Deep learning-based intelligent control method and system for freezing top-to-top penetration

Abstract

The invention provides an intelligent control method and system for freezing top-to-top penetration based on deep learning, and relates to the technical field of deep learning, and the method comprises the steps: determining the position of a machine head of a tube push bench through obtaining the parameters, images, environment data and construction target data of the tube push bench at a construction site, and generating a propulsion path based on a path planning module. And in the construction process, temperature and thickness distribution of the closure section position is obtained, the historical freezing scheme is adjusted, and a closure section freezing construction scheme is obtained. And as the frozen wall reaches the preset condition, machine head construction schemes are constructed based on the first information, and the machine head construction schemes comprise disassembly, welding and waterproof construction schemes. And finally, construction is conducted based on the machine head construction scheme until the waterproof effect reaches the preset standard, a closure section reinforced concrete and pressure supplementing grouting scheme is constructed according to construction data, and finally construction is completed. According to the invention, the efficiency and precision of freezing opposite-vertex through construction are obviously improved.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology, and in particular to an intelligent control method and system for freezing top penetration based on deep learning. Background Art

[0002] In the current field of underground engineering construction, especially in the process of pipe jacking construction, precise control of top penetration technology is particularly important. In the existing technology, the position of the machine head and path planning are usually determined by manual experience and simple automatic control systems. This method has significant limitations. For example, the subjectivity of manual operation leads to insufficient positioning accuracy, and the inaccuracy of path planning often leads to construction errors, affecting the quality of the project. In addition, the utilization rate of environmental data and machine head status data during the construction process is low, resulting in insufficient response capabilities to unforeseen problems, ultimately causing construction progress delays and cost increases. In order to address these problems, some existing technologies have introduced simple sensor monitoring and automated control methods, but due to the lack of support from deep learning and intelligent control technologies, these methods are still limited in adaptability and decision-making capabilities under complex working conditions, and it is difficult to meet the construction needs under high-precision and complex geological conditions.

[0003] Therefore, there is an urgent need for an intelligent control method and system for freezing top penetration based on deep learning to solve the above problem. Summary of the invention

[0004] The purpose of the present invention is to provide an intelligent control method and system for freezing top penetration based on deep learning to improve the above problems. In order to achieve the above purpose, the technical solutions adopted by the present invention are as follows:

[0005] In a first aspect, the present application provides an intelligent control method for freezing top penetration based on deep learning, comprising:

[0006] Acquire first information, wherein the first information includes parameter data, image data, environmental data, and construction target data of all pipe jacking machines at the construction site;

[0007] Determine the head position of the pipe jacking machine based on the first information, and send the determined head position and the first information to a path planning module for path planning to obtain the propulsion path information of the pipe jacking machine;

[0008] The construction is carried out based on the advancement path information of the pipe jacking machine until the heads of the two corresponding pipe jacking machines reach the preset closing section position, the temperature distribution and thickness distribution of the closing section position are obtained, and the preset historical freezing scheme is adjusted based on the temperature distribution and thickness distribution of the closing section position to obtain the freezing construction scheme of the closing section position;

[0009] Based on the freezing construction plan of the closing section position, construction is performed until the freezing wall at the closing section position reaches a preset condition, and a head construction plan is constructed based on the first information, wherein the head construction plan includes a head disassembly plan, a head shell welding plan, and a head waterproof construction plan;

[0010] Construction is carried out based on the nose construction plan until the waterproof effect of the nose shell reaches a preset threshold, and a reinforced concrete construction plan and a pressure-added grouting plan for the closing section are constructed based on the first information. Construction is then carried out in sequence according to the reinforced concrete construction plan and the pressure-added grouting plan for the closing section to obtain the final construction result.

[0011] In the second aspect, the present application also provides an intelligent control system for freezing top penetration based on deep learning, including:

[0012] An acquisition unit, used to acquire first information, wherein the first information includes parameter data, image data, environmental data and construction target data of all pipe jacking machines at the construction site;

[0013] A planning unit, configured to determine the head position of the pipe jacking machine based on the first information, and send the determined head position and the first information to a path planning module for path planning to obtain the propulsion path information of the pipe jacking machine;

[0014] An adjustment unit is used to perform construction based on the advancement path information of the pipe jacking machine until the heads of the two corresponding pipe jacking machines reach the preset closing section position, obtain the temperature distribution and thickness distribution of the closing section position, and adjust the preset historical freezing scheme based on the temperature distribution and thickness distribution of the closing section position to obtain a freezing construction scheme of the closing section position;

[0015] A construction unit is used to perform construction based on the frozen construction plan of the closing section position until the frozen wall at the closing section position reaches a preset condition, and to construct a head construction plan based on the first information, wherein the head construction plan includes a head disassembly plan, a head shell welding plan, and a head waterproof construction plan;

[0016] The processing unit is used to perform construction based on the machine head construction plan until the waterproof effect of the machine head shell reaches a preset threshold, construct a closing section reinforced concrete construction plan and a pressure-added grouting plan based on the first information, and perform construction in sequence according to the closing section reinforced concrete construction plan and the pressure-added grouting plan to obtain a final construction result.

[0017] The beneficial effects of the present invention are:

[0018] The present invention obtains the pipe jacking machine parameters, image data, environmental data and construction target data at the construction site, uses a deep learning algorithm to accurately locate the head position of the pipe jacking machine, and performs path planning based on a multi-objective optimization algorithm. The freezing construction plan of the closing section is optimized by dynamically adjusting the temperature distribution and thickness data obtained during the construction process. In addition, this method also combines advanced computing methods such as the analytic hierarchy process, particle swarm optimization algorithm and Bayesian optimization algorithm to construct a complete head disassembly, welding, waterproofing construction plan and a closing section reinforced concrete construction plan and a pressure-compensating grouting plan. Compared with the prior art, the present invention can greatly improve construction accuracy and efficiency, and significantly reduce construction risks and costs.

[0019] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or be understood by implementing the embodiments of the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 It is a schematic flow chart of an intelligent control method for freezing top penetration based on deep learning described in an embodiment of the present invention;

[0022] Figure 2 It is a schematic diagram of the structure of the intelligent control system for freezing top penetration based on deep learning described in an embodiment of the present invention.

[0023] In the figure: 701, acquisition unit; 702, planning unit; 703, adjustment unit; 704, construction unit; 705, processing unit. DETAILED DESCRIPTION

[0024] 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 are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0025] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0026] Embodiment 1:

[0027] This embodiment provides an intelligent control method for freezing top penetration based on deep learning.

[0028] See also Figure 1 , the figure shows that the method includes step S1, step S2, step S3, step S4 and step S5.

[0029] Step S1, obtaining first information, wherein the first information includes parameter data, image data, environmental data and construction target data of all pipe jacking machines at the construction site;

[0030] It is understandable that obtaining the first information is the basis of the entire intelligent control method. This information includes the parameter data, image data, environmental data and construction target data of the pipe jacking machine, covering the comprehensive information set required in the construction process. Specifically, the parameter data of the pipe jacking machine includes the specifications, performance indicators, real-time operating status, etc. of the machine, which can provide a basis for subsequent head position determination and path planning. The image data captures real-time images of the pipe jacking machine and the surrounding environment through high-precision cameras and sensors, and uses computer vision technology for processing and analysis to ensure visual monitoring of key parts during the construction process.

[0031] Environmental data involves key environmental variables such as temperature and humidity, soil properties, and groundwater levels at the construction site. This information has a direct impact on path planning and the selection of construction methods. Construction target data covers construction plans, design drawings, engineering specifications, etc. These data ensure the consistency of the construction process with the established goals. Through the comprehensive acquisition and analysis of these data, a panoramic control of the construction site can be achieved, providing reliable data support for each subsequent step.

[0032] Step S2, determining the head position of the pipe jacking machine based on the first information, and sending the determined head position and the first information to a path planning module for path planning to obtain the propulsion path information of the pipe jacking machine;

[0033] It is understandable that this step can significantly reduce the path deviation of the pipe jacking machine during construction, optimize the propulsion path, and thus improve construction efficiency, reduce energy consumption and construction costs through accurate head position determination and intelligent path planning. This combination of deep learning and intelligent path planning can greatly improve the level of automation and intelligence of construction in complex environments. In this step, step S2 includes step S21, step S22, step S23, step S24 and step S25.

[0034] Step S21, performing dimensionality reduction processing on the pipe jacking machine image data and environmental data acquired at the construction site to obtain the head image data and head environmental data of the pipe jacking machine;

[0035] It can be understood that this step calculates the intra-class scatter matrix (S_w) and inter-class scatter matrix (S_b) of each type of data for the high-dimensional pipe jacking machine image data and environmental data obtained at the construction site, where S_w represents the degree of dispersion of the same type of data, and S_b represents the degree of dispersion of different types of data. Then, by maximizing the ratio of the inter-class divergence to the intra-class divergence, the linear discriminant analysis algorithm finds a projection direction so that the inter-class distance of the projected data is as large as possible and the intra-class distance is as small as possible. By solving the eigenvalues ​​and eigenvectors, the eigenvectors that can maximize the inter-class separability are selected, and the original high-dimensional data are projected into a low-dimensional space, thereby obtaining the dimensionality reduction results of the head image data and environmental data of the pipe jacking machine. This process effectively extracts the most meaningful features for the differentiation task, greatly reduces the data dimension while retaining important information, and optimizes the computational efficiency of subsequent path planning.

[0036] Step S22: Analyze the head image data and head environment data of the pipe jacking machine based on a preset linear discriminant analysis method to obtain the head position information of the pipe jacking machine;

[0037] It can be understood that this step analyzes the head image data and head environment data of the pipe jacking machine based on the preset linear discriminant analysis method to accurately obtain the head position information of the pipe jacking machine. First, a classification model is constructed by linear discriminant analysis, and the reduced head image data and environmental data are used as input. The linear discriminant analysis method analyzes the intra-class and inter-class divergence of the data to find the optimal linear combination, so that the separability between different classes is maximized, thereby mapping the data to a new low-dimensional space. In this space, the classes that were originally difficult to distinguish are more separated in the new projection direction, and such processing greatly improves the accuracy of classification. Subsequently, the head position of the pipe jacking machine is determined by calculating the distance from each data point to the classification boundary in the reduced-dimensional data space. The uniqueness of this method is that it can effectively reduce the dimension of the data while maintaining the key information of the data and improving the efficiency of data processing and analysis. The technical effect is reflected in the precise positioning of the head position, ensuring the accuracy and reliability of path planning, and facilitating the smooth development of subsequent construction processes.

[0038] Step S23, constructing a three-dimensional model with the head position information, environmental data and construction target data of the pipe jacking machine to obtain first sub-information, wherein the first sub-information includes the three-dimensional coordinates of the head position, the three-dimensional coordinates of the environment and the three-dimensional coordinates of the construction target;

[0039] It can be understood that this step integrates the head position information, environmental data and construction target data of the pipe jacking machine, and generates the first sub-information by constructing a three-dimensional model. Specifically, using three-dimensional modeling technology, the acquired head position, environmental characteristics and construction target parameters are mapped to points or surfaces in three-dimensional space respectively to form a comprehensive model with spatial relationships. First, the head position data is parsed into three-dimensional coordinates, indicating the actual position in the construction space; then, the environmental data including surrounding terrain, obstacles and other information is modeled as a three-dimensional structure consistent with the actual scene; finally, the construction target data is converted into the three-dimensional coordinates and spatial range of the target position. These data are integrated into a unified three-dimensional model through coordinate transformation and spatial registration technology, and the generated first sub-information not only contains the spatial coordinates of each element, but also reflects the spatial relationship between them. This process ensures the accuracy and adaptability of the construction path planning of the pipe jacking machine, provides a complete three-dimensional reference framework, and lays the foundation for subsequent path planning and other construction decisions. The technical effect is reflected in the comprehensive capture of the spatial layout of the construction site, which is conducive to accurate construction path planning and risk prediction.

[0040] Step S24, mapping the first sub-information to a one-dimensional sequence corresponding to a preset Hilbert curve, and generating a propulsion path of the pipe jacking machine by obtaining the one-dimensional sequence through mapping;

[0041] It can be understood that the Hilbert curve in this step is a curve that continuously fills the space and has the ability to map multidimensional data to one-dimensional space while maintaining the spatial proximity of the data as much as possible. Specifically, the three-dimensional coordinates of the head position, environment and construction target in the first sub-information are first converted into discrete three-dimensional grid points. Then, by constructing the Hilbert curve, these grid points are arranged in the path order of the curve, so as to map the three-dimensional data into a one-dimensional space. This one-dimensional sequence not only simplifies the data structure, but also retains the geometric relationship in the original three-dimensional space. Next, by analyzing the continuity and change trend in the one-dimensional sequence, the propulsion path of the pipe jacking machine is generated, which can better fit the actual environment during construction and optimize the path length and obstacle avoidance ability. The technical effect is to utilize the space-preserving characteristics of the Hilbert curve to effectively reduce the data dimension and reduce the computational complexity, while ensuring the accuracy and reliability of the construction path.

[0042] Step S25: Perform a feasibility analysis on the propulsion path of the pipe jacking machine based on a convex hull algorithm to obtain a propulsion path of the pipe jacking machine with the shortest propulsion distance.

[0043] It can be understood that this step restores the one-dimensional sequence path data generated in the above steps to a set of path points in three-dimensional space. Then, the convex hull algorithm is applied to construct the minimum convex polyhedron covering all path points. This convex hull represents the boundary range of all possible paths. In this process, by extracting the vertices of the convex hull, the extreme points and turning points in the path are identified, and then the total length of the path is calculated and optimized. By comparing the lengths of all possible paths, the path with the shortest advancement distance is selected as the final advancement path. The technical effect of this method is that the convex hull algorithm can quickly and efficiently process multidimensional data and identify key points in the path, ensuring that the pipe jacking machine selects the optimal path in a complex construction environment, reducing construction time and resource consumption, and improving the accuracy and efficiency of path planning.

[0044] Step S3, performing construction based on the pipe jacking machine advancement path information until the heads of the two corresponding pipe jacking machines reach the preset closing section position, obtaining the temperature distribution and thickness distribution of the closing section position, and adjusting the preset historical freezing scheme based on the temperature distribution and thickness distribution of the closing section position to obtain a freezing construction scheme for the closing section position;

[0045] It can be understood that this step dynamically adjusts the freezing construction plan through accurate temperature monitoring and thickness analysis, thereby improving the effect and accuracy of freezing construction, reducing construction risks, and effectively avoiding construction problems caused by uneven temperature or soil changes. In this step, step S3 includes step S31, step S32, step S33, step S34 and step S35.

[0046] Step S31, constructing a finite element model based on the geological data information of the preset closing section position, wherein the attributes and boundary conditions of the geological data are defined to obtain a completed finite element model;

[0047] It can be understood that in this step, first, a finite element model is constructed based on the geological data information of the preset closing section position. The core purpose of this process is to accurately simulate the physical properties of the soil around the closing section and the complex processes such as heat conduction and frozen wall formation during the construction process. The first step in constructing the finite element model is to extract the geological data of the closing section, including physical properties such as soil type, density, humidity, thermal conductivity, and information such as groundwater distribution that affects the freezing process.

[0048] On this basis, the boundary conditions of the model are defined. The boundary conditions include the temperature field distribution at the closing section, the temperature of the freezing medium during the freezing construction process, the temperature of the surrounding environment, etc. These boundary conditions will affect the heat conduction equation in the model, thereby determining the thickness of the frozen wall, the expansion speed, and its interaction with the surrounding soil. The setting of boundary conditions usually needs to be combined with on-site environmental monitoring data and historical construction experience to ensure that the model can reflect the actual situation during the construction process as accurately as possible.

[0049] The heat conduction equation is as follows:

[0050]

[0051] in, is the Laplace operator of the temperature field, which represents the spatial second-order partial derivative of the temperature field and describes the propagation of heat in space. α is the thermal diffusion coefficient, which represents the ability of heat diffusion and is usually determined by the thermal conductivity and specific heat capacity of the material. T represents the temperature field, t represents the freezing construction time, and Q represents the heat source term.

[0052] By defining the above geological properties and boundary conditions, a finite element model is established. This model can not only simulate the formation process of the frozen wall, but also predict the temperature change, heat conduction efficiency and the formation speed of the frozen wall during the freezing process, providing theoretical support for the subsequent optimization of the construction plan. Through high-precision finite element simulation, various uncertainties in the freezing construction can be predicted in advance, providing a scientific basis for the adjustment and optimization of the construction plan, and ensuring the efficiency and safety of the frozen top penetration construction.

[0053] Step S32, meshing the established finite element model, and inputting the temperature information collected from each mesh into the finite element model to solve the temperature field by the heat conduction equation, so as to obtain the temperature field distribution information of the closing section;

[0054] It can be understood that this step first meshes the established finite element model. The purpose of this process is to discretize the continuous physical space into multiple small finite elements so as to perform accurate numerical calculations on the physical processes inside each element. The fineness of the meshing directly affects the accuracy of the simulation results and the efficiency of the calculation. Generally speaking, the finer the mesh, the higher the solution accuracy can be, but the amount of calculation will also increase accordingly. Next, the temperature information collected from each grid is input into the finite element model to solve the heat conduction equation. The heat conduction equation is the basic equation that describes the propagation of heat in a medium, and its form is usually a partial differential equation for time and space variables. In this process, the heat conduction equation is solved by inputting the initial temperature conditions of each grid (such as the soil temperature and the ambient temperature before construction) and the boundary conditions (such as the temperature of the freezing medium and the temperature change during the freezing construction).

[0055] The model calculates the temperature field distribution information of the closing section and displays the temperature distribution in the entire closing section. These temperature field information can intuitively reflect the formation process of the frozen wall and reveal the temperature change trend of each area in the soil or rock. In this way, construction personnel can understand which areas have better freezing effects and which areas may have temperature deviations, so as to adjust the freezing plan in time.

[0056] Step S33, selecting a temperature node related to the frozen wall based on the temperature field distribution information of the closing section, and performing regression analysis based on preset historical frozen wall data to obtain a mathematical relationship between the temperature and the frozen wall thickness;

[0057] The mathematical relationship is as follows:

[0058]

[0059] Where d(t) represents the thickness of frozen soil, represents the gradient of the soil temperature field, t represents the freezing construction time, k represents the thermal conductivity of the soil, ρ represents the density of the soil, c represents the specific heat capacity of the soil, T env Indicates the temperature of the surrounding environment, T freeze Indicates the temperature of the freezing medium.

[0060] It is understandable that this step first selects the temperature nodes related to the frozen wall based on the temperature field distribution information of the closing section. The temperature node refers to the specific point or area where the temperature information is collected in the finite element model. In the temperature field distribution of the closing section, the formation of the frozen wall is closely related to the temperature change of these temperature nodes, so it is necessary to select those important temperature nodes that can represent the change of the thickness of the frozen wall. These temperature nodes are located at the boundary or nearby area of ​​the frozen wall, which is the place where the temperature change is most significant. By selecting these nodes, it is possible to ensure that the change of the frozen wall is accurately monitored. Then, regression analysis is performed based on the preset historical frozen wall data. The historical frozen wall data is usually data measured through previous construction experience or experiments, which reflects how the thickness of the frozen wall changes under different temperature conditions. Regression analysis provides an effective mathematical model for predicting the thickness of the frozen wall by combining historical data with current temperature field information. This method not only enhances the accuracy of construction, but also improves the controllability of freezing construction, avoids the instability of the freezing effect caused by temperature fluctuations, and also provides a basis for the subsequent adjustment of the freezing construction plan.

[0061] Step S34, predicting the frozen wall thickness of each grid area based on the mathematical relationship, and determining the optimal freezing parameter combination based on the frozen wall thickness and temperature field distribution information of each grid area, wherein the freezing parameter combination includes a combination of freezing medium temperature, freezing time and freezing spacing;

[0062] It can be understood that this step can calculate the thickness of the frozen wall in each area by inputting the temperature information of each grid area into the mathematical model. Next, based on the frozen wall thickness and temperature field distribution information of each grid area, the optimal freezing parameter combination is determined. The optimal freezing parameter combination is to ensure that the frozen wall reaches the expected thickness and remains stable during the construction process. The freezing parameter combination includes three important parameters: freezing medium temperature, freezing time and freezing spacing. Among them, the freezing medium temperature determines the freezing rate. Too low a temperature may lead to too fast freezing, while too high a temperature may lead to incomplete freezing; the freezing time affects the continuity of the freezing process, and the appropriate freezing time helps to form a stable frozen wall; the freezing spacing is related to the layout and freezing efficiency of the freezing equipment. Reasonable spacing can improve the freezing effect and avoid energy waste. In order to determine the optimal freezing parameter combination, this step adopts a multi-objective optimization method, comprehensively considering factors such as frozen wall thickness, construction efficiency, and energy consumption, and looking for the optimal solution under these conditions.

[0063] It can be understood that the optimization objective function in this step is as follows:

[0064]

[0065] Among them, A represents the optimal solution, w1 represents the weight coefficient of the frozen wall thickness, d(t) represents the thickness of the frozen wall, and w 2 Represents the weight coefficient of construction time, T construction represents the total time required for construction, w 3 Represents the weight coefficient of energy consumption, Q freeze (T freeze ) represents the flow rate of the refrigeration medium, t freeze represents the time interval of the freezing process, and dt represents the small time increment in the integration.

[0066] This step can achieve precise control of the thickness of the frozen wall and optimize the selection of freezing parameters through the application of mathematical relationship prediction and optimization algorithms, thereby improving the accuracy and stability of freezing construction. In this way, the uncertainty in the construction process can be greatly reduced, and construction delays or waste of resources caused by uneven freezing or long freezing time can be avoided, achieving a more efficient and economical freezing construction process. At the same time, this also provides a basis for real-time adjustment and monitoring of the construction site to ensure the smooth achievement of construction goals.

[0067] Step S35: Send the optimal freezing parameter combination and historical freezing schemes to a preset neural network structure for learning and training, and predict the freezing construction scheme for the closing section position.

[0068] It can be understood that the neural network in this step is a deep neural network, and the structure of the deep neural network includes an input layer, a hidden layer, and an output layer, in which the neurons in each layer are connected to the neurons in the previous and next layers. In this step, the input layer receives data on the optimal freezing parameter combination and historical freezing schemes, which include freezing medium temperature, freezing time, freezing spacing, historical freezing construction data, etc. The hidden layer abstracts and learns the complex relationship between these input data through multiple nonlinear transformations, and finally the output layer predicts the freezing construction plan for the closing section position, including the predicted freezing parameters and construction steps. Among them, by taking the optimal freezing parameter combination and the historical freezing scheme as input, the deep neural network adjusts its weights and biases through the back propagation algorithm to minimize the error between the predicted value and the actual demand.

[0069] The deep neural network will continuously adjust the model parameters during the training process. By learning the rules in historical data, it can not only adapt to different construction environments, but also provide reasonable solutions for freezing construction in new environments. This deep learning-based approach makes the formulation of freezing construction plans more intelligent, avoiding the limitations of relying on experience and manual adjustments.

[0070] Step S4, performing construction based on the freezing construction plan at the closing section position until the freezing wall at the closing section position reaches a preset condition, and constructing a head construction plan based on the first information, wherein the head construction plan includes a head disassembly plan, a head shell welding plan, and a head waterproof construction plan;

[0071] It can be understood that through the construction and implementation of the above three solutions, the disassembly, welding and waterproofing construction process of the machine head can be effectively guaranteed to be completed smoothly, and technical support can be provided for subsequent construction links. In this step, step S4 includes step S41, step S42, step S43, step S44, step S45 and step S46.

[0072] Step S41, obtaining the head structure data information of the parameter data of the pipe jacking machine in the first information, and using the hierarchical analysis method to decompose the preset head disassembly task into a multi-level structure to obtain a hierarchical model of the disassembly steps;

[0073] It can be understood that this step first extracts the parameter data of the pipe jacking machine from the first information, especially the structural data information of the machine head. These structural data include the geometric shape of the machine head, the size of each part, the weight distribution, and the connection method with other parts. This information is the basis for the design of the disassembly task, ensuring that the disassembly plan can adapt to the structural characteristics of the machine head and be carried out effectively.

[0074] Next, the preset head disassembly task is decomposed into a multi-level structure using the analytic hierarchy process. The analytic hierarchy process is a multi-objective decision-making method that can assign a relative weight to each disassembly step based on factors such as importance, difficulty, and time requirements. In the specific operation, all the steps in the disassembly process are first identified, such as disassembling the head shell, disassembling internal components, cleaning and inspecting each component, etc. Then, these steps are sorted and divided into multiple levels to ensure that each step has the right priority and execution order. For example, the first level of the disassembly task may be to disassemble the shell, while the second level may be to deal with the internal electrical system or pipeline connection.

[0075] The analytic hierarchy process quantifies the weights of different dismantling tasks through expert judgment and mathematical models to form a hierarchical structure model. This structure model can not only help optimize the dismantling process and reduce unnecessary operations, but also improve the efficiency and safety of the dismantling process. It can provide a clear task allocation and implementation framework for subsequent construction and adjustments.

[0076] Step S42: Calculate the weight of each disassembly task step based on the preset relative importance weight of the nose disassembly task, and sort each nose disassembly task according to the weight calculation result, wherein the tasks are sorted in descending order of relative importance weight to obtain the nose disassembly plan;

[0077] It is understandable that in this step, the expert team first compares the preset nose disassembly tasks and evaluates the role of each step in the entire disassembly process. For example, some steps may be crucial in the disassembly process, such as ensuring that important internal components are not damaged when disassembling the nose shell, while other steps may be secondary, such as cleaning work or detail adjustments. Through the hierarchical analysis method, experts assign an importance weight to each disassembly task, reflecting the impact of the step on safety, efficiency or successful completion of the task during the disassembly process. Or directly assign an importance weight to each task based on historical data.

[0078] Next, based on these weight values, all disassembly tasks are sorted in order of relative importance from large to small. The sorting results will provide a clear priority for the head disassembly, ensuring that the most critical steps are executed first. The order and priority of each step will directly affect the efficiency and accuracy of the disassembly process. For example, if the step of disassembling the head shell has a greater weight than other steps, then the disassembly of the head shell will be performed first during the disassembly process.

[0079] Through this process, the obtained head disassembly plan not only ensures the efficient execution of the disassembly task, but also minimizes errors and risks. The sorted disassembly plan helps construction personnel understand the disassembly step arrangement, thereby improving the smoothness and safety of the operation. The technical effect is reflected in optimizing the disassembly process, saving time, and improving the overall construction efficiency through reasonable task allocation, reducing the occurrence of mechanical damage and personnel accidents.

[0080] Step S43, obtaining the geometric shape of the pipe jacking machine and preset welding requirement information in the image data of the pipe jacking machine in the first information;

[0081] It is understandable that image data is usually collected by a high-resolution camera or laser scanner, covering detailed images of various parts of the pipe jacking machine. In the image processing stage, computer vision technology is used for feature extraction. Through edge detection, contour recognition, image segmentation and other technologies, the appearance information and internal structure of the pipe jacking machine are extracted to construct a geometric model of the head. This process is implemented by the YOLOV3 algorithm adopted in the present invention, and this step also combines the preset welding requirement information with the geometric shape data. These welding requirement information generally include welding methods (such as TIG welding, MIG welding, etc.), welding quality standards, joint types (such as fillet welds, butt welds, etc.), and temperature and pressure requirements during welding. Welding requirements can be extracted from historical data, construction standards and design specifications. The welding requirements are integrated with the head geometry data to ensure that the welding scheme can meet the requirements of various aspects such as structural strength, durability and ease of operation.

[0082] Step S44, constructing a welding path model based on the geometric shape of the pipe jacking machine and preset welding requirement information, wherein a plurality of welding path information is obtained by defining the starting point, end point and constraint conditions of the path;

[0083] It can be understood that this step analyzes the geometric shape data of the pipe jacking machine to clarify the morphological characteristics of the welding part, especially the shape, size and possible complex geometry of the connection area. The head structure of the pipe jacking machine usually includes multiple connection parts, such as shell, support, joints, etc. In this process, computational geometry and three-dimensional modeling technology are used to construct the welding area model of the head according to the external and internal structure of the pipe jacking machine. Next, define the starting point and end point of the welding path. The starting point of the welding path usually selects the starting position of the joint, and the end point is the position where the welding ends. When selecting these starting and end points, it is necessary to ensure that the welding operation can cover the entire joint area and meet the preset welding requirements, such as welding depth, weld width, etc. In addition, in order to ensure welding quality and construction efficiency, the path model also needs to define a series of constraints. These constraints include welding speed, welding temperature range, welding angle, joint type, and external environmental factors (such as climate, pressure, etc.). These constraints ensure that the welding process can be carried out within a safe and accurate range to avoid welding defects.

[0084] Step S45, optimizing and selecting all welding path information based on the particle swarm optimization algorithm, wherein each particle represents a path selection, iteratively evaluating the fitness of each path, calculating its optimization degree, until the optimal welding path information is obtained, and using the optimal welding path information as the welding solution for the head shell;

[0085] It can be understood that the particle swarm optimization algorithm in this step regards each welding path as a "particle". Each particle represents a possible path selection and has a position (path selection) and a speed (path adjustment direction). Initially, all particles are randomly distributed in the search space, and the fitness of each particle is evaluated by a preset objective function. The goal of fitness evaluation is to optimize the various performances of the welding path, such as welding accuracy, welding speed, welding cost, etc.

[0086] It can be understood that the fitness function in this step is as follows:

[0087] f(X)=a·P precision +b·V welding +c·C welding

[0088] Where X is the position of the particle, represents the choice of welding path, P precision is the welding accuracy, the deviation between the calculated path and the target path. The higher the accuracy, the smaller the value. V weldingis the welding speed, C welding is the welding cost, a is the weight coefficient of welding accuracy, b is the weight coefficient of welding speed, and c is the weight coefficient of welding cost.

[0089] During the optimization process, each particle adjusts its current position based on its own historical optimal position and the optimal position of all particles in the group. At each iteration, the particle group will gradually converge, and the particles will focus on the optimal path selection. The fitness of each particle will be continuously evaluated, and its personal best position and the group best position will be updated. Through multiple rounds of iterations, the particle group will eventually converge to an optimal solution, that is, the optimal welding path. During the iteration process of the particle swarm optimization algorithm, the degree of optimization of the path is quantified by the objective function. The objective function usually includes multiple dimensions, such as the distance of the welding path, welding time, welding quality, energy consumption, possible interference during welding, etc. By comprehensively considering these factors, the particle swarm algorithm can assign a fitness value to each path, and then select the path with the lowest comprehensive cost and the highest welding quality.

[0090] Step S46, selecting all preset waterproof materials based on the environmental data of the pipe jacking machine in the first information, wherein the waterproof material with the lowest cost and meeting the preset conditions is selected as the final selected waterproof material, and the preset construction plan corresponding to the final selected waterproof material is used as the machine head waterproof construction plan.

[0091] It is understandable that this step first requires screening the preset waterproof materials according to the environmental data. Each waterproof material has different performance parameters, such as compressive strength, anti-penetration ability, chemical stability, temperature and moisture resistance, etc. Through these performance indicators, materials that do not meet the requirements of environmental conditions can be filtered out. For example, if the soil on site has high chemical corrosivity, some materials may be excluded due to insufficient corrosion resistance. Next, by selecting waterproof materials that are cost-effective and meet the requirements of environmental conditions, combined with the preset construction budget and resource conditions, the cost-effectiveness of the materials is further evaluated. At this time, cost-effectiveness considerations are very important because there are differences in the procurement and construction costs of different materials. By comprehensively considering the technical performance and economy of waterproof materials, the waterproof material with the highest cost-effectiveness is finally selected.

[0092] Step S5, performing construction based on the machine head construction plan until the waterproof effect of the machine head shell reaches a preset threshold, constructing a closing section reinforced concrete construction plan and a pressure-added grouting plan based on the first information, and performing construction in sequence according to the closing section reinforced concrete construction plan and the pressure-added grouting plan to obtain the final construction result.

[0093] It is understandable that this step, combined with the optimization of the reinforced concrete construction plan of the closing section and the pressure grouting plan, can effectively improve construction efficiency, reduce construction risks, and ensure that the project quality meets the expected standards. Ultimately, the entire construction process ensures that the project is completed as planned while ensuring construction safety and quality, providing a solid foundation for subsequent construction stages. In this step, step S5 includes step S51, step S52, and step S53.

[0094] Step S51, extracting geological parameters, environmental parameters and construction target data of the closing section location based on environmental data;

[0095] It is understandable that by combining the extraction of geological parameters, environmental parameters and construction target data, it can provide all-round information support for the subsequent optimization and execution of the construction plan. This step ensures that the formulation of the construction plan is more scientific and reasonable through precise data analysis and parameter extraction, and avoids problems caused by ignoring certain parameters in actual construction. Ultimately, these data will help the construction team make more flexible and precise adjustments when facing a complex construction environment, ensuring the smooth progress of the entire project.

[0096] Step S52: construct a multi-objective optimization function based on geological conditions, environmental parameters and construction target data at the closing section, wherein the multi-objective optimization model takes construction intensity, construction cost and construction time as optimization targets;

[0097] It can be understood that the construction strength in this step mainly refers to the structural strength requirements during the construction process, ensuring that the materials and processes used in the construction can meet the engineering safety requirements. For the construction of the closing section, the strength not only involves the compressive strength of reinforced concrete, but also the supporting capacity of the surrounding soil or rock formation. Differences in geological conditions, such as soil density, groundwater level and stability of rock formations, will directly affect the required construction strength. By incorporating construction strength into the optimization function, it can be ensured that the use of materials during the construction process can meet the structural safety requirements. Construction cost is one of the key factors of the entire project, covering the purchase cost of construction materials, labor costs, equipment use costs, etc. Different construction methods and technical choices will have different effects on costs. For example, the use of high-strength materials and advanced equipment may improve construction efficiency, but it will also increase the cost of single materials and equipment. Therefore, optimizing construction costs means minimizing unnecessary costs such as material waste and equipment downtime while ensuring construction strength and construction time. Construction time is an important indicator for evaluating construction progress and directly affects the delivery cycle of the project. Reasonable arrangement of construction sequence, optimization of construction technology, and reduction of waiting and downtime during construction can effectively reduce construction time. Especially in complex geological environments and extreme climatic conditions, the optimization of construction time is particularly important. By introducing construction time as an optimization target, it is possible to avoid waste of resources and increase in construction costs due to excessive construction periods. By constructing and solving multi-objective optimization functions, the construction team can consider geological conditions, environmental factors, and project goals while comprehensively weighing the pros and cons of various aspects, and ultimately obtain an optimal construction plan.

[0098] Step S53, solving the multi-objective optimization function to obtain the optimal reinforced concrete construction plan for the closing section.

[0099] It can be understood that in this step, when solving the multi-objective optimization problem of the reinforced concrete construction plan of the closing section, the initial solution is randomly generated through the simulated annealing algorithm and gradually adjusted to find the optimal balance of objectives such as construction intensity, construction cost and construction time.

[0100] First, initialize a solution S 0 , and set the initial temperature T 0 And the cooling parameter α. Then, based on the current solution S, a neighborhood solution S' is generated, and the energy difference ΔE = E(S') - E(S) is calculated, where E(S) is the objective function value of the current solution and E(S') is the objective function value of the new solution. If the new solution S' is better (ΔE < 0), it is directly accepted; otherwise, it is accepted with probability Accept inferior solutions to avoid falling into local optimal solutions. As the temperature T gradually decreases, the algorithm gradually converges to the global optimal solution. The stopping condition of simulated annealing is usually when the temperature reaches the set minimum value or the number of iterations reaches the upper limit, and finally outputs the optimal solution S*, such as a construction plan that balances cost, strength and time. By continuously adjusting the weights and parameters of the objective function, the simulated annealing algorithm can provide an effective global search solution in multi-objective optimization to find the optimal construction path and material configuration.

[0101] In this step, step S5 also includes step S54, step S55 and step S56.

[0102] Step S54, extracting geological parameters, crack distribution and water pressure information of the closing section based on the environmental data, and using the geological parameters, crack distribution and water pressure information as input variables;

[0103] It is understandable that this step uses the extracted geological parameters, fracture distribution and water pressure information as input variables to provide accurate basic data for the subsequent pressure grouting scheme. These input variables will help build a detailed underground environment model to ensure that the pressure grouting scheme can effectively cope with the instability of the underground structure and water pressure changes in actual construction, thereby optimizing the key parameters such as the amount, pressure and flow rate of grouting to ensure the smooth progress of construction. Through this process, the construction environment under underground conditions can be accurately simulated, providing theoretical support for subsequent grouting operations and freezing construction, and ensuring the safety and effectiveness of the construction plan.

[0104] Step S55, constructing an initial grouting model based on preset historical pressure-replenishing grouting data, and defining grouting volume, grouting pressure and grouting flow rate parameters;

[0105] It can be understood that this step is obtained through the analysis of past projects, such as groundwater infiltration, crack expansion and underground stability, etc., to help establish a preliminary grouting model. In the process of building the initial grouting model, three key parameters need to be defined: grouting volume, grouting pressure and grouting flow rate. Grouting volume refers to the total amount of grouting liquid required in each grouting process, which directly affects the construction effect and material cost. Grouting pressure refers to the pressure applied during the grouting process.

[0106] Step S56: input the input variables into the initial grouting model for processing, wherein the initial grouting model is optimized based on the Bayesian optimization algorithm to predict the optimal pressure-replenishing grouting construction plan.

[0107] It is understandable that in this step, Bayesian optimization is a method of global optimization through a probability model, which is usually used to solve optimization problems with high dimensions, high computational costs, and no explicit analytical solutions. In this step, the Bayesian optimization algorithm establishes a probabilistic relationship between input variables and output effects (the relationship between grouting effects and grouting parameters), and infers the optimal parameter combination based on this relationship. Bayesian optimization can gradually find the parameter combination that can maximize the grouting effect by continuously updating and improving the probability model with a small number of trials.

[0108] In the Bayesian optimization process, this step first defines the Gaussian process model, and then uses the model to predict the output of the initial grouting model. By optimizing the proxy model to minimize the expected loss target, Bayesian optimization can effectively determine which parameter combinations are most likely to bring the best grouting effect and provide the optimal pressure-replenishing grouting solution for actual construction.

[0109] Through this optimization method, not only can the number of experiments be reduced and construction costs be lowered, but also the accuracy and effect of construction can be improved, ensuring the best grouting effect in a complex construction environment and meeting the requirements of geological conditions, crack distribution and other aspects to the greatest extent.

[0110] Embodiment 2:

[0111] like Figure 2 As shown, this embodiment provides an intelligent control system for freezing top penetration based on deep learning, see Figure 2 The system includes an acquisition unit 701 , a planning unit 702 , an adjustment unit 703 , a construction unit 704 and a processing unit 705 .

[0112] An acquisition unit 701 is used to acquire first information, wherein the first information includes parameter data, image data, environmental data and construction target data of all pipe jacking machines at the construction site;

[0113] A planning unit 702 is used to determine the head position of the pipe jacking machine based on the first information, and send the determined head position and the first information to a path planning module for path planning to obtain the propulsion path information of the pipe jacking machine;

[0114] The adjustment unit 703 is used to perform construction based on the advancement path information of the pipe jacking machine until the heads of the two corresponding pipe jacking machines reach the preset closing section position, obtain the temperature distribution and thickness distribution of the closing section position, and adjust the preset historical freezing scheme based on the temperature distribution and thickness distribution of the closing section position to obtain the freezing construction scheme of the closing section position;

[0115] A construction unit 704 is used to perform construction based on the frozen construction plan at the closing section position until the frozen wall at the closing section position reaches a preset condition, and to construct a head construction plan based on the first information, wherein the head construction plan includes a head disassembly plan, a head shell welding plan, and a head waterproof construction plan;

[0116] Processing unit 705 is used to perform construction based on the nose construction plan until the waterproof effect of the nose shell reaches a preset threshold, construct a closing section reinforced concrete construction plan and a pressure-added grouting plan based on the first information, and perform construction in sequence according to the closing section reinforced concrete construction plan and the pressure-added grouting plan to obtain the final construction result.

[0117] It should be noted that, regarding the system in the 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 elaborated here.

[0118] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0119] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. An intelligent control method for freezing top penetration based on deep learning, characterized in that: include: Acquire first information, wherein the first information includes parameter data, image data, environmental data, and construction target data of all pipe jacking machines at the construction site; Determine the head position of the pipe jacking machine based on the first information, and send the determined head position and the first information to a path planning module for path planning to obtain the propulsion path information of the pipe jacking machine; The construction is carried out based on the advancement path information of the pipe jacking machine until the heads of the two corresponding pipe jacking machines reach the preset closing section position, the temperature distribution and thickness distribution of the closing section position are obtained, and the preset historical freezing scheme is adjusted based on the temperature distribution and thickness distribution of the closing section position to obtain the freezing construction scheme of the closing section position; Based on the freezing construction plan of the closing section position, construction is performed until the freezing wall at the closing section position reaches a preset condition, and a head construction plan is constructed based on the first information, wherein the head construction plan includes a head disassembly plan, a head shell welding plan, and a head waterproof construction plan; Construction is carried out based on the nose construction plan until the waterproof effect of the nose shell reaches a preset threshold, and a reinforced concrete construction plan and a pressure-added grouting plan for the closing section are constructed based on the first information. Construction is then carried out in sequence according to the reinforced concrete construction plan and the pressure-added grouting plan for the closing section to obtain the final construction result.

2. The intelligent control method for freezing top penetration based on deep learning according to claim 1 is characterized in that , based on the first information, the head position of the pipe jacking machine is determined, and the determined head position and the first information are sent to the path planning module for path planning to obtain the propulsion path information of the pipe jacking machine, including: Perform dimensionality reduction processing on the pipe jacking machine image data and environmental data acquired at the construction site to obtain the head image data and head environmental data of the pipe jacking machine; Based on the preset linear discriminant analysis method, the head image data and head environment data of the pipe jacking machine are analyzed to obtain the head position information of the pipe jacking machine; The head position information, environmental data and construction target data of the pipe jacking machine are used to construct a three-dimensional model to obtain first sub-information, wherein the first sub-information includes the three-dimensional coordinates of the head position, the three-dimensional coordinates of the environment and the three-dimensional coordinates of the construction target; Mapping the first sub-information to a one-dimensional sequence corresponding to a preset Hilbert curve, and generating a propulsion path of the pipe jacking machine by obtaining the one-dimensional sequence through mapping; The feasibility analysis of the propulsion path of the pipe jacking machine is performed based on the convex hull algorithm to obtain the propulsion path of the pipe jacking machine with the shortest propulsion distance.

3. The intelligent control method for freezing top penetration based on deep learning according to claim 1 is characterized in that , obtain the temperature distribution and thickness distribution of the closing section, and adjust the preset historical freezing scheme based on the temperature distribution and thickness distribution of the closing section to obtain the freezing construction scheme of the closing section, including: A finite element model is constructed based on the geological data information of the preset closing section position, wherein the attributes and boundary conditions of the geological data are defined to obtain a completed finite element model; The established finite element model is meshed, and the temperature information collected from each mesh is input into the finite element model to solve the temperature field by the heat conduction equation, so as to obtain the temperature field distribution information of the closing section; Based on the temperature field distribution information of the closing section, the temperature nodes related to the frozen wall are selected, and regression analysis is performed based on the preset historical frozen wall data to obtain the mathematical relationship between the temperature and the frozen wall thickness; Predicting the frozen wall thickness of each grid area based on the mathematical relationship, and determining an optimal freezing parameter combination based on the frozen wall thickness and temperature field distribution information of each grid area, wherein the freezing parameter combination includes a combination of freezing medium temperature, freezing time and freezing spacing; The optimal freezing parameter combination and historical freezing scheme are sent to a preset neural network structure for learning and training, and a freezing construction scheme for the closing section position is predicted.

4. The intelligent control method for freezing top penetration based on deep learning according to claim 1 is characterized in that Based on the first information, a head construction plan is constructed, wherein the head construction plan includes a head disassembly plan, a head shell welding plan, and a head waterproof construction plan, including: Obtaining the machine head structure data information of the parameter data of the pipe jacking machine in the first information, and using the hierarchical analysis method to decompose the preset machine head disassembly task into a multi-level structure to obtain a hierarchical model of the disassembly steps; The weight of each disassembly task step is calculated based on the preset relative importance weight of the nose disassembly task, and each nose disassembly task is sorted according to the weight calculation result, wherein the relative importance weight is sorted from large to small to obtain the nose disassembly plan; Acquire the geometric shape of the pipe jacking machine and preset welding requirement information in the image data of the pipe jacking machine in the first information; A welding path model is constructed based on the geometric shape of the pipe jacking machine and preset welding requirement information, wherein a plurality of welding path information is obtained by defining the starting point, end point and constraint conditions of the path; All welding path information is optimized and selected based on the particle swarm optimization algorithm, where each particle represents a path selection. The fitness of each path is evaluated iteratively, and its optimization degree is calculated until the optimal welding path information is obtained. The optimal welding path information is used as the welding solution for the head shell. Based on the environmental data of the pipe jacking machine in the first information, all preset waterproof materials are selected, wherein the waterproof material with the lowest cost and meeting the preset conditions is selected as the final selected waterproof material, and the preset construction plan corresponding to the final selected waterproof material is used as the machine head waterproof construction plan.

5. The intelligent control method for freezing top penetration based on deep learning according to claim 1 is characterized in that ,Based on the first information, a reinforced concrete construction plan and a pressure grouting plan for the closing section are constructed, and the construction is carried out in sequence according to the reinforced concrete construction plan and the pressure grouting plan for the closing section to obtain the final construction results, including: Extract geological parameters, environmental parameters and construction target data of the closing section location based on environmental data; A multi-objective optimization function is constructed based on geological conditions, environmental parameters and construction target data of the closure section, wherein the multi-objective optimization model takes construction intensity, construction cost and construction time as optimization targets; The multi-objective optimization function is solved to obtain the optimal reinforced concrete construction plan for the closing section.

6. An intelligent control system for freezing top penetration based on deep learning, characterized in that: include: An acquisition unit, used to acquire first information, wherein the first information includes parameter data, image data, environmental data and construction target data of all pipe jacking machines at the construction site; A planning unit, configured to determine the head position of the pipe jacking machine based on the first information, and send the determined head position and the first information to a path planning module for path planning to obtain the propulsion path information of the pipe jacking machine; An adjustment unit is used to perform construction based on the advancement path information of the pipe jacking machine until the heads of the two corresponding pipe jacking machines reach the preset closing section position, obtain the temperature distribution and thickness distribution of the closing section position, and adjust the preset historical freezing scheme based on the temperature distribution and thickness distribution of the closing section position to obtain a freezing construction scheme of the closing section position; A construction unit is used to perform construction based on the frozen construction plan of the closing section position until the frozen wall at the closing section position reaches a preset condition, and to construct a head construction plan based on the first information, wherein the head construction plan includes a head disassembly plan, a head shell welding plan, and a head waterproof construction plan; The processing unit is used to perform construction based on the machine head construction plan until the waterproof effect of the machine head shell reaches a preset threshold, construct a closing section reinforced concrete construction plan and a pressure-added grouting plan based on the first information, and perform construction in sequence according to the closing section reinforced concrete construction plan and the pressure-added grouting plan to obtain a final construction result.

7. The deep learning-based intelligent control system for freezing top penetration according to claim 6 is characterized in that: The planning unit comprises: The first planning subunit is used to perform dimensionality reduction processing on the pipe jacking machine image data and environmental data acquired at the construction site to obtain the head image data and head environmental data of the pipe jacking machine; The second planning subunit is used to analyze the head image data and head environment data of the pipe jacking machine based on a preset linear discriminant analysis method to obtain the head position information of the pipe jacking machine; The third planning subunit is used to construct a three-dimensional model with the head position information, environmental data and construction target data of the pipe jacking machine to obtain first sub-information, wherein the first sub-information includes the three-dimensional coordinates of the head position, the three-dimensional coordinates of the environment and the three-dimensional coordinates of the construction target; A fourth planning subunit is used to map the first sub-information to a one-dimensional sequence corresponding to a preset Hilbert curve, and generate a propulsion path of the pipe jacking machine by obtaining the one-dimensional sequence through mapping; The fifth planning subunit is used to perform a feasibility analysis on the propulsion path of the pipe jacking machine based on a convex hull algorithm to obtain a propulsion path of the pipe jacking machine with the shortest propulsion distance.

8. The deep learning-based intelligent control system for freezing top penetration according to claim 6 is characterized in that: The adjustment unit comprises: The first adjustment subunit is used to construct a finite element model based on the geological data information of the preset closing section position, wherein the attributes and boundary conditions of the geological data are defined to obtain the established finite element model; The second adjustment subunit is used to mesh the established finite element model, and input the temperature information collected from each mesh into the finite element model to solve the temperature field by the heat conduction equation, so as to obtain the temperature field distribution information of the closing section; The third adjustment subunit is used to select the temperature node related to the frozen wall based on the temperature field distribution information of the closing section, and perform regression analysis based on the preset historical frozen wall data to obtain the mathematical relationship between the temperature and the frozen wall thickness; a fourth adjustment subunit, for predicting the frozen wall thickness of each grid area based on the mathematical relationship, and determining an optimal freezing parameter combination based on the frozen wall thickness and temperature field distribution information of each grid area, wherein the freezing parameter combination includes a combination of freezing medium temperature, freezing time and freezing spacing; The fifth adjustment subunit is used to send the optimal freezing parameter combination and historical freezing scheme to a preset neural network structure for learning and training, and predict the freezing construction scheme for the closing section position.

9. The deep learning-based intelligent control system for freezing top penetration according to claim 6 is characterized in that: The construction unit comprises: The first construction subunit is used to obtain the head structure data information of the parameter data of the pipe jacking machine in the first information, and use the hierarchical analysis method to decompose the preset head disassembly task into a multi-level structure to obtain a hierarchical model of the disassembly steps; The second construction subunit is used to calculate the weight of each disassembly task step based on the preset relative importance weight of the nose disassembly task, and sort each nose disassembly task according to the weight calculation result, wherein the order of relative importance weight is from large to small, and the nose disassembly plan is obtained; A third construction subunit is used to obtain the geometric shape of the pipe jacking machine and preset welding requirement information in the image data of the pipe jacking machine in the first information; The fourth construction subunit is used to construct a welding path model based on the geometric shape of the pipe jacking machine and preset welding requirement information, wherein a plurality of welding path information is obtained by defining the starting point, the end point and the constraint conditions of the path; The fifth construction subunit is used to optimize and select all welding path information based on the particle swarm optimization algorithm, wherein each particle represents a path selection, and the fitness of each path is iteratively evaluated to calculate its optimization degree until the optimal welding path information is obtained, and the optimal welding path information is used as the welding solution for the head shell; The sixth construction subunit is used to select all preset waterproof materials based on the environmental data of the pipe jacking machine in the first information, wherein the waterproof material with the lowest cost and meeting the preset conditions is selected as the final selected waterproof material, and the preset construction plan corresponding to the final selected waterproof material is used as the machine head waterproof construction plan.

10. The deep learning-based intelligent control system for freezing top penetration according to claim 6 is characterized in that: The processing unit comprises: A first processing subunit is used to extract geological parameters, environmental parameters and construction target data of the closing section location based on the environmental data; A second processing subunit is used to construct a multi-objective optimization function based on geological conditions, environmental parameters and construction target data of the closing section, wherein the multi-objective optimization model takes construction intensity, construction cost and construction time as optimization targets; The third processing subunit is used to solve the multi-objective optimization function to obtain the optimal reinforced concrete construction plan for the closing section.

Images