A pile foundation recommendation and dataset design method based on a deep neural network

Abstract

The application relates to the technical field of neural networks and bridges, and discloses a pile foundation recommendation and dataset design method based on a deep neural network, which comprises the following steps: acquiring characteristic parameters for determining pile foundation classification; inputting the characteristic parameters into a neural network model of a group pile foundation recommendation algorithm to recommend a pile foundation configuration type, so that a recommended pile foundation configuration scheme is obtained; integrating the characteristic parameters and the recommended pile foundation configuration scheme into a database to form sample data records; and adding the sample data records to an original dataset to obtain a new dataset. According to the characteristic parameters for determining pile foundation classification, the neural network model of the group pile foundation recommendation algorithm can recommend a pile foundation configuration scheme, and the dataset can be updated according to the recommended pile foundation configuration scheme, so that the problem that the current data sample size is limited and the accuracy of the group pile foundation neural network is not high is solved, the requirement for the sample size is low, and the model recommendation accuracy is high.

Description

Technical Field

[0001] The present invention relates to the field of neural networks and bridge technology, and in particular to a pile foundation recommendation and data set design method based on deep neural networks. Background Art

[0002] With the development of artificial intelligence (AI), scientists have proposed various neural networks, including LeNet, AlexNet, VGG, ResNet, and GoogLeNet. These neural networks typically handle classification problems, identifying objects such as people, cats, and dogs. Depending on the network width, these problems can be binary or multi-class. As AI penetrates various industries, more and more sectors are realizing digital transformation through the empowerment of intelligent technologies. However, the greatest challenge currently lies in applying intelligent technologies to practical industry applications. Existing neural networks cannot fully address practical challenges encountered during industry implementation, such as cold start issues caused by small or even missing data sample sets and complex business logic.

[0003] Taking the design of pile group foundations for high-speed railway bridges as an example, if a neural network is used to solve the classification problem of pile group foundations, a "combination explosion" problem will occur. There are more than 100 classification results, and the currently accumulated data sample size is limited, resulting in the accuracy of the pile group foundation neural network not being very ideal. It is necessary to propose a new neural network structure to solve the classification "combination explosion" problem. Summary of the Invention

[0004] The main purpose of the present invention is to provide a pile foundation recommendation and data set design method based on deep neural network, aiming to solve the technical problem that the current data sample size in the existing technology is limited, resulting in low accuracy of the pile group foundation neural network.

[0005] To achieve the above objectives, the present invention provides a pile foundation recommendation and dataset design method based on a deep neural network, the method comprising the following steps:

[0006] Obtain characteristic parameters that determine pile foundation classification;

[0007] Inputting the characteristic parameters into a neural network model of a pile group foundation recommendation algorithm to recommend a pile foundation configuration type, thereby obtaining a recommended pile foundation configuration scheme, wherein the neural network model of the pile group foundation recommendation algorithm is constructed based on a deep neural network, the deep neural network includes an input layer, a hidden layer, and an output layer, the output layer of the deep neural network is composed of a single softmax layer, and the output layer of the neural network model of the pile group foundation recommendation algorithm is composed of multiple softmax layers;

[0008] Incorporating the characteristic parameters and the recommended pile foundation configuration scheme into a database to form a sample data record;

[0009] The sample data records are added to the original data set to obtain a new data set, wherein the new data set is used to continuously optimize and upgrade the neural network model of the pile group foundation recommendation algorithm by re-calling the new data set for model training.

[0010] Optionally, before obtaining the characteristic parameters for determining the pile foundation classification, the method further includes:

[0011] Obtain bridge design plans;

[0012] Determining bridge scheme elements that determine pile foundation configuration schemes based on the bridge design scheme;

[0013] The basic characteristic parameters and the number of parameters are determined according to the bridge scheme elements that determine the pile foundation configuration scheme.

[0014] Optionally, obtaining characteristic parameters for determining pile foundation classification includes:

[0015] Extracting bridge schemes from existing bridge engineering projects to obtain initial characteristic parameters, wherein the initial characteristic parameters include at least one of main beam type, main span length, main beam length, pier height, pier location, and terrain characteristics;

[0016] Determining target feature parameters from the initial feature parameters based on the basic feature parameters and the number of parameters;

[0017] The target characteristic parameters are used as characteristic parameters for determining pile foundation classification.

[0018] Optionally, inputting the characteristic parameters into a neural network model of a pile group foundation recommendation algorithm to recommend a pile foundation configuration type to obtain a recommended pile foundation configuration scheme includes:

[0019] Inputting the characteristic parameters determining pile foundation classification into the input layer of the neural network model of the pile group foundation recommendation algorithm;

[0020] Encoding the characteristic parameters determining the pile foundation classification through the input layer to obtain an encoded input vector;

[0021] Inputting the encoded input vector into the hidden layer of the neural network model of the pile group foundation recommendation algorithm;

[0022] Inputting the encoded input vector into the output layer of the neural network model of the pile group foundation recommendation algorithm through the hidden layer;

[0023] The recommended pile foundation configuration scheme is outputted through the output layer, wherein the recommended pile foundation configuration scheme consists of a pile diameter and the number of piles.

[0024] Optionally, the output layer includes a first softmax layer and a second softmax layer, and outputting the recommended pile foundation configuration solution through the output layer includes:

[0025] Determine the pile diameter through one of the first softmax layer and the second softmax layer;

[0026] Determining the number of pile foundations through the other of the first softmax layer and the second softmax layer;

[0027] A recommended pile foundation configuration scheme is obtained based on the pile foundation diameter and the number of pile foundations.

[0028] Optionally, obtaining a recommended pile foundation configuration scheme based on the pile foundation diameter and the number of pile foundations includes:

[0029] Arrange and combine the pile foundation diameters and the number of pile foundations to obtain a preset number of pile foundation configuration solutions;

[0030] Performing two-hot encoding on the preset number of pile foundation configuration schemes, wherein a two-hot encoding is formed by concatenating two one-hot vectors;

[0031] Identifying the pile diameter by one of the two one-hot vectors to obtain the effective pile diameter;

[0032] Obtain the number of valid foundation piles by identifying the number of foundation piles in the other of the two one-hot vectors;

[0033] A recommended pile foundation configuration scheme is obtained from the preset number of pile foundation configuration schemes according to the effective foundation pile diameter and the effective foundation pile quantity.

[0034] Optionally, before inputting the characteristic parameters into a neural network model of a pile group foundation recommendation algorithm to recommend a pile foundation configuration type and obtaining a recommended pile foundation configuration solution, the method further includes:

[0035] Obtain sample format requirements;

[0036] Based on the sample format requirements, parameters required by the sample format and corresponding pile foundation configuration schemes are collected from existing bridge engineering projects to form a data set;

[0037] Constructing a neural network model of an initial pile group foundation recommendation algorithm according to the parameters required by the sample format, the sample format requirements, and the pile foundation configuration scheme;

[0038] The neural network model of the initial pile group foundation recommendation algorithm is trained using the data set to obtain a neural network model of the target pile group foundation recommendation algorithm.

[0039] In addition, to achieve the above-mentioned purpose, the present invention also proposes a pile foundation recommendation and dataset design device based on a deep neural network, and the pile foundation recommendation and dataset design device based on a deep neural network comprises:

[0040] An acquisition module is used to obtain characteristic parameters that determine pile foundation classification;

[0041] An input module is used to input the characteristic parameters into a neural network model of a pile group foundation recommendation algorithm to recommend a pile foundation configuration type and obtain a recommended pile foundation configuration scheme, wherein the neural network model of the pile group foundation recommendation algorithm is constructed based on a deep neural network, and the deep neural network includes an input layer, a hidden layer, and an output layer, and the output layer of the deep neural network is composed of a single softmax layer, and the output layer of the neural network model of the pile group foundation recommendation algorithm is composed of multiple softmax layers;

[0042] a storage module for integrating the characteristic parameters and the recommended pile foundation configuration scheme into a database to form a sample data record;

[0043] An adding module is used to add the sample data records to the original data set to obtain a new data set, wherein the new data set is used to continuously optimize and upgrade the neural network model of the pile group foundation recommendation algorithm by re-calling the new data set for model training.

[0044] In addition, to achieve the above-mentioned purpose, the present invention also proposes a pile foundation recommendation and dataset design device based on a deep neural network, and the pile foundation recommendation and dataset design device based on a deep neural network includes: a memory, a processor, and a pile foundation recommendation and dataset design program based on a deep neural network stored on the memory and runnable on the processor, and the pile foundation recommendation and dataset design program based on a deep neural network is configured to implement the steps of the pile foundation recommendation and dataset design method based on a deep neural network as described above.

[0045] In addition, to achieve the above-mentioned purpose, the present invention also proposes a storage medium, on which a pile foundation recommendation and dataset design program based on a deep neural network is stored. When the pile foundation recommendation and dataset design program based on a deep neural network is executed by a processor, the steps of the pile foundation recommendation and dataset design method based on a deep neural network as described above are implemented.

[0046] The present invention obtains characteristic parameters that determine pile foundation classification; inputs the characteristic parameters into a neural network model of a pile group foundation recommendation algorithm to recommend pile foundation configuration types, thereby obtaining a recommended pile foundation configuration scheme; merges the characteristic parameters and the recommended pile foundation configuration scheme into a database to form a sample data record; and adds the sample data record to the original data set to obtain a new data set. In this manner, the neural network model of the pile group foundation recommendation algorithm recommends a pile foundation configuration scheme based on the characteristic parameters that determine pile foundation classification, and updates the data set based on the recommended pile foundation configuration scheme. This solves the current problem of low accuracy of the pile group foundation neural network due to limited data sample size, reduces sample size requirements, and achieves high model recommendation accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 Schematic diagram of the structure of a deep neural network-based pile foundation recommendation and data set design device in the hardware operating environment involved in an embodiment of the present invention;

[0048] Figure 2 This is a flow chart of the first embodiment of the pile foundation recommendation and data set design method based on deep neural network of the present invention;

[0049] Figure 3 This is a flow chart of a second embodiment of the method for pile foundation recommendation and data set design based on a deep neural network according to the present invention;

[0050] Figure 4 A schematic diagram of a bridge scheme according to a second embodiment of a pile foundation recommendation and data set design method based on a deep neural network of the present invention;

[0051] Figure 5a-5d A schematic diagram of pile foundation arrangement according to a second embodiment of a pile foundation recommendation and data set design method based on a deep neural network of the present invention;

[0052] Figure 6 This is a flow chart of a third embodiment of the method for pile foundation recommendation and data set design based on a deep neural network according to the present invention;

[0053] Figure 7 A schematic diagram of a traditional neural network according to a third embodiment of the method for pile foundation recommendation and data set design based on a deep neural network of the present invention;

[0054] Figure 8 A schematic diagram of a double-tail neural network according to a third embodiment of the method for pile foundation recommendation and data set design based on a deep neural network of the present invention;

[0055] Figure 9 This is a structural block diagram of the first embodiment of the pile foundation recommendation and dataset design device based on deep neural network of the present invention.

[0056] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0057] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0058] Reference Figure 1 , Figure 1 Schematic diagram of the structure of the equipment for pile foundation recommendation and data set design based on deep neural network in the hardware operating environment involved in the embodiment of the present invention.

[0059] like Figure 1 As shown, the deep neural network-based pile foundation recommendation and dataset design device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display and an input unit such as a keyboard. Optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a wireless fidelity (Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk storage device. The memory 1005 may also be a storage device independent of the processor 1001.

[0060] Those skilled in the art will understand that Figure 1 The structure shown in the figure does not constitute a limitation on the pile foundation recommendation and dataset design device based on deep neural network, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0061] like Figure 1 As shown, the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and a pile foundation recommendation and data set design program based on a deep neural network.

[0062] exist Figure 1In the pile foundation recommendation and dataset design device based on deep neural network shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the pile foundation recommendation and dataset design device based on deep neural network of the present invention can be set in the pile foundation recommendation and dataset design device based on deep neural network, and the pile foundation recommendation and dataset design device based on deep neural network calls the pile foundation recommendation and dataset design program based on deep neural network stored in the memory 1005 through the processor 1001, and executes the pile foundation recommendation and dataset design method based on deep neural network provided by the embodiment of the present invention.

[0063] The embodiment of the present invention provides a pile foundation recommendation and data set design method based on deep neural network, referring to Figure 2 , Figure 2 This is a flow chart of the first embodiment of the pile foundation recommendation and dataset design method based on deep neural network of the present invention.

[0064] In this embodiment, the pile foundation recommendation and data set design method based on deep neural network includes the following steps:

[0065] Step S10: Acquire characteristic parameters that determine pile foundation classification.

[0066] It should be noted that the executor of this embodiment is a pile foundation recommendation and dataset design device based on a deep neural network, and it can also be other devices that can achieve the same or similar functions. This example does not limit this. This embodiment is explained using a pile foundation recommendation and dataset design device based on a deep neural network as an example.

[0067] The characteristic parameters that determine the pile foundation classification may include main beam type, main beam length, main span length, pier height, pier position, etc., which are not specifically limited in this embodiment.

[0068] Step S20: Input the characteristic parameters into the neural network model of the pile group foundation recommendation algorithm to recommend the pile foundation configuration type and obtain a recommended pile foundation configuration scheme, wherein the neural network model of the pile group foundation recommendation algorithm is constructed based on a deep neural network, and the deep neural network includes an input layer, a hidden layer and an output layer. The output layer of the deep neural network is composed of a single softmax layer, and the output layer of the neural network model of the pile group foundation recommendation algorithm is composed of multiple softmax layers.

[0069] It should be noted that the neural network model of the group pile foundation recommendation algorithm is constructed based on a deep neural network. The deep neural network includes an input layer, a hidden layer and an output layer. The output layer of the deep neural network is composed of a single softmax layer. Multiple softmax layers are added in parallel to the output layer of the deep neural network, and a multi-tail deep neural network model is obtained. For example, two softmax layers are added to obtain a double-tail deep neural network model, that is, the neural network model of the group pile foundation recommendation algorithm. This embodiment does not impose any specific restrictions on this.

[0070] It is understandable that deep learning frameworks such as pytorch and tensorflow can be used to define the neural network model of the pile group foundation recommendation algorithm.

[0071] In the specific implementation, in order to solve the problem of combined classification, this embodiment transforms the single softmax layer of the traditional deep neural network and adds two softmax layers in parallel to form a two-tail neural network (Two-Tail Net, TTNet). This allows multiple elements of the combined classification to be recommended simultaneously to form a recommendation plan, and adjusts the classification scale of the multiplication of multiple classification elements to the classification scale of the addition of multiple classification elements, greatly reducing the number of parameters of the neural network and thereby reducing the requirements for the sample size.

[0072] Step S30: Incorporating the characteristic parameters and the recommended pile foundation configuration scheme into a database to form a sample data record.

[0073] It should be noted that before the recommended pile foundation configuration scheme is incorporated into the database, it is reviewed based on experience to determine whether the pile group scheme meets the project requirements. If not, the recommended pile foundation configuration scheme is optimized according to the review opinions to meet the project requirements.

[0074] It is understandable that characteristic parameters such as main beam type, main span length, main beam length, pier height, pier position, terrain characteristics, etc. are incorporated into the database with the recommended pile foundation configuration plan after review and optimization to form a new sample record.

[0075] Step S40: adding the sample data records to the original data set to obtain a new data set, wherein the new data set is used to continuously optimize and upgrade the neural network model of the pile group foundation recommendation algorithm by re-calling the new data set for model training.

[0076] It should be noted that with the design of new pile foundation configuration schemes for bridges, a large number of new sample data records are added to the original data set to obtain a new data set. The new data set is re-called for model training, and the model is continuously optimized and upgraded to improve the accuracy of pile foundation configuration scheme recommendations.

[0077] This embodiment obtains characteristic parameters that determine pile foundation classification; inputs the characteristic parameters into the neural network model of the pile group foundation recommendation algorithm to recommend pile foundation configuration types, thereby obtaining a recommended pile foundation configuration scheme; merges the characteristic parameters and the recommended pile foundation configuration scheme into a database to form a sample data record; and adds the sample data record to the original data set to obtain a new data set. In this manner, the neural network model of the pile group foundation recommendation algorithm recommends a pile foundation configuration scheme based on the characteristic parameters that determine pile foundation classification, and updates the data set based on the recommended pile foundation configuration scheme. This solves the current problem of low accuracy of the pile group foundation neural network due to limited data sample size, reduces sample size requirements, and increases model recommendation accuracy.

[0078] refer to Figure 3 , Figure 3 This is a flow chart of the second embodiment of the pile foundation recommendation and dataset design method based on deep neural network of the present invention.

[0079] Based on the first embodiment described above, the pile foundation recommendation and data set design method based on a deep neural network in this embodiment further includes, before step S10:

[0080] Step S01: Obtain a bridge design plan.

[0081] It should be noted that the bridge design scheme includes the main beam scheme, pier scheme, foundation scheme, etc.

[0082] like Figure 4 As shown, Figure 4 This is a schematic diagram of a bridge scheme for the pile foundation recommendation and dataset design method based on a deep neural network in this embodiment, where the main beam type is a continuous rigid frame arch.

[0083] Step S02: Determine bridge scheme elements that determine the pile foundation configuration scheme based on the bridge design scheme.

[0084] It should be noted that this embodiment is described using the foundation scheme as an example. The foundation scheme includes pile foundation, open-cut foundation, and well foundation. For example, if the foundation scheme is a pile foundation and the pile foundation is mainly a group pile foundation, then the bridge scheme elements that affect the group pile foundation scheme can be obtained.

[0085] Step S03: determining basic characteristic parameters and parameter quantities based on the bridge scheme elements that determine the pile foundation configuration scheme.

[0086] It should be noted that the number of parameters in the input layer of the neural network model of the pile group foundation recommendation algorithm is consistent with the number of parameters determined by the bridge scheme elements that determine the pile foundation configuration scheme.

[0087] Furthermore, the step S10 includes: extracting bridge schemes in existing bridge engineering projects to obtain initial characteristic parameters, wherein the initial characteristic parameters include at least one of main beam type, main span length, main beam length, pier height, pier position and terrain characteristics; determining target characteristic parameters from the initial characteristic parameters based on the basic characteristic parameters and the number of parameters; and using the target characteristic parameters as characteristic parameters for determining pile foundation classification.

[0088] It should be noted that the initial characteristic parameters are determined based on the bridge scheme in the existing bridge engineering project, which may include one or more of the main beam type, main span length, main beam length, pier height, pier location and terrain characteristics. This embodiment does not impose specific restrictions on this.

[0089] It is understood that main beam types may include simply supported beams, continuous beams, continuous rigid frames, continuous beam arches, continuous rigid frame arches, low-tower cable-stayed bridges, arch bridges, and cable-stayed bridges, and this embodiment does not impose specific restrictions on this. The main span length refers to the characteristic span of a specific beam type. For simply supported beams, the main span is the calculated span, generally referring to the theoretical length between supports. For continuous structures such as continuous beams, continuous rigid frames, low-tower cable-stayed bridges, continuous beam arches, and continuous rigid frame arches, particularly three-span continuous structures, the main span refers to the theoretical calculated length of the middle span, which is also the theoretical length of the longest span. For other bridge types, it generally refers to the theoretical length of the longest span. The main beam length is the continuous length of the bridge deck. Pier locations may include middle piers, side piers, and secondary side piers, and this embodiment does not impose specific restrictions on this. Topographic features are classified according to bridge business rules. Based on the characteristics of the bridge plan layout, they can be divided into conventional areas, plains, difficult mountainous areas, eastern coastal areas, and densely populated urban areas, and this embodiment does not impose specific restrictions on this.

[0090] In a specific implementation, target characteristic parameters are selected from the initial characteristic parameters according to the basic characteristic parameters and the number of parameters determined by the bridge scheme elements that determine the pile foundation configuration scheme, and the target characteristic parameters are used as characteristic parameters for determining the pile foundation classification.

[0091] Furthermore, the inputting of the characteristic parameters into the neural network model of the pile group foundation recommendation algorithm to recommend pile foundation configuration types and obtain a recommended pile foundation configuration scheme includes: inputting the characteristic parameters that determine the pile foundation classification into the input layer of the neural network model of the pile group foundation recommendation algorithm; encoding the characteristic parameters that determine the pile foundation classification through the input layer to obtain an encoded input vector; inputting the encoded input vector into the hidden layer of the neural network model of the pile group foundation recommendation algorithm; inputting the encoded input vector into the output layer of the neural network model of the pile group foundation recommendation algorithm through the hidden layer; and outputting a recommended pile foundation configuration scheme through the output layer, wherein the recommended pile foundation configuration scheme consists of a pile diameter and the number of piles.

[0092] It should be noted that the characteristic parameters that determine the pile foundation classification are encoded through the input layer to obtain an encoded input vector. When the characteristic parameter is the main beam type, encoding is performed according to specific rules, such as simply supported beam -10, continuous beam -20, continuous rigid frame -30, etc. When the characteristic parameter is the main beam length, the encoding method is the sum of the theoretical lengths of each span, and the length unit is preferably meters. When the characteristic parameter is the pier height, the encoding is the specific pier height, and the length unit is meters. When the characteristic parameter is the pier location, the encoding method uses specific mapping rules for digital mapping, such as side pier -1, secondary side pier -2, secondary middle pier -3, and middle pier -4. When the characteristic parameter is a terrain label, the encoding method uses specific mapping rules for digital mapping, such as conventional area -1, plain area -2, difficult mountainous area -3, eastern coastal area -4, and dense urban area -5.

[0093] After encoding, the input vector is Input=[main beam type code, beam length code, main span code, pier height code, pier position code, terrain code].

[0094] As shown in Table 1, Table 1 is an example table of a pile foundation sample set before encoding. The input data in the sample set, namely the characteristic parameters, include main beam type, main span, main beam length, pier location, and terrain label. The sample label includes the pile foundation type.

[0095] Table 1

[0096]

[0097] As shown in Table 2, Table 2 is an example table of the coded pile foundation sample set. The input data in the sample set, namely the characteristic parameters, include the main beam type, main span, main beam length, pier location, and terrain label. The sample label includes the pile foundation type. The simply supported beam in the main beam type is coded as 10, and the edge in the pier location is coded as 1.

[0098] Table 2

[0099]

[0100] It is worth noting that the input layer is used to output recommended pile foundation configuration schemes, where the pile foundation configuration scheme is composed of the pile diameter and the number of piles. Commonly used pile diameters include more than 10 types, such as 1.0m, 1.25m, 1.5m, 1.8m, 2.0m, 2.2m, 2.5m, and 3.0m. The number of piles in each diameter group is more than 10 types, such as 8, 9, 10, 11, 12, 15, 16, and 18. This embodiment does not impose specific restrictions on this. Pile groups are formed by combining and arranging the pile diameters and the number of piles to form different types of pile foundation configuration schemes. For example, 8-1.0m and 10-1.25m represent eight pile groups with a diameter of 1m and ten pile groups with a diameter of 1.25m, respectively. Therefore, for the output layer of a single softmax layer, the generated pile foundation configuration schemes have a scale of 10×10=100.

[0101] Furthermore, the output layer includes a first softmax layer and a second softmax layer, and outputting the recommended pile foundation configuration scheme through the output layer includes: determining the pile foundation diameter through one of the first softmax layer and the second softmax layer; determining the number of pile foundations through the other of the first softmax layer and the second softmax layer; and obtaining the recommended pile foundation configuration scheme based on the pile foundation diameter and the number of pile foundations.

[0102] It should be noted that, in this embodiment, the output layer of the neural network model of the pile group foundation recommendation algorithm is composed of two parallel softmax layers, namely the first softmax layer and the second softmax layer, where one softmax layer is used to determine the diameter of the pile foundation, and the other softmax layer is used to determine the number of pile foundations. In this way, the parameters of the output layer are reduced from the scale of 10×10=100 of the traditional neural network to 10+10=20.

[0103] Furthermore, obtaining a recommended pile foundation configuration scheme based on the pile foundation diameter and the pile foundation quantity includes: permuting and combining the pile foundation diameter and the pile foundation quantity to obtain a preset number of pile foundation configuration schemes; performing two-hot encoding on the preset number of pile foundation configuration schemes, wherein two-hot is composed of two one-hot vectors; obtaining an effective pile foundation diameter by identifying the pile foundation diameter by one of the two one-hot vectors; obtaining an effective pile foundation number by identifying the pile foundation number by the other of the two one-hot vectors; and obtaining a recommended pile foundation configuration scheme from the preset number of pile foundation configuration schemes based on the effective pile foundation diameter and the effective pile foundation number.

[0104] It should be noted that pile foundation configurations are formed by combining pile diameters and pile numbers. The output layer of the neural network model of the pile group foundation recommendation algorithm uses a two-tail network. One one-tail network component is used to classify pile diameters, and the other one-tail network component is used to classify pile numbers. The corresponding samples are two-hot encoded, consisting of two one-hot vectors: one identifying the pile diameter, and the other identifying the pile number. The identified pile diameters and pile numbers are considered valid pile diameters and valid pile numbers, and the pile foundation configuration consisting of these valid pile diameters and valid pile numbers is the recommended pile foundation configuration.

[0105] In a specific implementation, after obtaining the recommended pile foundation configuration scheme, the pile foundation is arranged according to the pile diameter and the number of piles in the recommended pile foundation configuration scheme. The pile foundation arrangement includes determinant, quincunx, etc. For example: 9 determinant piles with a diameter of 1.0m, 10 quincunx piles with a diameter of 1.0m, 11 quincunx piles with a diameter of 1.0m, 12 determinant piles with a diameter of 1.0m, etc. Figure 5a-5d As shown, Figure 5a-5d Schematic diagram of pile foundation arrangement of the pile foundation recommendation and data set design method based on deep neural network in this embodiment, wherein: Figure 5a This is a schematic diagram of the arrangement of 9 plum blossom piles with a diameter of 1.0m. Figure 5b The schematic diagram of the arrangement of 10 piles with a diameter of 1.0m is shown in Figure 1. Figure 5c This is a schematic diagram of the arrangement of 11 plum blossom piles with a diameter of 1.0m. Figure 5d Schematic diagram of the arrangement of 12 determinant piles with a diameter of 1.0m.

[0106] This embodiment obtains a bridge design plan; determines bridge design elements that determine the pile foundation configuration plan based on the bridge design plan; and determines basic characteristic parameters and the number of parameters based on the bridge design elements that determine the pile foundation configuration plan. In this manner, characteristic parameters that determine the pile foundation configuration plan are input into the model based on the characteristic parameters and the number of parameters determined based on the bridge design elements that determine the pile foundation configuration plan, thereby recommending pile foundation configuration plans and improving the accuracy of model recommendations.

[0107] refer to Figure 6 , Figure 6 This is a flow chart of the third embodiment of the pile foundation recommendation and dataset design method based on deep neural network of the present invention.

[0108] Based on the first embodiment described above, the pile foundation recommendation and data set design method based on a deep neural network in this embodiment further includes, before step S20:

[0109] Step S11: Obtain sample format requirements.

[0110] It should be noted that the sample format requirements can be abstracted and refined based on bridge scheme elements that influence the design solution. Design solutions can include main beam solutions, pier solutions, foundation solutions, etc. This embodiment does not limit this. This embodiment uses the foundation solution as an example for illustration. Foundation solutions include pile foundations, open-cut foundations, and well-dug foundations. For example, if the foundation solution is a pile foundation, and the pile foundation is primarily a group pile foundation, then the bridge scheme elements that influence the group pile foundation solution can be obtained and abstracted and refined to form the sample format requirements.

[0111] Step S12: Based on the sample format requirements, parameters required by the sample format and corresponding pile foundation configuration schemes are collected from existing bridge engineering projects to form a data set.

[0112] It should be noted that after obtaining the sample format requirements, a data sample set, namely a data set, can be constructed. When the design scheme is a pile group foundation scheme, the data sample set is a pile group foundation sample set.

[0113] The data sample set can be formed by collecting parameters required by the sample format and corresponding pile group foundation scheme types from existing bridge engineering projects based on the sample format requirements.

[0114] In a specific implementation, the samples corresponding to the data sample set are two-hot encoded and composed of two one-hot vectors. Each one-hot vector corresponds to one of the elements of the solution for the combined classification problem. The sample number vector in this example is: SampleLabel = [0, 1, 0…0, 1, 0]. The data sample set includes sample labels, which are mainly used to distinguish between pile group solutions. The sample labels are composed of pile diameter and pile number. The pile diameter and pile number are both composed of a dictionary and a one-hot vector. For example, the pile diameter dictionary vector = [1, 1.25, 1.5, 1.8, 2.0, 2.2, 2.5, 3.0]. The corresponding pile diameter one-hot vector is a vector of the same length as the pile diameter vector, with only one position being 1 and the other positions being zero. For example, the one-hot vector for a pile diameter of 1.8m is [0, 0, 0, 1, 0, 0, 0]. For example, if the dictionary vector for the number of piles is [8,9,10,11,12,14,15,18], the corresponding one-hot vector for the number of piles is a vector of the same length, with only one position set to 1 and all other positions set to zero. For example, the one-hot vector for 15 piles is [0,0,0,0,0,0,1,0]. The two-hot vector for the sample labels for the 15-1.8m solution is [0,0,0,1,0,0,0,0,0,0,0,0,0,0,1,0].

[0115] Step S13: constructing a neural network model of an initial pile group foundation recommendation algorithm according to the parameters required by the sample format, the sample format requirements, and the pile foundation configuration scheme.

[0116] In specific implementation, a neural network model of the pile group foundation recommendation algorithm can be constructed according to the parameters required by the verification format, the sample format requirements and the pile foundation configuration scheme. The neural network model is a neural network model of the design scheme recommendation algorithm. For example, if the design scheme is a pile group foundation scheme, the neural network model is a neural network model of the pile group foundation recommendation algorithm. The neural network model can be constructed using deep learning frameworks such as PyTorch, TensorFlow, Theano, and MXNet.

[0117] In a specific implementation, the output layer of the neural network model is a plurality of parallel softmax layers. An N-Tail (multi-tail) network layer can be used, which is composed of a plurality of one-tail (single-tail) network layers. For example, a Two-Tail (dual-tail) network layer can be used, which is composed of two one-tail (single-tail) network layers. This greatly reduces the number of parameters of the neural network and lowers the requirements for the sample size. This embodiment does not impose any specific restrictions on this.

[0118] Furthermore, step S13 includes: obtaining the number of parameters required by the sample format; constructing the input layer of the neural network model of the initial group pile foundation recommendation algorithm according to the parameters required by the sample format and the number of parameters required by the sample format; setting the size of the hidden layer according to the number of parameters required by the sample format to form the hidden layer of the multi-tail neural network; determining the number of softmax layers of the output layer according to the type of the pile group foundation; constructing the output layer of the neural network model of the initial group pile foundation recommendation algorithm based on the parameters required by the sample format and the number of softmax layers of the output layer; determining the target loss function based on the number of softmax layers of the output layer; and constructing the neural network model of the initial group pile foundation recommendation algorithm through the input layer, the hidden layer, the output layer, and the loss function.

[0119] It should be noted that the neural network model of this embodiment is different from the traditional neural network model. The neural network model consists of an input layer, a hidden layer and an output layer, and corresponds to a data set in a special format. The output layer in this embodiment is different from the output layer of the traditional neural network model. The output layer of the traditional neural network model has only one softmax layer, and the output layer of the neural network model in this embodiment has multiple softmax layers, for example, it can be 2 softmax layers, 3 softmax layers, etc. This embodiment does not impose specific restrictions on this.

[0120] When building a neural network model, it is necessary to build the input layer, hidden layer, and output layer. Therefore, the number of parameters required by the sample format can be obtained, which makes it convenient to build the input layer, hidden layer, and output layer according to the number of parameters.

[0121] In a specific implementation, the input layer of the neural network model of the initial group pile foundation recommendation algorithm can be constructed by the parameters required by the sample format and the number of parameters required by the sample format, and the number of parameters of the input layer is required to be the same as the number of parameters required by the sample format. The size of the hidden layer can be set according to the number of parameters required by the sample format, and the construction of the hidden layer can be determined with reference to a traditional neural network, thereby forming a hidden layer of a multi-tail neural network, for example, a hidden layer of a double-tail neural network. When the hidden layer of the double-tail neural network is formed, the output layer of the neural network model of this embodiment is composed of two softmax layers in parallel, and the number of softmax layers of the output layer can be determined by the type of group pile foundation. For example, if the type of group pile foundation includes a pile foundation arrangement-diameter combination, the number of softmax layers of the output layer is 2 layers, and the output layer adopts a two-tail network, in which one-tail network part is used to classify the diameter of the pile foundation, and the other one-tail network part is used to classify the number of pile foundations. The pile foundation type parameters can be determined according to the parameters required by the sample format, so that the pile foundation type parameters of multiple output network layers can be obtained according to the number of softmax layers of the output layer. For example, if the pile foundation type parameters are pile foundation arrangement and diameter, the pile foundation type parameters of the multiple output network layers include pile foundation arrangement parameters and diameter parameters, and the pile foundation type parameters of the multiple output network layers are multi-bit effectively encoded to obtain an encoded output vector. The encoded output vector is Output=[0.2, 0.98, …, 0.9, 0.1, …, 0, 0.95, …], so that the output layer of the neural network model of the initial pile group foundation recommendation algorithm is constructed through the encoded output vector.

[0122] It should be noted that the target loss function can be determined by the number of softmax layers in the output layer. By designing three loss functions in advance, the target loss function among the three functions can be determined according to the number of softmax layers.

[0123] Optionally, determining the target loss function based on the number of softmax layers in the output layer includes: when the number of softmax layers in the output layer is 2, splicing the two softmax layers to obtain a first loss function; adjusting the weights of each softmax layer in the network through feedback to obtain a second loss function; when the number of softmax layers in the output layer is single, splicing the single softmax layer to obtain a third loss function; and using the first loss function, the second loss function and the third loss function as target loss functions.

[0124] It should be understood that the loss function is the sum of the loss functions of two one-tail neural networks. There are three specific solutions. When the output layer has two softmax layers, Solution A concatenates the two softmax layers to obtain the first loss function. Solution B does not concatenate the two softmax layers and adjusts the weights of each softmax layer in the network through feedback, resulting in the second loss function. When the output layer has only one softmax layer, the single softmax layer is concatenated to obtain the third loss function.

[0125] Specifically, if the number of softmax layers is two, the loss function corresponding to scheme A is as follows:

[0126] (Formula 1)

[0127] The loss function corresponding to scheme B is as follows:

[0128] (Equation 2)

[0129] The loss function corresponding to scheme C is as follows:

[0130] (Equation 3)

[0131] In a specific implementation, the first loss function, the second loss function and the third loss function are used as target loss functions, and a loss function corresponding to one of schemes A, B and C is selected from the target loss functions.

[0132] After constructing the input layer, hidden layer, and output layer, a neural network model is constructed through the input layer, hidden layer, output layer, and loss function. Figure 7 As shown, Figure 7 This is a diagram of a traditional neural network. The output layer in a traditional neural network has only one softmax layer. Figure 8 As shown, Figure 8 This is a schematic diagram of a double-tailed neural network. The output layer of the double-tailed neural network consists of two parallel softmax layers, such as output layer A and output layer B. The classification scale of multiple classification factors is adjusted from multiplying the classification factors to the classification scale of multiple classification factors adding together, which greatly reduces the number of parameters of the neural network and thus reduces the requirements for sample size.

[0133] Furthermore, the input layer of the neural network model of the initial pile group foundation recommendation algorithm is constructed according to the parameters required by the sample format and the number of parameters required by the sample format, including: obtaining characteristic parameters that determine the classification of the pile foundation according to the parameters required by the sample format and the number of parameters required by the sample format, wherein the number of parameters required by the sample format is the same as the number of the characteristic parameters, and the characteristic parameters include one or more of main beam type, main beam length, main span length, pier height, pier position and terrain label; encoding the characteristic parameters to obtain an encoded input vector; and constructing the input layer of the neural network model of the initial pile group foundation recommendation algorithm through the encoded input vector.

[0134] It should be understood that the characteristic parameters for determining pile foundation classification can be obtained according to the parameters required by the sample format. Since the number of parameters in the input layer is the same as the number of parameters required by the sample format, the number of characteristic parameters can be determined according to the number of parameters required by the sample format.

[0135] Step S14: training the neural network model of the initial pile group foundation recommendation algorithm using the data set to obtain the neural network model of the target pile group foundation recommendation algorithm.

[0136] In a specific implementation, after obtaining a data sample set, the data samples can be input into a neural network model for training. When the training results reach the set evaluation index, a neural network model of the target pile group foundation recommendation algorithm is obtained.

[0137] After constructing the neural network model for the target pile group foundation recommendation algorithm, the neural network model can be used to process the combination classification of pile groups. The neural network model for the target pile group foundation recommendation algorithm is connected to the business system by deploying a server.

[0138] Furthermore, the step of obtaining the neural network model of the target pile group foundation recommendation algorithm specifically includes: obtaining an evaluation threshold of an evaluation indicator, wherein the evaluation indicator includes at least one of precision, recall rate and score value; training the neural network model of the initial pile group foundation recommendation algorithm through the data sample set to obtain a training result; evaluating the training result through the evaluation indicator; and obtaining the neural network model of the target pile group foundation recommendation algorithm when the training result meets the evaluation threshold of the evaluation indicator.

[0139] In a specific implementation, an evaluation threshold for an evaluation indicator can be obtained. The evaluation threshold can be set as needed. The evaluation indicator includes at least one of precision, recall, and score, and may also include other evaluation indicators, which are not limited in this embodiment. The score is the F1 value. The evaluation threshold includes a precision threshold, a recall threshold, and a score threshold.

[0140] It should be noted that when training the neural network model using a data sample set, real-time training results can be obtained and evaluated using evaluation indicators to determine whether the current neural network model meets the requirements. When the training results meet the evaluation threshold of the evaluation indicator, the neural network model is considered to have reached the evaluation depth requirement, and training is stopped to obtain the neural network model of the target pile group foundation recommendation algorithm. For example, if the accuracy of the training results is greater than the accuracy threshold and the score value is greater than the score threshold, the training results are considered to meet the evaluation threshold of the evaluation indicator.

[0141] This embodiment obtains sample format requirements; based on the sample format requirements, collects parameters and corresponding pile foundation configuration plans from existing bridge engineering projects to form a dataset; constructs a neural network model for an initial pile group foundation recommendation algorithm based on the parameters, sample format requirements, and pile foundation configuration plans; and trains the neural network model for the initial pile group foundation recommendation algorithm using the dataset to obtain a neural network model for a target pile group foundation recommendation algorithm. By constructing multiple parallel softmax layers in the output layer, the classification scale is adjusted from the multiplication of multiple classification elements to the sum of multiple classification elements, significantly reducing the number of neural network parameters and, in turn, the sample size requirement.

[0142] Reference Figure 9 , Figure 9 This is a structural block diagram of the first embodiment of the pile foundation recommendation and dataset design device based on deep neural network of the present invention.

[0143] like Figure 9 As shown, the pile foundation recommendation and data set design device based on deep neural network proposed in the embodiment of the present invention includes:

[0144] An acquisition module 10 is used to acquire characteristic parameters that determine pile foundation classification;

[0145] An input module 20 is configured to input the characteristic parameters into a neural network model of a pile group foundation recommendation algorithm to recommend a pile foundation configuration type, thereby obtaining a recommended pile foundation configuration scheme, wherein the neural network model of the pile group foundation recommendation algorithm is constructed based on a deep neural network, the deep neural network including an input layer, a hidden layer, and an output layer, the output layer of the deep neural network being composed of a single softmax layer, and the output layer of the neural network model of the pile group foundation recommendation algorithm being composed of multiple softmax layers;

[0146] The storage module 30 is used to merge the characteristic parameters and the recommended pile foundation configuration scheme into a database to form a sample data record;

[0147] The adding module 40 is used to add the sample data records to the original data set to obtain a new data set, wherein the new data set is used to continuously optimize and upgrade the neural network model of the pile group foundation recommendation algorithm by re-calling the new data set for model training.

[0148] This embodiment obtains characteristic parameters that determine pile foundation classification; inputs the characteristic parameters into the neural network model of the pile group foundation recommendation algorithm to recommend pile foundation configuration types, thereby obtaining a recommended pile foundation configuration scheme; merges the characteristic parameters and the recommended pile foundation configuration scheme into a database to form a sample data record; and adds the sample data record to the original data set to obtain a new data set. In this manner, the neural network model of the pile group foundation recommendation algorithm recommends a pile foundation configuration scheme based on the characteristic parameters that determine pile foundation classification, and updates the data set based on the recommended pile foundation configuration scheme. This solves the current problem of low accuracy of the pile group foundation neural network due to limited data sample size, reduces sample size requirements, and increases model recommendation accuracy.

[0149] In one embodiment, the acquisition module 10 is further used to obtain a bridge design scheme; determine bridge scheme elements that determine a pile foundation configuration scheme based on the bridge design scheme; and determine basic characteristic parameters and the number of parameters based on the bridge scheme elements that determine the pile foundation configuration scheme.

[0150] In one embodiment, the acquisition module 10 is also used to extract bridge schemes in existing bridge engineering projects to obtain initial characteristic parameters, wherein the initial characteristic parameters include at least one of the main beam type, main span length, main beam length, pier height, pier location and terrain characteristics; determine target characteristic parameters from the initial characteristic parameters based on the basic characteristic parameters and the number of parameters; and use the target characteristic parameters as characteristic parameters to determine the classification of pile foundations.

[0151] In one embodiment, the input module 20 is further used to input the characteristic parameters that determine the pile foundation classification into the input layer of the neural network model of the pile group foundation recommendation algorithm; encode the characteristic parameters that determine the pile foundation classification through the input layer to obtain an encoded input vector; input the encoded input vector into the hidden layer of the neural network model of the pile group foundation recommendation algorithm; input the encoded input vector into the output layer of the neural network model of the pile group foundation recommendation algorithm through the hidden layer; and output a recommended pile foundation configuration scheme through the output layer, wherein the recommended pile foundation configuration scheme consists of a pile diameter and the number of piles.

[0152] In one embodiment, the output layer includes a first softmax layer and a second softmax layer, and the input module 20 is further used to determine the pile foundation diameter through one of the first softmax layer and the second softmax layer; determine the number of pile foundations through the other of the first softmax layer and the second softmax layer; and obtain a recommended pile foundation configuration scheme based on the pile foundation diameter and the number of pile foundations.

[0153] In one embodiment, the input module 20 is further configured to permutate and combine the pile foundation diameter and the number of pile foundations to obtain a preset number of pile foundation configuration schemes; perform two-hot encoding on the preset number of pile foundation configuration schemes, wherein two-hot is composed of two one-hot vectors; obtain an effective pile foundation diameter by identifying the pile foundation diameter by one of the two one-hot vectors; obtain an effective pile foundation number by identifying the pile foundation number by the other of the two one-hot vectors; and obtain a recommended pile foundation configuration scheme from the preset number of pile foundation configuration schemes based on the effective pile foundation diameter and the effective pile foundation number.

[0154] In one embodiment, the input module 20 is further used to obtain sample format requirements; based on the sample format requirements, the parameters of the sample format requirements and the corresponding pile foundation configuration schemes are collected from existing bridge engineering projects to form a data set; a neural network model of an initial pile group foundation recommendation algorithm is constructed according to the parameters of the sample format requirements, the sample format requirements and the pile foundation configuration scheme; the neural network model of the initial pile group foundation recommendation algorithm is trained using the data set to obtain a neural network model of a target pile group foundation recommendation algorithm.

[0155] In addition, to achieve the above-mentioned purpose, the present invention also proposes a pile foundation recommendation and dataset design device based on a deep neural network, and the pile foundation recommendation and dataset design device based on a deep neural network includes: a memory, a processor, and a pile foundation recommendation and dataset design program based on a deep neural network stored on the memory and runnable on the processor, and the pile foundation recommendation and dataset design program based on a deep neural network is configured to implement the steps of the pile foundation recommendation and dataset design method based on a deep neural network as described above.

[0156] Since the present deep neural network-based pile foundation recommendation and data set design device adopts all the technical solutions of all the above embodiments, it has at least all the beneficial effects brought by the technical solutions of the above embodiments, which will not be described one by one here.

[0157] In addition, an embodiment of the present invention also proposes a storage medium, on which a pile foundation recommendation and dataset design program based on a deep neural network is stored. When the pile foundation recommendation and dataset design program based on a deep neural network is executed by a processor, the steps of the pile foundation recommendation and dataset design method based on a deep neural network as described above are implemented.

[0158] Since the storage medium adopts all the technical solutions of all the above embodiments, it has at least all the beneficial effects brought by the technical solutions of the above embodiments, which will not be described one by one here.

[0159] It should be understood that the above is only an example and does not constitute any limitation to the technical solution of the present invention. In specific applications, those skilled in the art can make settings as needed, and the present invention does not impose any limitation on this.

[0160] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of the present invention. In practical applications, technicians in this field can select part or all of it according to actual needs to achieve the purpose of the embodiment scheme, and no limitation is made here.

[0161] In addition, for technical details not fully described in this embodiment, please refer to the pile foundation recommendation and dataset design method based on deep neural network provided in any embodiment of the present invention, which will not be repeated here.

[0162] In addition, it should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or system. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or system comprising the element.

[0163] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0164] Through the above description of the embodiments, those skilled in the art will clearly understand that the methods of the above embodiments can be implemented using software plus the necessary general-purpose hardware platform. Of course, hardware can also be used, but in many cases the former is the more preferred implementation method. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory (ROM) / RAM, a magnetic disk, or an optical disk) and includes a number of instructions for enabling a terminal device (such as a mobile phone, computer, server, or network device) to execute the methods described in the various embodiments of the present invention.

[0165] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A pile foundation recommendation and dataset design method based on deep neural network, characterized in that: The method comprises: Obtain bridge design plans; Determining bridge scheme elements that determine pile foundation configuration schemes based on the bridge design scheme; Determining basic characteristic parameters and parameter quantities based on the bridge scheme elements that determine the pile foundation configuration scheme; Extracting bridge schemes from existing bridge engineering projects to obtain initial characteristic parameters, wherein the initial characteristic parameters include at least one of main beam type, main span length, main beam length, pier height, pier location, and terrain characteristics; Determining target feature parameters from the initial feature parameters based on the basic feature parameters and the number of parameters; Using the target characteristic parameters as characteristic parameters for determining pile foundation classification; Inputting the characteristic parameters into a neural network model of a pile group foundation recommendation algorithm to recommend a pile foundation configuration type, thereby obtaining a recommended pile foundation configuration scheme, wherein the neural network model of the pile group foundation recommendation algorithm is constructed based on a deep neural network, the deep neural network includes an input layer, a hidden layer, and an output layer, the output layer of the deep neural network is composed of a single softmax layer, and the output layer of the neural network model of the pile group foundation recommendation algorithm is composed of multiple softmax layers; Incorporating the characteristic parameters and the recommended pile foundation configuration scheme into a database to form a sample data record; The sample data records are added to the original data set to obtain a new data set, wherein the new data set is used to continuously optimize and upgrade the neural network model of the pile group foundation recommendation algorithm by re-calling the new data set for model training.

2. The method according to claim 1, wherein The characteristic parameters are input into a neural network model of a pile group foundation recommendation algorithm to recommend a pile foundation configuration type, thereby obtaining a recommended pile foundation configuration scheme, including: Inputting the characteristic parameters determining pile foundation classification into the input layer of the neural network model of the pile group foundation recommendation algorithm; Encoding the characteristic parameters determining the pile foundation classification through the input layer to obtain an encoded input vector; Inputting the encoded input vector into the hidden layer of the neural network model of the pile group foundation recommendation algorithm; Inputting the encoded input vector into the output layer of the neural network model of the pile group foundation recommendation algorithm through the hidden layer; The recommended pile foundation configuration scheme is outputted through the output layer, wherein the recommended pile foundation configuration scheme consists of a pile diameter and the number of piles.

3. The method according to claim 2, wherein The output layer includes a first softmax layer and a second softmax layer, and outputting the recommended pile foundation configuration solution through the output layer includes: Determine the pile diameter through one of the first softmax layer and the second softmax layer; Determining the number of pile foundations through the other of the first softmax layer and the second softmax layer; A recommended pile foundation configuration scheme is obtained based on the pile foundation diameter and the number of pile foundations.

4. The method according to claim 3, wherein The recommended pile foundation configuration scheme is obtained according to the pile foundation diameter and the number of pile foundations, including: Arrange and combine the pile foundation diameters and the number of pile foundations to obtain a preset number of pile foundation configuration solutions; Performing two-hot encoding on the preset number of pile foundation configuration schemes, wherein a two-hot encoding is formed by concatenating two one-hot vectors; Identifying the pile diameter by one of the two one-hot vectors to obtain the effective pile diameter; Obtain the number of valid foundation piles by identifying the number of foundation piles in the other of the two one-hot vectors; A recommended pile foundation configuration scheme is obtained from the preset number of pile foundation configuration schemes according to the effective foundation pile diameter and the effective foundation pile quantity.

5. The method according to claim 1, wherein Before inputting the characteristic parameters into the neural network model of the pile group foundation recommendation algorithm to recommend pile foundation configuration types and obtaining a recommended pile foundation configuration scheme, the method further includes: Obtain sample format requirements; Based on the sample format requirements, parameters required by the sample format and corresponding pile foundation configuration schemes are collected from existing bridge engineering projects to form a data set; Constructing a neural network model of an initial pile group foundation recommendation algorithm according to the parameters required by the sample format, the sample format requirements, and the pile foundation configuration scheme; The neural network model of the initial pile group foundation recommendation algorithm is trained using the data set to obtain a neural network model of the target pile group foundation recommendation algorithm.

6. A pile foundation recommendation and data set design device based on deep neural network, characterized in that: The pile foundation recommendation and data set design device based on deep neural network includes: an acquisition module, configured to acquire a bridge design scheme; determine bridge scheme elements that determine a pile foundation configuration scheme based on the bridge design scheme; determine basic characteristic parameters and a number of parameters based on the bridge scheme elements that determine the pile foundation configuration scheme; extract bridge schemes from existing bridge engineering projects to obtain initial characteristic parameters, wherein the initial characteristic parameters include at least one of main beam type, main span length, main beam length, pier height, pier location, and terrain characteristics; determine target characteristic parameters from the initial characteristic parameters based on the basic characteristic parameters and the number of parameters; and use the target characteristic parameters as characteristic parameters for determining pile foundation classification; An input module is used to input the characteristic parameters into a neural network model of a pile group foundation recommendation algorithm to recommend a pile foundation configuration type and obtain a recommended pile foundation configuration scheme, wherein the neural network model of the pile group foundation recommendation algorithm is constructed based on a deep neural network, and the deep neural network includes an input layer, a hidden layer, and an output layer, and the output layer of the deep neural network is composed of a single softmax layer, and the output layer of the neural network model of the pile group foundation recommendation algorithm is composed of multiple softmax layers; a storage module for integrating the characteristic parameters and the recommended pile foundation configuration scheme into a database to form a sample data record; An adding module is used to add the sample data records to the original data set to obtain a new data set, wherein the new data set is used to continuously optimize and upgrade the neural network model of the pile group foundation recommendation algorithm by re-calling the new data set for model training.

7. A pile foundation recommendation and data set design device based on deep neural network, characterized in that: The pile foundation recommendation and dataset design device based on deep neural network includes: a memory, a processor, and a pile foundation recommendation and dataset design program based on deep neural network stored in the memory and executable on the processor. The pile foundation recommendation and dataset design program based on deep neural network is configured to implement the pile foundation recommendation and dataset design method based on deep neural network as described in any one of claims 1 to 5.

8. A storage medium, characterized in that: The storage medium stores a pile foundation recommendation and data set design program based on a deep neural network. When the pile foundation recommendation and data set design program based on a deep neural network is executed by a processor, the pile foundation recommendation and data set design method based on a deep neural network as described in any one of claims 1 to 5 is implemented.

Images