CN117390381A - Method and device for predicting joint leakage of underground diaphragm walls based on deep learning - Google Patents

Abstract

The invention provides a method and a device for predicting joint seam leakage of a underground diaphragm wall based on deep learning, and relates to the technical field of foundation pit support waterproofing, wherein the method comprises the steps of obtaining first information and second information; preprocessing and converting the first information to obtain underground continuous wall engineering space-time data; establishing a leakage quantity prediction model according to the underground continuous wall engineering space-time data and a preset deep learning algorithm; and obtaining a recommended scheme according to the second information and the leakage quantity prediction model. According to the invention, a leakage prediction model of the joint seam of the underground continuous wall is established based on a deep learning algorithm, and the leakage of the joint seam of the underground continuous wall is predicted by acquiring a stratum distribution map of a history project, a reinforcing scheme of the joint seam of the underground continuous wall and leakage monitoring data and a stratum distribution map of a current construction area, so that recommended construction parameters and corresponding leakage prediction values are obtained, and reliable leakage prediction and guidance of optimizing a construction scheme are provided for underground continuous wall construction.

Description

Method and device for predicting underground continuous wall joint joint leakage based on deep learning

Technical field

The present invention relates to the technical field of foundation pit support and waterproofing, and specifically to a deep learning-based method and device for predicting joint leakage of underground diaphragm wall joints.

Background technique

In recent years, with the comprehensive development of urbanization in our country, the construction technology of underground engineering has become more and more mature, and the requirements for limiting precipitation in construction projects have become increasingly strict. The leakage problem in underground diaphragm wall projects has always been an important hidden danger for project quality and safety. Construction parameters need to be optimized to reduce the risk of leakage. Traditional methods are usually based on experience and expert knowledge, lacking scientificity and reliability, and Multi-factor interactions were not considered.

The present invention proposes a method for predicting underground continuous wall joint joint leakage based on deep learning. It establishes prediction models and decision rules based on historical project data to optimize and recommend construction parameters.

Contents of the invention

The purpose of the present invention is to provide a method and device for predicting underground continuous wall joint joint leakage based on deep learning to improve the above problems. In order to achieve the above objects, the technical solutions adopted by the present invention are as follows:

In the first aspect, this application provides a method for predicting joint leakage of underground diaphragm wall joints based on deep learning, including:

Obtain first information and second information. The first information includes stratigraphic distribution maps of historical projects, underground diaphragm wall joint reinforcement plans and leakage monitoring data. The underground diaphragm wall joint reinforcement plan includes the use of high-pressure jet spraying. Parameters for grouting reinforcement method construction, the second information includes the stratigraphic distribution map of the current construction area;

Preprocess and convert the first information to obtain spatio-temporal data of the underground diaphragm wall project. The spatio-temporal data of the underground diaphragm wall project includes first stratum distribution spatial data, leakage information time series data and construction plan model data;

Establish a leakage prediction model based on the spatiotemporal data of the underground diaphragm wall project and the preset deep learning algorithm;

A recommended solution is obtained based on the second information and the leakage amount prediction model, and the recommended solution includes at least one recommended construction parameter and a corresponding leakage amount prediction value.

In the second aspect, this application also provides a device for predicting joint leakage of underground diaphragm wall joints based on deep learning, including:

Acquisition module, used to acquire first information and second information. The first information includes stratigraphic distribution maps of historical projects, underground diaphragm wall joint reinforcement plans and leakage monitoring data. The underground diaphragm wall joint reinforcement plan Including parameters for construction using high-pressure jet grouting reinforcement method, and the second information includes a stratigraphic distribution map of the current construction area;

A conversion module for preprocessing and converting the first information to obtain spatiotemporal data of the underground diaphragm wall project. The spatiotemporal data of the underground diaphragm wall project includes first stratum distribution spatial data, leakage information time series data and construction plan model data. ;

A building module for establishing a leakage prediction model based on the spatiotemporal data of the underground diaphragm wall project and the preset deep learning algorithm;

An output module is configured to obtain a recommended solution based on the second information and the leakage amount prediction model, where the recommended solution includes at least one recommended construction parameter and a corresponding leakage amount prediction value.

The beneficial effects of the present invention are:

This invention establishes an underground diaphragm wall joint joint leakage prediction model based on a deep learning algorithm. By obtaining the stratigraphic distribution map of historical projects, the underground diaphragm wall joint reinforcement plan and leakage monitoring data, as well as the stratigraphic distribution map of the current construction area, the prediction The leakage amount of underground diaphragm wall joints can be used to obtain recommended construction parameters and corresponding leakage amount prediction values, providing reliable leakage amount prediction and guidance for optimizing construction plans for underground diaphragm wall construction, and improving construction quality and safety. properties, saving construction costs and time.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of embodiments of the invention. The objectives and other advantages of the invention may be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

Description of the drawings

In order to explain the technical solutions of the embodiments of the present invention more clearly, the drawings required to be used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and therefore do not It should be regarded as a limitation of the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without exerting creative efforts.

Figure 1 is a schematic flow chart of the underground continuous wall joint joint leakage prediction method based on deep learning described in the embodiment of the present invention;

Figure 2 is a schematic structural diagram of the underground diaphragm wall joint joint leakage prediction device based on deep learning described in the embodiment of the present invention.

Labels in the figure: 1. Acquisition module; 2. Transformation module; 21. First processing unit; 22. Second processing unit; 23. Third processing unit; 24. First integration unit; 241. First assignment unit; 242 , the second assignment unit; 243. the first calculation unit; 244. the first matching unit; 3. building module; 31. the first extraction unit; 32. the fourth processing unit; 33. the first modeling unit; 34. 2. Modeling unit; 4. Output module; 41. Fifth processing unit; 42. Second calculation unit; 43. Sixth processing unit; 431. First construction unit; 432. Seventh processing unit; 433. Eighth processing Unit; 434, ninth processing unit; 44, second integration unit.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments These are some embodiments of the present invention, rather than all embodiments. The components of the embodiments of the invention generally described and illustrated in the figures herein may be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of the invention provided in the appended drawings is not intended to limit the scope of the claimed invention, but rather to represent selected embodiments of the invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present invention.

It should be noted that similar reference numerals and letters represent similar items in the following figures, therefore, once an item is defined in one figure, it does not need further definition and explanation in subsequent figures. Meanwhile, in the description of the present invention, the terms "first", "second", etc. are only used to differentiate the description and cannot be understood as indicating or implying relative importance.

Example 1:

This embodiment provides a method for predicting joint leakage of underground diaphragm wall joints based on deep learning.

Referring to Figure 1, the figure shows that the method includes step S100, step S200, step S300 and step S400.

Step S100, obtain the first information and the second information. The first information includes the stratigraphic distribution map of the historical project, the underground diaphragm wall joint reinforcement plan and the leakage monitoring data. The underground diaphragm wall joint reinforcement plan includes the use of high-pressure jet injection. The parameters of the slurry reinforcement method construction, and the second information includes the stratigraphic distribution map of the current construction area.

It can be understood that in this step, multiple underground diaphragm wall projects are selected from historical projects, their stratigraphic distribution maps, leakage monitoring data and reinforcement plans are obtained, and then the required third layer is obtained by analyzing and processing these data. a message. At the same time, the stratigraphic distribution map of the current construction area is obtained as second information through on-site survey and data collection.

Step S200: Preprocess and transform the first information to obtain spatio-temporal data of the underground diaphragm wall project. The spatio-temporal data of the underground diaphragm wall project includes the first stratum distribution spatial data, leakage information time series data and construction plan model data.

It can be understood that in this step, the stratigraphic distribution map of the historical project is digitized, the geological features are extracted, the leakage monitoring data is analyzed and processed in time series, the leakage patterns and trends are extracted, and the connection is Analyze and sort out the seam reinforcement schemes and extract the reinforcement parameters. Through processing and transformation, the obtained spatiotemporal data of underground diaphragm wall engineering can provide basic data support for subsequent modeling and prediction. This step preprocesses and transforms the spatiotemporal data of underground continuous wall engineering in historical projects to obtain data formats and input data suitable for modeling and prediction, thereby improving the accuracy and efficiency of subsequent deep learning algorithm modeling and prediction. Provide more reliable data support for leakage prediction. It should be noted that step S200 includes step S210, step S220, step S230 and step S240.

Step S210: Convert the stratigraphic distribution map into a grid-like data structure to obtain first stratigraphic distribution spatial data. The first stratigraphic distribution spatial data includes grid nodes and connections between adjacent grid nodes. The grid nodes represent geological units. The connection is the spatial relationship and interaction between two geological units.

It can be understood that in this step, the stratigraphic distribution map is discretized, converted into the form of grid nodes, and the spatial relationship and interaction relationship between adjacent grid nodes are described through connections, thereby Obtain the first stratigraphic distribution spatial data. Preferably, the stratigraphic distribution map is divided into multiple grid cells, each cell represents a geological unit, and the adjacent relationship between the cells is represented as a connection line. On the basis of the connection, parameters such as the distance and contact area between adjacent geological units can be further calculated to more accurately describe the geological characteristics of the underground diaphragm wall construction area. This step converts the stratigraphic distribution map into a grid-like data structure, allowing the spatial relationship of geological units to be quantified and visualized, thereby supporting subsequent data processing and analysis.

Step S220: Convert the leakage amount monitoring data into leakage information time series data. Each time point in the leakage information time series data represents the leakage amount within a preset time period.

It can be understood that in this step, the leakage amount monitoring data is converted into leakage information time series data, and the leakage amount information at each time point is obtained through time series analysis and processing. Preferably, the monitoring data is smoothed to eliminate noise interference, and then a time series model is used to fit and predict the leakage data. These prediction results can be used to formulate reasonable construction parameters and leakage control plans, thereby improving project quality and efficiency.

Step S230: Convert the underground diaphragm wall joint reinforcement plan into construction plan model data. The construction plan model data includes three-dimensional model data and construction parameters. The three-dimensional model data includes three-dimensional graphics of bored piles and three-dimensional graphics of jet grouting piles.

It can be understood that in this step, the three-dimensional model data includes the three-dimensional graphics of the bored piles and the three-dimensional graphics of the jet grouting piles, and the construction parameters include the diameter of the borehole, the length of the cast-in-place piles, the diameter of the jet grouting piles, the spacing between the jet grouting piles, etc. This step converts the underground diaphragm wall joint reinforcement plan into an operable three-dimensional model and construction parameters, which provides important data support for subsequent modeling and prediction. Based on the three-dimensional model data and construction parameters, the spatial coordinates and reinforcement length of each reinforcement node are calculated to predict the amount of leakage. Through this step, complex reinforcement plans can be converted into operable data forms, providing important data support for subsequent modeling and prediction.

Step S240: Integrate the first stratum distribution spatial data, leakage information time series data and construction plan model data to obtain underground diaphragm wall engineering spatiotemporal data.

It can be understood that in this step, this step integrates the geological data, construction plans and leakage monitoring data of historical projects to obtain more comprehensive and accurate spatiotemporal data of underground diaphragm wall projects, which can provide guidance for subsequent leakage prediction and The recommendation of recommended construction parameters provides a more accurate and reliable data basis. At the same time, this step improves the utilization and value of data and provides support for the efficiency and safety of project construction. It should be noted that step S240 includes step S241, step S242, step S243 and step S244.

Step S241: Combine the geological attribute information and coordinate information of each grid node of the first formation distribution spatial data with the leakage amount monitoring data, add the leakage information time series data to the corresponding geological unit, and obtain the leakage amount monitoring data. Geological unit data.

It can be understood that in this step, the geological attribute information and coordinate information of each grid node in the first stratigraphic distribution spatial data are combined with the leakage amount monitoring data, and the leakage information time series data is added to the corresponding geological In the unit, the geological unit data with leakage information is obtained. In this way, by integrating different data, a richer and more complete spatiotemporal data of underground diaphragm wall engineering is formed, which provides a more accurate and reliable basis for subsequent leakage prediction and recommendation of recommended construction parameters.

Step S242: Add the void volume information in the construction parameters to the three-dimensional graph of the bored pile, and add the slurry orifice information in the construction parameters to the three-dimensional graph of the jet grouting pile to obtain an integrated data model.

It can be understood that in underground diaphragm wall projects, bored piles and jet grouting piles are commonly used reinforcement methods, and their construction parameters include void volume and slurry orifices. In this step, these construction parameters are added to the three-dimensional graphics to obtain an integrated data model containing complete construction information, which is beneficial to subsequent modeling and analysis, and can be used to support more accurate leakage predictions and recommended construction parameters. recommendation.

Step S243: Calculate the leakage calculation value according to the integrated data model and the preset finite element calculation formula.

It can be understood that in this step, for an underground diaphragm wall engineering system with a complex structure and multiple parameters, under the influence of many factors such as geological unit physical properties, construction parameters, leakage monitoring data, etc., the method can be adopted Finite element method is used to numerically simulate and analyze it. This method can divide the engineering system into multiple finite element units, solve the seepage flow field, stress field and other related parameters of each unit separately, and then combine them into the finite element solution of the entire system. Among them, the calculation formula of the seepage flow field is Darcy's law, and the calculation formula of the stress field is the elastic mechanics equation. The final leakage calculation value can be obtained through iterative calculation. The formulas involved are as follows:

q＝K×A×(u ₂ -u ₁ )/L;

Among them, q is the volume of liquid passing through unit cross-sectional area per unit time, K is the permeability coefficient, A is the cross-sectional area, u ₁ and u ₂ are the pressures of the liquid at both ends, and L is the length.

Step S244: Match the geological unit data, the integrated data model and the leakage calculation values according to the corresponding geological units to obtain the spatiotemporal data of the underground diaphragm wall project.

It can be understood that in this step, first, the leakage information is added to the corresponding geological unit data based on the coordinate information of the geological unit and the leakage information time series data, and then the construction parameter information in the integrated data model is combined with the geological unit The data is matched to determine the specific construction parameters of each geological unit. Finally, the matched geological unit data, integrated data model and leakage calculation values are integrated to obtain the spatiotemporal data of the underground diaphragm wall project, which can be used to predict subsequent leakage and recommend construction. Parameter recommendations are provided.

Step S300: Establish a leakage prediction model based on the spatiotemporal data of the underground diaphragm wall project and the preset deep learning algorithm.

It can be understood that in this step, the leakage amount prediction model can use the stratigraphic distribution map of historical projects, underground diaphragm wall joint reinforcement plans and leakage amount monitoring data, as well as the stratigraphic distribution map of the current construction area, to predict leakage quantity prediction. By training and optimizing deep learning algorithms, leakage volume can be predicted more accurately. It should be noted that step S300 includes step S310, step S320, step S330 and step S340.

Step S310: Feature extraction is performed on the first stratigraphic distribution spatial data and construction plan model data, and the extracted feature information is matched with the leakage information time series data to obtain a feature set.

It can be understood that in this step, the purpose of feature extraction on the spatiotemporal data of the underground diaphragm wall project is to transform the data into a form that can be processed by a computer. Preferably, this step can be completed through feature engineering, including extracting features related to leakage volume, such as the type, thickness, and depth of geological units, from stratigraphic distribution maps, construction plan model data, and leakage information time series data. and other geological attributes, as well as parameter information of the construction plan, such as the diameter, spacing, depth, etc. of the jet grouting piles. Corresponding these feature information with the leakage information time series data, a feature set is obtained for subsequent processing and analysis by the deep learning algorithm.

Step S320: Divide the feature set into a training set and a test set according to a preset ratio.

It can be understood that in this step, the training set is a data set used to train the deep learning algorithm, and the test set is a data set used to evaluate the prediction accuracy and generalization ability of the trained model. Preferably, the data in the feature set can be divided into a training set and a test set in a ratio of 80:20. Among them, 80% of the data is used as a training set to train the model; 20% of the data is used as a test set to test the prediction effect of the model.

Step S330: Use the training set to train the preset first recurrent neural network model, and use the preset backpropagation algorithm to optimize the first recurrent neural network model to obtain an initial model.

It can be understood that in this step, the first recurrent neural network model is a neural network model that can process data with time series characteristics, and the model parameters can be optimized by using the back propagation algorithm to improve the accuracy of model prediction. Preferably, during specific implementation, the model is trained and optimized by setting different model hyperparameters, number of training rounds and other parameters based on the feature set of the training set and the corresponding leakage information time series data. During model training and optimization, the mean squared error is used to evaluate the model's predictive accuracy in order to validate it on subsequent test sets. The beneficial effect of this step is that the initial model obtained through training can improve the accuracy and accuracy of leakage prediction, thereby providing reliable data support for subsequent leakage prediction and recommendation of recommended construction parameters.

Among them, the first recurrent neural network model used in this step includes an input layer, a serialization layer, a long short-term memory network layer, a fully connected layer and an output layer. The input layer is used to receive the first stratum distribution spatial data, construction plan model data and leakage information time series data. The serialization layer is used to serialize the input data so that it meets the input requirements of the recurrent neural network. The long short-term memory network layer performs cyclic processing on the serialized input data and extracts feature information in the sequence. The fully connected layer is used to flatten the feature information extracted from the long short-term memory network layer, and perform feature fusion and mapping through the fully connected layer to obtain the feature vector. The output layer is used to map the feature vector into the preset leakage prediction value space to obtain the corresponding leakage prediction value.

Preferably, in this embodiment, according to the actual situation of the data set, the number of neurons in the input layer is 3, which respectively correspond to the first stratum distribution spatial data, construction plan model data and leakage information time series data. The number of neurons in the long short-term memory network layer is 64, and the length of the time series in the corresponding data set is 64. The number of neurons in the fully connected layer is 128, which is set based on factors such as data complexity and required accuracy. The final output layer has only one neuron, representing the predicted leakage value.

Step S340: Use the test set to verify the initial model, evaluate the performance of the first recurrent neural network model, and perform parameter adjustment and optimization of the model based on the evaluation results to obtain a leakage prediction model.

It can be understood that in this step, the preset test set is used to verify and evaluate the initial model to understand the performance of the first recurrent neural network model, including the accuracy, generalization ability, stability, etc. of the model. By evaluating the model performance, the optimization direction of the model and the parameters that need to be adjusted can be determined, and the parameters of the model can be further adjusted and optimized to obtain a more accurate and reliable leakage prediction model.

Step S400: Obtain a recommended solution based on the second information and the leakage amount prediction model. The recommended solution includes at least one recommended construction parameter and a corresponding leakage amount prediction value.

It can be understood that in this step, the corresponding feature information is extracted based on the stratigraphic distribution map of the current construction area, and then input into the trained model to obtain the corresponding leakage amount prediction value. Based on the predicted values and preset thresholds, the degree of leakage risk in the current construction area can be judged, and then recommended construction parameters can be proposed to reduce the risk of leakage. It should be noted that step S400 includes step S410, step S420, step S430 and step S440.

Step S410: Convert the second information into a network-like data structure to obtain second stratigraphic distribution spatial data.

It can be understood that in this step, the geological distribution map in the second information is converted into a grid-like data structure. Each grid node represents a geological unit, and the connection between the nodes represents the relationship between the geological units. Spatial relationships and interaction relationships are used to obtain the second stratigraphic distribution spatial data. This step converts the second information into a network-like data structure so that the geological distribution information can be processed and analyzed by computers, providing basic data support for subsequent steps.

Step S420: Use the second formation distribution spatial data as an input value of the leakage amount prediction model to obtain a leakage amount prediction result, where the leakage amount prediction result includes at least one leakage amount prediction value.

It can be understood that in this step, the second formation distribution spatial data may be geological information data obtained through geological exploration, geological surveying and other means, such as formation type, thickness, porosity and other parameters at different depths. Using these parameters as input, the leakage prediction model can predict the leakage prediction value at the corresponding depth, thereby recommending suitable recommended construction parameters.

Step S430: Obtain at least one recommended construction parameter based on the leakage amount prediction result and the preset decision rule.

It can be understood that in this step, based on the leakage volume prediction value predicted by the leakage volume prediction model and combined with the preset decision rules, the optimal or suboptimal construction parameter combination is obtained as part of the recommended plan. Preferably, according to the actual situation and needs, different decision rules can be set, such as giving different combinations of recommended construction parameters based on the predicted value of leakage, or making trade-offs and calculations based on other factors, such as cost, feasibility, etc. decision making. It should be noted that step S430 includes step S431, step 432, step S433 and step S434.

Step S431: Establish decision rules based on the first information and the preset decision tree algorithm, and classify the construction parameters in the underground diaphragm wall joint reinforcement plan according to the decision rules to obtain the construction parameter range corresponding to each category.

It can be understood that in this step, the decision rules are first established based on the first information and the preset decision tree algorithm, that is, the known construction information is classified and organized, and the corresponding decision rules are formulated to facilitate the construction of the underground continuous wall. The construction parameters in the joint reinforcement scheme are classified. Decision tree algorithms usually branch information according to specific attributes and values to build a decision tree to classify information and make decisions. Then, according to the decision rules, the construction parameters in the underground diaphragm wall joint reinforcement scheme are classified, and the construction parameter range corresponding to each category is obtained. For example, such as bored pile length, hole spacing, pouring depth, jet grouting pile spacing, spray pressure, etc., and then set the corresponding construction parameter range for each category, such as bored pile length of 10-15 meters, hole spacing The depth is 1.5-2.0 meters, the filling depth is 20-25 meters, etc. Finally, based on the leakage prediction results and known construction parameter classifications, preset decision rules are used for reasoning to obtain at least one recommended construction parameter. For example, when the predicted leakage amount is large, it can be inferred based on the preset decision rules that more stringent construction parameters need to be adopted, such as increasing the number of bored piles, reducing the spacing of grouting holes, increasing the depth of grouting, etc. This step quickly and effectively obtains at least one recommended construction parameter based on the second information and the preset decision rules, which improves the accuracy and efficiency of decision-making, reduces the cost of trial and error and adjustment, and can also optimize the construction plan. Engineering risks and quality hazards are reduced.

Step S432: Divide the construction parameter range into at least two sub-intervals, and perform parameter sensitivity analysis on each sub-interval to obtain analysis results. The analysis results include the degree of influence of each parameter on the leakage amount prediction result.

It can be understood that in this step, the sensitivity analysis is to evaluate the impact of small changes in construction parameters on the leakage prediction results. Preferably, in this embodiment, global sensitivity analysis is used to select the void volume and slurry orifice parameters in the construction plan model data and the first stratum distribution spatial data in the underground diaphragm wall engineering spatio-temporal data as variables for analysis, and multiple Comprehensive evaluation of the impact of parameters on leakage volume prediction results. In the underground diaphragm wall project, the influence of multiple parameters on the leakage prediction results is considered, such as permeability coefficient, water head, groundwater level difference, etc. Through global sensitivity analysis, the impact of each parameter on the leakage prediction results can be obtained, so that the importance of each parameter can be understood. Using the Latin hypercube sampling method, the parameter space is sampled, numerical simulation is performed, and the main effect and interaction effect of each parameter are calculated, and finally the degree of influence of each parameter on the model output is obtained. In this way, the selection of construction parameters can be more scientific and reasonable, and the quality and efficiency of the project can be improved.

Step S433: According to the leakage amount prediction results and analysis results, the construction parameters of each sub-interval are optimized to obtain the corresponding leakage amount prediction value.

It can be understood that, preferably, in this embodiment, the leakage amount prediction model is applied to the underground diaphragm wall project, and the second stratigraphic distribution spatial data generated according to the second information is used as the input of the model, and the Latin hypercube sampling method is used. A set of representative construction parameter combinations are generated, and then these parameter combinations are input into the leakage prediction model to obtain the corresponding leakage prediction value. Then, the influence degree of each construction parameter is evaluated through the global sensitivity analysis method, and the sensitivity value of each parameter is obtained. Finally, based on the leakage prediction results and the sensitivity values of each parameter, the construction parameters of each sub-interval are optimized to obtain the corresponding leakage prediction value and the best construction parameter combination. The beneficial effect of this step is to improve the construction efficiency and construction quality of the underground diaphragm wall project, reduce the construction cost and construction period, reduce the risk of leakage, and improve the stability and safety of the underground diaphragm wall. .

Step S434: Obtain recommended construction parameters based on the optimized construction parameters and corresponding leakage prediction values.

It can be understood that in this step, the optimal construction parameters for each sub-interval are determined based on the leakage prediction results and analysis results obtained in the previous steps, and recommended construction parameters are obtained.

Step S440: Integrate all recommended construction parameters and corresponding leakage prediction values to obtain a recommended solution.

It can be understood that in this step, the recommended plan includes a variety of recommended construction parameters, and each parameter corresponds to a predicted value of leakage, which is used as a basis to provide decision support. In this step, by integrating various recommended parameters and corresponding leakage prediction values, multiple solutions can be comprehensively evaluated and the optimal solution can be selected, thereby effectively guiding the construction practice of underground diaphragm wall engineering and improving its construction. Quality and efficiency.

Example 2:

As shown in Figure 2, this embodiment provides a device for predicting underground continuous wall joint joint leakage based on deep learning. The device includes:

Acquisition module 1 is used to obtain the first information and the second information. The first information includes the stratigraphic distribution map of the historical project, the underground diaphragm wall joint reinforcement plan and the leakage monitoring data. The underground diaphragm wall joint reinforcement plan includes the use of high pressure The parameters of the jet grouting reinforcement method construction, the second information includes the stratigraphic distribution map of the current construction area.

The conversion module 2 is used to preprocess and convert the first information to obtain the spatio-temporal data of the underground diaphragm wall project. The spatio-temporal data of the underground diaphragm wall project includes the first stratum distribution spatial data, leakage information time series data and construction plan model data.

Building module 3 is used to establish a leakage prediction model based on the spatiotemporal data of the underground diaphragm wall project and the preset deep learning algorithm.

The output module 4 is used to obtain a recommended plan based on the second information and the leakage amount prediction model. The recommended plan includes at least one recommended construction parameter and a corresponding leakage amount prediction value.

In a specific implementation of the present disclosure, the conversion module 2 includes:

The first processing unit 21 is used to convert the stratigraphic distribution map into a grid-like data structure to obtain first stratigraphic distribution spatial data. The first stratigraphic distribution spatial data includes grid nodes and connections between adjacent grid nodes. The grid Nodes represent geological units, and lines represent the spatial relationship and interaction between two geological units.

The second processing unit 22 is used to convert the leakage amount monitoring data into leakage information time series data. Each time point in the leakage information time series data represents the leakage amount within a preset time period.

The third processing unit 23 is used to convert the underground diaphragm wall joint reinforcement plan into construction plan model data. The construction plan model data includes three-dimensional model data and construction parameters. The three-dimensional model data includes three-dimensional graphics of bored piles and three-dimensional jet grouting piles. graphics.

The first integration unit 24 is used to integrate the first stratum distribution spatial data, leakage information time series data and construction plan model data to obtain underground continuous wall engineering spatiotemporal data.

In a specific implementation of the present disclosure, the first integration unit 24 includes:

The first assignment unit 241 is used to combine the geological attribute information and coordinate information of each grid node of the first stratigraphic distribution spatial data with the leakage amount monitoring data, and add the leakage information time series data to the corresponding geological unit, to obtain Geological unit data with seepage information.

The second assignment unit 242 is used to add the void volume information in the construction parameters to the three-dimensional graph of the bored pile, and add the slurry orifice information in the construction parameters to the three-dimensional graph of the jet grouting pile to obtain an integrated data model.

The first calculation unit 243 is used to calculate the leakage calculation value according to the integrated data model and the preset finite element calculation formula.

The first matching unit 244 is used to match the geological unit data, the integrated data model and the leakage calculation value according to the corresponding geological unit to obtain the spatiotemporal data of the underground diaphragm wall project.

In a specific implementation of the present disclosure, building module 3 includes:

The first extraction unit 31 is used to extract features from the first stratigraphic distribution spatial data and construction plan model data, and associate the extracted feature information with the leakage information time series data to obtain a feature set.

The fourth processing unit 32 is used to divide the feature set into a training set and a test set according to a preset ratio.

The first modeling unit 33 uses the training set to train the preset first recurrent neural network model, and uses the preset backpropagation algorithm to optimize the first recurrent neural network model to obtain an initial model.

The second modeling unit 34 uses the test set to verify the initial model, evaluates the performance of the first recurrent neural network model, and performs parameter adjustment and optimization of the model based on the evaluation results to obtain a leakage prediction model.

In a specific implementation of the present disclosure, the output module 4 includes:

The fifth processing unit 41 is used to convert the second information into a network-like data structure to obtain the second stratigraphic distribution spatial data.

The second calculation unit 42 is configured to use the second formation distribution spatial data as an input value of the leakage amount prediction model to obtain a leakage amount prediction result, where the leakage amount prediction result includes at least one leakage amount prediction value.

The sixth processing unit 43 is used to derive at least one recommended construction parameter based on the leakage amount prediction results and the preset decision rules.

The second integration unit 44 is used to integrate all recommended construction parameters and corresponding leakage prediction values to obtain a recommended solution.

In a specific implementation of the present disclosure, the sixth processing unit 43 includes:

The first construction unit 431 is used to establish decision rules based on the first information and the preset decision tree algorithm, and classify the construction parameters in the underground diaphragm wall joint reinforcement plan according to the decision rules to obtain the construction parameter range corresponding to each category. .

The seventh processing unit 432 is used to divide the construction parameter range into at least two sub-intervals, and perform parameter sensitivity analysis on each sub-interval to obtain analysis results. The analysis results include the degree of influence of each parameter on the leakage amount prediction result.

The eighth processing unit 433 is used to optimize the construction parameters of each sub-interval based on the leakage amount prediction results and analysis results to obtain the corresponding leakage amount prediction value.

The ninth processing unit 434 is used to obtain recommended construction parameters based on the optimized construction parameters and corresponding leakage prediction values.

It should be noted that, regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be described in detail here.

The above descriptions are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention. should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims (10)

1. A method for predicting joint leakage of underground diaphragm walls based on deep learning, which is characterized by including: Obtain first information and second information. The first information includes stratigraphic distribution maps of historical projects, underground diaphragm wall joint reinforcement plans and leakage monitoring data. The underground diaphragm wall joint reinforcement plan includes the use of high-pressure jet spraying. Parameters for the grouting reinforcement method construction, the second information includes the stratigraphic distribution map of the current construction area; Preprocess and convert the first information to obtain spatio-temporal data of the underground diaphragm wall project. The spatio-temporal data of the underground diaphragm wall project includes first stratum distribution spatial data, leakage information time series data and construction plan model data; Establish a leakage prediction model based on the spatiotemporal data of the underground diaphragm wall project and the preset deep learning algorithm; A recommended solution is obtained based on the second information and the leakage amount prediction model, and the recommended solution includes at least one recommended construction parameter and a corresponding leakage amount prediction value. 2. The deep learning-based underground diaphragm wall joint joint leakage prediction method according to claim 1, characterized in that the first information is preprocessed and transformed to obtain the underground diaphragm wall engineering spatiotemporal data, including: The formation distribution map is converted into a grid-like data structure to obtain first formation distribution spatial data. The first formation distribution spatial data includes grid nodes and connections between adjacent grid nodes. The grid The nodes represent geological units, and the connection lines are the spatial relationship and interaction relationship between the two geological units; Convert the leakage amount monitoring data into leakage information time series data, where each time point in the leakage information time series data represents the leakage amount within a preset time period; Convert the underground diaphragm wall joint reinforcement plan into construction plan model data. The construction plan model data includes three-dimensional model data and construction parameters. The three-dimensional model data includes three-dimensional graphics of bored piles and three-dimensional graphics of jet grouting piles; The first stratigraphic distribution spatial data, the leakage information time series data and the construction plan model data are integrated to obtain the underground diaphragm wall project spatio-temporal data. 3. The deep learning-based underground diaphragm wall joint joint leakage prediction method according to claim 2, characterized in that the first stratum distribution spatial data, the leakage information time series data and the construction plan model are The data is integrated to obtain the spatiotemporal data of the underground diaphragm wall project, including: Combine the geological attribute information and coordinate information of each grid node of the first formation distribution spatial data with the leakage amount monitoring data, and add the leakage information time series data to the corresponding geological unit to obtain leakage information. Geological unit data; Add the void volume information in the construction parameters to the three-dimensional graph of the bored pile, and add the slurry orifice information in the construction parameters to the three-dimensional graph of the jet grouting pile to obtain an integrated data model; The leakage calculation value is calculated according to the integrated data model and the preset finite element calculation formula; The geological unit data, the integrated data model and the leakage calculation value are matched according to the corresponding geological unit to obtain the spatiotemporal data of the underground diaphragm wall project. 4. The deep learning-based method for predicting underground diaphragm wall joint joint leakage according to claim 1, characterized in that a recommended solution is obtained according to the second information and the leakage amount prediction model, and the recommended solution includes At least one recommended construction parameter and corresponding leakage prediction value, including: Convert the second information into a network-like data structure to obtain second stratigraphic distribution spatial data; Using the second formation distribution spatial data as the input value of the leakage prediction model, a leakage prediction result is obtained, and the leakage prediction result includes at least one leakage prediction value; According to the leakage amount prediction results and the preset decision rules, at least one recommended construction parameter is obtained; All the recommended construction parameters and the corresponding predicted leakage values are integrated to obtain a recommended solution. 5. The method for predicting joint leakage of underground diaphragm wall joints based on deep learning according to claim 4, characterized in that at least one recommended construction parameter is obtained based on the leakage prediction result and the preset decision rule. ,include: Establish decision rules based on the first information and a preset decision tree algorithm, and classify the construction parameters in the underground diaphragm wall joint reinforcement scheme according to the decision rules to obtain the construction parameter range corresponding to each category; Divide the construction parameter range into at least two sub-intervals, and perform parameter sensitivity analysis on each of the sub-intervals to obtain analysis results, where the analysis results include the degree of influence of each parameter on the leakage amount prediction result; According to the leakage amount prediction results and the analysis results, the construction parameters of each sub-interval are optimized to obtain the corresponding leakage amount prediction value; Based on the optimized construction parameters and corresponding leakage prediction values, recommended construction parameters are obtained. 6. A device for predicting joint leakage of underground diaphragm walls based on deep learning, which is characterized by including: Acquisition module, used to acquire first information and second information. The first information includes stratigraphic distribution maps of historical projects, underground diaphragm wall joint reinforcement plans and leakage monitoring data. The underground diaphragm wall joint reinforcement plan Including parameters for construction using high-pressure jet grouting reinforcement method, and the second information includes a stratigraphic distribution map of the current construction area; A conversion module for preprocessing and converting the first information to obtain spatio-temporal data of the underground diaphragm wall project. The spatio-temporal data of the underground diaphragm wall project includes first stratum distribution spatial data, leakage information time series data and construction plan model data. ; A building module for establishing a leakage prediction model based on the spatiotemporal data of the underground diaphragm wall project and the preset deep learning algorithm; An output module is configured to obtain a recommended solution based on the second information and the leakage amount prediction model, where the recommended solution includes at least one recommended construction parameter and a corresponding leakage amount prediction value. 7. The deep learning-based underground diaphragm wall joint joint leakage prediction device according to claim 6, characterized in that the conversion module includes: The first processing unit is used to convert the stratigraphic distribution map into a grid-like data structure to obtain first stratigraphic distribution spatial data. The first stratigraphic distribution spatial data includes grid nodes and adjacent grid nodes. Connection lines, the grid nodes represent geological units, and the connection lines are the spatial relationship and interaction relationship between the two geological units; A second processing unit, configured to convert the leakage amount monitoring data into leakage information time series data, where each time point in the leakage information time series data represents the leakage amount within a preset time period; The third processing unit is used to convert the underground diaphragm wall joint reinforcement plan into construction plan model data. The construction plan model data includes three-dimensional model data and construction parameters. The three-dimensional model data includes three-dimensional graphics of bored piles. and three-dimensional graphics of jet grouting piles; The first integration unit is used to integrate the first stratigraphic distribution spatial data, the leakage information time series data and the construction plan model data to obtain underground continuous wall engineering spatiotemporal data. 8. The deep learning-based underground diaphragm wall joint joint leakage prediction device according to claim 7, characterized in that the first integration unit includes: The first assignment unit is used to combine the geological attribute information and coordinate information of each grid node of the first formation distribution spatial data with the leakage amount monitoring data, and add the leakage information time series data to the corresponding geological unit. , obtain geological unit data with leakage information; The second assignment unit is used to add the void volume information in the construction parameters to the three-dimensional graph of the bored pile, and add the slurry orifice information in the construction parameters to the three-dimensional graph of the jet grouting pile, to obtain Integrated data model; The first calculation unit is used to calculate the leakage calculation value according to the integrated data model and the preset finite element calculation formula; The first matching unit is used to match the geological unit data, the integrated data model and the leakage calculation value according to the corresponding geological unit to obtain the spatiotemporal data of the underground diaphragm wall project. 9. The deep learning-based underground diaphragm wall joint joint leakage prediction device according to claim 6, characterized in that the output module includes: A fifth processing unit, configured to convert the second information into a network-like data structure to obtain second stratigraphic distribution spatial data; A second calculation unit, configured to use the second formation distribution spatial data as an input value of the leakage prediction model to obtain a leakage prediction result, where the leakage prediction result includes at least one leakage prediction value ; A sixth processing unit, configured to derive at least one recommended construction parameter based on the leakage prediction results and the preset decision rules; The second integration unit is used to integrate all the recommended construction parameters and the corresponding predicted leakage values to obtain a recommended solution. 10. The deep learning-based underground diaphragm wall joint joint leakage prediction device according to claim 9, characterized in that the sixth processing unit includes: A first building unit, configured to establish decision rules based on the first information and a preset decision tree algorithm, and classify the construction parameters in the underground diaphragm wall joint reinforcement scheme according to the decision rules to obtain each category Corresponding construction parameter range; The seventh processing unit is used to divide the construction parameter range into at least two sub-intervals, and perform parameter sensitivity analysis on each of the sub-intervals to obtain analysis results. The analysis results include the effect of each parameter on the leakage amount. The degree of impact of the predicted results; An eighth processing unit is used to optimize the construction parameters of each sub-interval according to the leakage amount prediction results and the analysis results, and obtain the corresponding leakage amount prediction value; The ninth processing unit is used to obtain recommended construction parameters based on the optimized construction parameters and corresponding leakage prediction values.