Real-time dynamic prediction system and method for three-dimensional configuration of high-pressure jet grouting pile forming

Abstract

Provided in the present invention are a real-time dynamic prediction system and method for three-dimensional configuration of high-pressure jet grouting pile forming. The method comprises: acquiring a training dataset; a model construction module building a diameter prediction model for a high-pressure jet grouting pile on the basis of BRNN and GRU fusion; a model training module training the diameter prediction model for a high-pressure jet grouting pile by means of the training dataset; a prediction module performing prediction on the basis of the trained diameter prediction model for a high-pressure jet grouting pile, so as to obtain diameter prediction information during a construction process of a building project; and a diameter output module for a high-pressure jet grouting pile determining whether the diameter prediction information matches a diameter mode, and if the two match, outputting the diameter prediction information, and if the two do not match, adjusting operation parameters of the diameter prediction model for a high-pressure jet grouting pile, and performing prediction again. The real-time dynamic prediction system and method for three-dimensional configuration of high-pressure jet grouting pile forming provided in the present invention can realize real-time dynamic evaluation for the diameter of a high-pressure jet grouting pile.

Description

A real-time dynamic prediction system and method for the three-dimensional shape of high-pressure jet grouting piles Technical Field

[0001] The present invention relates to the technical field of construction engineering, and in particular to a real-time dynamic prediction system and method for the three-dimensional form of high-pressure jet-jet piles. Background Art

[0002] With the continuous development of my country's infrastructure, high-pressure rotary jet grouting has become a popular soil stabilization technology, widely used in the construction industry to meet various geotechnical engineering challenges. High-pressure rotary jet grouting uses a high-pressure rotating nozzle to spray cement slurry into the soil layer and mix it with the soil, forming a continuous overlapping cement reinforcement. This method requires a small construction footprint, has low vibration, and is relatively quiet, but it is prone to environmental pollution and is costly. It is not suitable for special soil types that cannot solidify the sprayed slurry. Therefore, it is necessary to accurately predict the parameters of high-pressure rotary jet piles to ensure the safety and reliability of the construction process.

[0003] Most existing research fails to consider the impact of complex strata on the parameters of high-pressure jet grouting piles. Using the average diameter of high-pressure jet grouting piles as a key indicator can lead to errors, making it difficult to dynamically monitor the diameter of the piles in real time. Therefore, it is crucial to design a real-time dynamic prediction system and method for the three-dimensional morphology of high-pressure jet grouting piles.

[0004] Summary of the Invention

[0005] The purpose of the present invention is to provide a real-time dynamic prediction system and method for the three-dimensional morphology of high-pressure rotary jet grouting piles, which can realize real-time dynamic evaluation of the diameter of high-pressure rotary jet grouting piles, can provide feedback and correction on the input data of the prediction model, and can effectively improve the prediction accuracy.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A real-time dynamic prediction system for the three-dimensional shape of high-pressure jet grouting piles, comprising: a model construction module, a model training module, a prediction module, and a high-pressure jet grouting pile diameter output module, wherein the model construction module is connected to the model training module, the model training module is connected to the prediction module, and the prediction module is connected to the high-pressure jet grouting pile diameter output module;

[0008] The model building module is used to build a high-pressure jet grouting pile diameter prediction model based on BRNN and GRU fusion;

[0009] The model training module is used to obtain a training data set and train a high-pressure jet grouting pile diameter prediction model based on the training data set;

[0010] The prediction module is used to make predictions based on the trained high-pressure jet grouting pile diameter prediction model to obtain diameter prediction information during the construction process of the building project;

[0011] The high-pressure rotary jet pile diameter output module is used to determine whether the acquired diameter prediction information matches the diameter pattern. If not, the operating parameters of the high-pressure rotary jet pile diameter prediction model are adjusted and the prediction is performed again. If they match, the diameter prediction information is output.

[0012] The present invention also provides a real-time dynamic prediction method for the three-dimensional shape of high-pressure rotary jet grouting piles, which is applied to the above-mentioned real-time dynamic prediction system for the three-dimensional shape of high-pressure rotary jet grouting piles, comprising the following steps:

[0013] Step 1: Get the training dataset;

[0014] Step 2: The model building module builds a high-pressure jet grouting pile diameter prediction model based on the fusion of BRNN and GRU;

[0015] Step 3: The model training module trains the high-pressure jet grouting pile diameter prediction model using the training data set;

[0016] Step 4: The prediction module performs prediction based on the trained high-pressure jet grouting pile diameter prediction model to obtain diameter prediction information during the construction process of the building project;

[0017] Step 5: The high-pressure rotary jet pile diameter output module determines whether the diameter prediction information matches the diameter pattern. If so, the diameter prediction information is output. If not, the operating parameters of the high-pressure rotary jet pile diameter prediction model are adjusted and the prediction is performed again until the diameter prediction information matches the diameter pattern. At this time, the diameter prediction information is output.

[0018] Optionally, in step 1, obtain the training data set, specifically:

[0019] Relevant soil data were collected based on the on-site investigation to obtain soil layer parameters. Based on the high-pressure jet grouting pile test, the injection parameters and diameter of the high-pressure jet grouting pile, namely the pile test parameters, were obtained. A training dataset was constructed based on the soil layer parameters and the pile test parameters.

[0020] Optionally, in step 2, the model building module builds a high-pressure jet grouting pile diameter prediction model based on the fusion of BRNN and GRU, specifically:

[0021] The model construction module constructs a BRNN and GRU fusion model based on BRNN and GRU, namely the high-pressure rotary jet pile diameter prediction model. Among them, the BRNN and GRU fusion model connects two opposite hidden layers to the same output. Based on generative deep learning, the output layer receives information from both forward and backward directions.

[0022] Optionally, in step 3, the model training module trains the high-pressure jet grouting pile diameter prediction model using the training data set, specifically:

[0023] Step 301: The model training module obtains a training data set and screens soil layer parameters and pile test parameters to select valid data parameters so that the BRNN network in the high-pressure jet grouting pile diameter prediction model consists of 300 hidden layers;

[0024] Step 302: Set input variables, including injection parameters, increment time, soil depth, and porosity. The injection parameters include the number of rotations per lift step Rot, flow rate Q, number of nozzles N, nozzle diameter d, injection time per lift step Dt, average rotation speed w, and injected grout volume V. j ' and lifting speed v, the output is the column diameter at a specific depth, where for soil layer parameters, all diameter values ​​collected from the first six columns are arranged as a soil layer training vector sequence, and for pile test parameters, all diameter values ​​collected from the last six columns are arranged as a test training vector sequence;

[0025] Step 303: Input the soil layer training vector sequence and the test training vector sequence into the high-pressure jet grouting pile diameter prediction model at the corresponding time, and train the high-pressure jet grouting pile diameter prediction model. At this time, the BRNN and GRU in the high-pressure jet grouting pile diameter prediction model are both trained for 300 epochs;

[0026] Step 304: During the training process, if there is an error between the output training result and the preset annotation structure output, the error is transferred step by step to the recurrent neural network of the high-pressure jet grouting pile diameter prediction model through the backward algorithm. The high-pressure jet grouting pile diameter prediction model automatically adjusts the weight parameters of each neuron, and stops training after the accuracy of the training result reaches the preset threshold, thereby completing the training of the high-pressure jet grouting pile diameter prediction model.

[0027] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: the real-time dynamic prediction system and method of the three-dimensional form of high-pressure rotary jet piles provided by the present invention include obtaining a training data set, a model construction module building a high-pressure rotary jet pile diameter prediction model based on BRNN and GRU fusion, a model training module training the high-pressure rotary jet pile diameter prediction model through the training data set, a prediction module predicting based on the trained high-pressure rotary jet pile diameter prediction model to obtain diameter prediction information during the construction process of the construction project, and a high-pressure rotary jet pile diameter output module judging whether the diameter prediction information matches the diameter pattern. If it matches, the diameter prediction information is output; if it does not match, the operating parameters of the high-pressure rotary jet pile diameter prediction model are adjusted and the prediction is re-performed until the diameter prediction information matches the diameter pattern, at which time the diameter prediction information is output; the system and method adopt bidirectional recursive prediction, which can greatly improve the training efficiency, can provide feedback and correction to the data input to the prediction model, can accurately reflect the changes in the actual data of the rotary jet pile over time, can effectively reduce the effectiveness of the high-pressure rotary jet grouting process modeling, and facilitate the proposal of improving the current high-pressure rotary jet pile design based on the analysis and prediction results, thereby reducing the potential risks in engineering practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0029] FIG1 is a schematic structural diagram of a real-time dynamic prediction system for the three-dimensional form of high-pressure jet grouting piles according to an embodiment of the present invention;

[0030] FIG2 is a schematic diagram of a first process flow of a real-time dynamic prediction method for the three-dimensional shape of high-pressure jet grouting piles provided by an embodiment of the present invention;

[0031] 3 is a schematic diagram of a second process flow of a real-time dynamic prediction method for the three-dimensional shape of high-pressure jet grouting piles provided by an embodiment of the present invention;

[0032] Figure 4 is a schematic diagram of the BRNN and GRU fusion model. DETAILED DESCRIPTION

[0033] The purpose of the present invention is to provide a real-time dynamic prediction system and method for the three-dimensional morphology of high-pressure rotary jet grouting piles, which can realize real-time dynamic evaluation of the diameter of high-pressure rotary jet grouting piles, can provide feedback and correction on the input data of the prediction model, and can effectively improve the prediction accuracy.

[0034] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0035] As shown in Figure 1, the real-time dynamic prediction system for the three-dimensional shape of high-pressure rotary jet piles provided by an embodiment of the present invention includes: a model construction module, a model training module, a prediction module and a high-pressure rotary jet pile diameter output module, wherein the model construction module is connected to the model training module, the model training module is connected to the prediction module, and the prediction module is connected to the high-pressure rotary jet pile diameter output module;

[0036] The model building module is used to build a high-pressure jet grouting pile diameter prediction model based on BRNN and GRU fusion;

[0037] The model training module is used to obtain a training data set and train a high-pressure jet grouting pile diameter prediction model based on the training data set;

[0038] The prediction module is used to make predictions based on the trained high-pressure jet grouting pile diameter prediction model to obtain diameter prediction information during the construction process of the building project;

[0039] The high-pressure rotary jet pile diameter output module is used to determine whether the acquired diameter prediction information matches the diameter pattern. If not, the operating parameters of the high-pressure rotary jet pile diameter prediction model are adjusted and the prediction is performed again. If they match, the diameter prediction information is output.

[0040] The present invention uses BRNN and GRU to predict the diameter of high-pressure jet grouting piles. The algorithm can perform a learning process from two directions, combining the relevant parameters of the high-pressure jet grouting piles with the BRNN and GRU fusion model, so that the diameter information of the high-pressure jet grouting piles from previous and future measurements can be saved at any time point, and the data results are more reliable and global.

[0041] The present invention relies on both the preceding and following databases to predict high-pressure jet grouting piles. Furthermore, the fused machine learning model can make timely adjustments to mismatched parameter structures, performing these adjustments because it is able to learn high-level features from the data. This can be considered a unique advancement of deep learning over traditional machine learning techniques. Another significant advantage of deep learning models over traditional machine learning is their ability to handle time series problems. In other words, they can predict future values ​​based on previous observations, addressing issues with feedforward networks that cannot handle the concept of temporal order or recurrent neural networks that are hampered by the vanishing gradient problem.

[0042] As shown in FIG2 and FIG3, the present invention also provides a real-time dynamic prediction method of the three-dimensional shape of high-pressure rotary jet grouting piles, which is applied to the real-time dynamic prediction system of the three-dimensional shape of high-pressure rotary jet grouting piles, comprising the following steps:

[0043] Step 1: Get the training dataset;

[0044] Step 2: The model building module builds a high-pressure jet grouting pile diameter prediction model based on the fusion of BRNN and GRU;

[0045] Step 3: The model training module trains the high-pressure jet grouting pile diameter prediction model using the training data set;

[0046] Step 4: The prediction module performs prediction based on the trained high-pressure jet grouting pile diameter prediction model to obtain diameter prediction information during the construction process of the building project;

[0047] Step 5: The high-pressure rotary jet pile diameter output module determines whether the diameter prediction information matches the diameter pattern. If so, the diameter prediction information is output. If not, the operating parameters of the high-pressure rotary jet pile diameter prediction model are adjusted and the prediction is performed again until the diameter prediction information matches the diameter pattern. At this time, the diameter prediction information is output.

[0048] In step 1, obtain the training data set, specifically:

[0049] Relevant soil data were collected based on the on-site investigation to obtain soil layer parameters. Based on the high-pressure jet grouting pile test, the injection parameters and diameter of the high-pressure jet grouting pile, namely the pile test parameters, were obtained. A training dataset was constructed based on the soil layer parameters and the pile test parameters.

[0050] In step 2, the model building module builds a high-pressure jet grouting pile diameter prediction model based on the fusion of BRNN and GRU, specifically:

[0051] The model construction module constructs a BRNN and GRU fusion model based on BRNN and GRU, namely the high-pressure rotary jet pile diameter prediction model. Among them, the BRNN and GRU fusion model connects two opposite hidden layers to the same output. Based on generative deep learning, the output layer receives information from both forward and backward directions.

[0052] As shown in Figure 4, the BRNN and GRU fusion model is introduced:

[0053] The BRNN and GRU fusion model connects two opposing hidden layers to the same output. Based on generative deep learning, the output layer can simultaneously receive information from both the forward and backward directions. Standard recurrent neural networks have limitations: future data cannot be expressed using the current state. The BRNN and GRU fusion model overcomes this disadvantage by not requiring fixed input data and, at the same time, allowing future input data to be received from the current state.

[0054] The BRNN and GRU fusion model works by dividing the neurons of the regular RNN into two directions: one for the positive time direction (forward state) and the other for the negative time direction (reverse state). The outputs of these two states are not connected to the inputs of the opposite state. Through these two time directions, the past and future of the current time frame can be used as input information. Considering that the current output is related not only to the previous sequence elements but also to the subsequent sequence elements, and the fast convergence speed of this fusion model can accelerate the test process and achieve rapid iteration, it has great practical significance in pile test prediction.

[0055] When in use, the soil layer parameter sequence to be processed in the soil layer parameter is converted into a corresponding vector data sequence through the BRNN and GRU fusion model, and the vector data sequence is input into the BRNN and GRU fusion model at the corresponding time in the model in a forward and reverse manner;

[0056] The pile test parameters are converted into corresponding vector data sequences through the BRNN and GRU fusion model, and the vector data sequences are input into the BRNN and GRU fusion model at the corresponding time in both forward and reverse directions;

[0057] The BRNN and GRU fusion model outputs a vector sequence to be processed, and the high-pressure jet grouting pile diameter output module outputs the high-pressure jet grouting pile diameter according to the BRNN and GRU fusion model;

[0058] The system structure of the present invention is complete and can automatically convert soil layer parameters and pile test parameters into corresponding sequences recognized by the computer, thereby obtaining the diameter of the high-pressure jet grouting pile. Two data sorting schemes are used to comprehensively investigate the ability of the BRNN and GRU fusion model to dynamically predict the diameter of the high-pressure jet grouting pile.

[0059] Further explanation: The soil layer parameters are passed to the hidden layer of the BRNN and GRU fusion model, wherein the hidden layer is a two-dimensional matrix, and each vector of the matrix corresponds to a soil parameter in the soil layer parameters. The correspondence between the soil layer parameters and the vectors is set when constructing the hidden layer. After the soil layer parameters are processed by the hidden layer, a vector data sequence is formed, and the vector data sequence is input into the BRNN and GRU fusion model;

[0060] The pile test parameters are passed to the hidden layer of the BRNN and GRU fusion model, wherein the hidden layer is a two-dimensional matrix, and each vector of the matrix corresponds to a high-pressure jet grouting pile test diameter and jet grouting parameters. The correspondence between the pile test parameters and the vectors is set when constructing the hidden layer. After the pile test parameters are processed by the hidden layer, a vector data sequence is formed, and the vector data sequence is input into the BRNN and GRU fusion model;

[0061] After the two vector data sequences are predicted and processed by the BRNN and GRU fusion model, they are converted into the corresponding high-pressure jet grouting pile diameter parameters, overcoming the inaccuracy of unidirectional data parameter prediction.

[0062] Specifically, the neural network algorithm formula in the BRNN module is:

[0063] Where f, g are activation functions. Common activation functions include logistic function and hyperbolic tangent function. n}, the expansion length of BRNN is n; Represents matrix concatenation; for time step t, the recurrent unit of RNN: h (t) =f(s (t-1) ,X (t) ,θ); h is the system state of BRNN. From the perspective of dynamical systems, the system state describes the changes of all points in a given space over time steps;

[0064] It should be noted that the BRNN module includes two RNN hidden layers, one forward RNN hidden layer and one backward RNN hidden layer, and the output y is obtained by merging the states of the two hidden layers;

[0065] Specifically, the neural network algorithm formula in the GRU module is:

[0066] z t =σ(W z x t +U z h t-1 +b z )

[0067] r t =σ(W r x t +U r h t-1 +b r )

[0068] When z t= 0, the current state h t and the previous state h t-1 It is a nonlinear relationship; when z t =1, h t and h t-1 It is a linear relationship.

[0069] In step 3, the model training module trains the high-pressure jet grouting pile diameter prediction model using the training data set, specifically:

[0070] Step 301: The model training module obtains a training data set and screens soil layer parameters and pile test parameters to select valid data parameters so that the BRNN network in the high-pressure jet grouting pile diameter prediction model consists of 300 hidden layers;

[0071] Step 302: Set input variables, including injection parameters, increment time, soil depth, and porosity. The injection parameters include the number of rotations per lift step Rot, flow rate Q, number of nozzles N, nozzle diameter d, injection time per lift step Dt, average rotation speed w, and injected grout volume V. j ' and lifting speed v, the output is the column diameter at a specific depth, where for soil layer parameters, all diameter values ​​collected from the first six columns are arranged as a soil layer training vector sequence, and for pile test parameters, all diameter values ​​collected from the last six columns are arranged as a test training vector sequence;

[0072] Step 303: Input the soil layer training vector sequence and the test training vector sequence into the high-pressure jet grouting pile diameter prediction model at the corresponding time, and train the high-pressure jet grouting pile diameter prediction model. At this time, the BRNN and GRU in the high-pressure jet grouting pile diameter prediction model are trained for 300 epochs, wherein the corresponding input includes not only the soil layer parameters and the pile test parameters, but also the vector data in the hidden layer;

[0073] Step 304: During the training process, if there is an error between the output training result and the preset annotation structure output, the error is transferred step by step to the recurrent neural network of the high-pressure jet grouting pile diameter prediction model through the backward algorithm. The high-pressure jet grouting pile diameter prediction model automatically adjusts the weight parameters of each neuron, and stops training after the accuracy of the training result reaches the preset threshold, thereby completing the training of the high-pressure jet grouting pile diameter prediction model. The initial learning rate is set to 0.01. In addition, the gradient threshold is set to 1 to prevent gradient explosion.

[0074] The system of the present invention uses the above-mentioned forward algorithm to transmit the calculation data layer by layer in the BRNN and GRU fusion model, and obtains the recognition diameter prediction data at the output layer. When the prediction result deviates from the labeled result of the training sample, the classic error back propagation algorithm in the neural network is used to adjust the various weights in the neural network. The error back propagation method distributes the error step by step back propagating to all neurons in each layer, obtains the error signal of the neurons in each layer, and then corrects the weight of each neuron. The process of transmitting the calculation data layer by layer through the forward algorithm and gradually modifying the weight of each neuron through the backward algorithm is the training process of the neural network; the above process is repeated until the accuracy of the prediction result reaches the set threshold, and the training is stopped. At this time, the BRNN and GRU fusion model can be considered to have been trained.

[0075] The present invention further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method of the present invention when executing the computer program;

[0076] Preferably, the electronic device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program executes the above steps, and the column diameter prediction is achieved when the processor executes the computer program.

[0077] The present invention provides a real-time dynamic prediction system and method for the three-dimensional form of high-pressure rotary jet grouting piles. The method comprises obtaining a training data set, a model construction module building a high-pressure rotary jet grouting pile diameter prediction model based on a fusion of BRNN and GRU, a model training module training the high-pressure rotary jet grouting pile diameter prediction model using the training data set, a prediction module performing prediction based on the trained high-pressure rotary jet grouting pile diameter prediction model to obtain diameter prediction information during the construction process of a building project, and a high-pressure rotary jet grouting pile diameter output module judging whether the diameter prediction information matches a diameter pattern. If so, the diameter prediction information is output; if not, the operating parameters of the high-pressure rotary jet grouting pile diameter prediction model are adjusted and prediction is performed again until the diameter prediction information matches the diameter pattern, at which point the diameter prediction information is output. The system and method adopt bidirectional recursive prediction, which can greatly improve training efficiency, provide feedback and correction to data input to the prediction model, accurately reflect changes in actual data of the rotary jet grouting pile over time, effectively reduce the effectiveness of modeling of the high-pressure rotary jet grouting process, facilitate the proposal of improving the current high-pressure rotary jet grouting pile design based on the analysis and prediction results, and reduce potential risks in engineering practice.

[0078] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A real-time dynamic prediction system for the three-dimensional shape of high-pressure jet grouting piles, characterized in that: include: A model construction module, a model training module, a prediction module and a high-pressure rotary jet pile diameter output module, wherein the model construction module is connected to the model training module, the model training module is connected to the prediction module, and the prediction module is connected to the high-pressure rotary jet pile diameter output module; The model building module is used to build a high-pressure jet grouting pile diameter prediction model based on BRNN and GRU fusion; The model training module is used to obtain a training data set and train a high-pressure jet grouting pile diameter prediction model according to the training data set; The prediction module is used to make predictions based on the trained high-pressure jet grouting pile diameter prediction model to obtain diameter prediction information during the construction process of the building project; The high-pressure rotary jet grouting pile diameter output module is used to determine whether the acquired diameter prediction information matches the diameter pattern. If not, the operating parameters of the high-pressure rotary jet grouting pile diameter prediction model are adjusted and the prediction is performed again. If they match, the diameter prediction information is output.

2. A real-time dynamic prediction method for the three-dimensional shape of high-pressure jet grouting piles, applied to the real-time dynamic prediction system for the three-dimensional shape of high-pressure jet grouting piles as claimed in claim 1, characterized in that: The steps include: Step 1: Get the training data set; Step 2: The model building module builds a high-pressure jet grouting pile diameter prediction model based on BRNN and GRU fusion; Step 3: The model training module trains the high-pressure jet grouting pile diameter prediction model through the training data set; Step 4: The prediction module performs prediction based on the trained high-pressure jet grouting pile diameter prediction model to obtain diameter prediction information during the construction process of the building project; Step 5: The high-pressure rotary jet grouting pile diameter output module determines whether the diameter prediction information matches the diameter pattern. If so, the diameter prediction information is output. If not, the operating parameters of the high-pressure rotary jet grouting pile diameter prediction model are adjusted and the prediction is performed again until the diameter prediction information matches the diameter pattern. At this time, the diameter prediction information is output.

3. The real-time dynamic prediction method of the three-dimensional morphology of high-pressure jet grouting piles according to claim 2 is characterized in that: In step 1, obtain the training data set, specifically: Relevant soil data were collected based on the on-site investigation to obtain soil layer parameters. The injection parameters and diameter of the high-pressure jet grouting piles, i.e., the pile test parameters, were obtained based on the high-pressure jet grouting pile test. A training data set was constructed based on the soil layer parameters and the pile test parameters.

4. The real-time dynamic prediction method of the three-dimensional morphology of high-pressure jet grouting piles according to claim 3 is characterized in that: In step 2, the model building module builds a high-pressure jet grouting pile diameter prediction model based on BRNN and GRU fusion, specifically: The model building module constructs a BRNN and GRU fusion model based on BRNN and GRU, that is, a high-pressure rotary jet pile diameter prediction model. The BRNN and GRU fusion model connects two opposite hidden layers to the same output. Based on generative deep learning, the output layer receives information from both forward and backward directions.

5. The real-time dynamic prediction method of the three-dimensional shape of high-pressure jet grouting piles according to claim 4 is characterized in that: In step 3, the model training module trains the high-pressure jet grouting pile diameter prediction model through the training data set, specifically: Step 301: The model training module obtains a training data set, and screens soil layer parameters and pile test parameters, and screens effective data parameters therein, so that the BRNN network in the high-pressure jet grouting pile diameter prediction model consists of 300 hidden layers; Step 302: Set input variables, including injection parameters, increment time, soil depth and porosity, where the injection parameters include the number of revolutions per lifting step Rot, flow rate Q, number of nozzles N, nozzle diameter d, injection time per lifting step Dt, average rotation speed w, injected grouting volume V j ' and lifting speed v, the output is the column diameter at a specific depth, where for soil layer parameters, all diameter values ​​collected from the first six columns are arranged as a soil layer training vector sequence, and for pile test parameters, all diameter values ​​collected from the last six columns are arranged as a test training vector sequence; Step 303: input the soil layer training vector sequence and the test training vector sequence into the high-pressure jet grouting pile diameter prediction model at the corresponding time, and train the high-pressure jet grouting pile diameter prediction model. At this time, the BRNN and GRU in the high-pressure jet grouting pile diameter prediction model are trained for 300 epochs; Step 304: During the training process, if there is an error between the output training result and the preset annotation structure output, the error is transmitted step by step to the recurrent neural network of the high-pressure rotary jet pile diameter prediction model through the backward algorithm. The high-pressure rotary jet pile diameter prediction model automatically adjusts the weight parameters of each neuron, and stops training after the accuracy of the training result reaches the preset threshold, thereby completing the training of the high-pressure rotary jet pile diameter prediction model.

Images