes of all concrete components. The amount of steel used in any steel reinforcement component of the target building structure is greater than the minimum amount of steel required for the structure. The amount of masonry blocks used in the target building structure is less than the sum of the volumes of all masonry components; The amount of composite material used in any composite material reinforced component in the target building structure is greater than the minimum amount of composite material required for the structure. The specific steps for training the material usage prediction model include: The material usage dataset is used as the training set. Based on prior knowledge, the material usage loss is calculated using a first preset formula, and the prior knowledge loss is calculated using a second preset formula. The training objective is to minimize the sum of the material usage loss and the prior knowledge loss, thereby completing the training of the material usage prediction model. The first preset formula is as follows: L M =MSELoss(Q adjust ,Q real ) Q adjust =Q orig -λ·ReLU(Q orig -Q max )+λ·ReLU(Q min -Q orig ) Among them, L M Q represents the material usage loss. adjust To adjust the material usage, Q real Q represents the actual amount of material used. orig Q represents the initial material usage. max Q represents the maximum expected material usage. min λ represents the expected minimum material usage, λ is the material loss correction weight, ReLU(·) is the activation function, and MSELoss(·) is the mean squared error loss function. The second preset formula is as follows: 50 K =μ·Sum(ReLU(Q orig -Q mmx ))+μ·Sum(ReLU(Q min -Q orig )) Among them, L K For prior knowledge loss, Q orig Q represents the initial material usage. max Q represents the maximum expected material usage. min denoted as the expected minimum material usage, μ is the prior loss correction weight, ReLU(·) is the activation function, and Sum(·) is the summation function.

2. The method for predicting the amount of building structural materials according to claim 1, characterized in that, The process involves inputting the design conditions and the structural design scheme into a pre-trained material usage prediction model to obtain the material usage of various materials in the target building structure, including: Extract the first high-dimensional feature from the structural design scheme; Extract the second high-dimensional feature from the design conditions; By concatenating the first high-dimensional feature and the second high-dimensional feature, a fused high-dimensional feature is obtained; The fused high-dimensional features are mapped to material usage to obtain the material usage of various materials in the target building structure.

3. The method for predicting the amount of building structural materials according to claim 2, characterized in that, The material usage prediction model includes a first sub-model, a second sub-model, and a third sub-model; The extraction of the first high-dimensional feature from the structural design scheme includes: The structural design scheme is represented as a structural design graph, wherein the structural design graph uses the nodes of the structural components as graph nodes, the node coordinates of the structural components as node features, the structural components as graph edges, and the geometric dimensions of the structural components as edge features. The structural design map is input into the first sub-model to obtain the first high-dimensional feature.

4. The method for predicting the amount of building structural materials according to claim 2, characterized in that, The material usage prediction model includes a first sub-model, a second sub-model, and a third sub-model; The extraction of the second high-dimensional feature from the design conditions includes: The design conditions are characterized as a design feature vector, which includes seismic load, wind load, building height, and building width. The design feature vector is input into the second sub-model to obtain the second high-dimensional feature.

5. The method for predicting the amount of building structural materials according to claim 2, characterized in that, The material usage prediction model includes a first sub-model, a second sub-model, and a third sub-model; The process of mapping the fused high-dimensional features to material usage to obtain the material usage of various materials in the target building structure includes: The fused high-dimensional features are input into the third sub-model to obtain the material usage of various materials in the target building structure.

6. The method for predicting the amount of building structural materials according to claim 1, characterized in that, The specific steps for constructing the material usage dataset include: Based on historical engineering projects, obtain the corresponding historical design conditions, historical building design schemes, and historical structural design schemes; Using the historical design conditions and the historical building design schemes as input data, and the historical structural design schemes as real label data, a historical structural scheme dataset is constructed. The historical structural scheme dataset is used as the training set to train a pre-built structural scheme design model, with the training objective being to minimize the difference between the structural design schemes output by the structural scheme design model and their corresponding real label data. The historical building design scheme and a preset number of augmented design conditions are input into the trained structural scheme design model to obtain the preset number of augmented structural design schemes. Based on the constructed structural parametric modeling and analysis process, the corresponding material consumption data are calculated according to the augmented design conditions and the augmented structural design scheme. Using the augmented structure design scheme and the augmented design conditions as input data, and the material usage data as real label data, the material usage dataset is constructed. The structural parametric modeling and analysis process includes: reading modeling information, creating a structural model, carrying out structural design, and outputting material usage.

7. A device for predicting the amount of building structural materials used, characterized in that, include: The design conditions acquisition module is used to acquire the design conditions of the target building structure, including seismic design conditions, wind resistance design conditions, and overall structural geometric design conditions. The structural design scheme acquisition module is used to acquire the structural design scheme of the target building structure, which includes the spatial location and geometric dimensions of walls, columns, beams, slabs, and supports. The material usage prediction module is used to input the design conditions and the structural design scheme into a pre-trained material usage prediction model to obtain the material usage of various materials in the target building structure. The material usage prediction model is obtained by embedding prior knowledge, using the material usage dataset as the training set, and optimizing the training with the goal of minimizing the sum of material usage loss and prior knowledge loss. The prior knowledge includes: The amount of concrete used in the target building structure is less than the sum of the volumes of all concrete components. The amount of steel used in any steel reinforcement component of the target building structure is greater than the minimum amount of steel required for the structure. The amount of masonry blocks used in the target building structure is less than the sum of the volumes of all masonry components; The amount of composite material used in any composite material reinforced component in the target building structure is greater than the minimum amount of composite material required for the structure. The specific steps for training the material usage prediction model include: The material usage dataset is used as the training set. Based on prior knowledge, the material usage loss is calculated using a first preset formula, and the prior knowledge loss is calculated using a second preset formula. The training objective is to minimize the sum of the material usage loss and the prior knowledge loss, thereby completing the training of the material usage prediction model. The first preset formula is as follows: L M =MSELoss(Q adjust ,Q real ) Q adjust =Q orig -λ·ReLU(Q orig -Q max )+λ·ReLU(Q min -Q orig ) Among them, L M Q represents the material usage loss. adjust To adjust the material usage, Q real Q represents the actual amount of material used. orig Q represents the initial material usage. max Q represents the maximum expected material usage. min λ represents the expected minimum material usage, λ is the material loss correction weight, ReLU(·) is the activation function, and MSELoss(·) is the mean squared error loss function. The second preset formula is as follows: 50 K =μ·Sum(ReLU(Q orig -Q max ))+μ·Sum(ReLU(Q min -Q orig )) Among them, L K For prior knowledge loss, Q orig Q represents the initial material usage. max Q represents the maximum expected material usage. min denoted as the expected minimum material usage, μ is the prior loss correction weight, ReLU(·) is the activation function, and Sum(·) is the summation function.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method for predicting the amount of building structural materials as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for predicting the amount of building structural materials as described in any one of claims 1 to 6.

Images