A tunnel shotcrete construction method and device based on machine learning optimization

Abstract

This application belongs to the field of tunnel construction technology and specifically discloses a tunnel shotcrete construction method and device based on machine learning optimization. The method includes: performing quality deviation analysis based on the tunnel design profile and tunnel point cloud data to determine the tunnel's over-break and under-break; constructing a dynamic deposition model for shotcrete based on the nozzle's spray trajectory, the dynamic deposition model is used to determine the deposition thickness at any position on the spray plane based on the nozzle's spray parameters; solving the objective function using an improved particle swarm optimization algorithm to obtain the nozzle's target spray parameters, the objective function minimizing the difference between the deposition thickness and the over-break and under-break as the optimization goal. The improved particle swarm optimization algorithm is determined by introducing a quantum random jump mechanism into the particle swarm optimization algorithm; and regulating the spraying equipment based on the nozzle's target spray parameters. Through this application, it is possible to effectively model and optimize the shotcrete construction process.

Description

Technical Field

[0001] The present application belongs to the field of tunnel construction technology, and more specifically, relates to a tunnel shotcrete construction method and device based on machine learning optimization. Background Art

[0002] Shotcrete is a crucial process for the initial support of mountain tunnels. Shotcrete construction is primarily performed using remotely controlled wet shotcrete machines. Due to a lack of precise modeling and optimization of the shotcrete process, construction quality is highly dependent on the operator's skill level, which can easily lead to problems such as concrete waste and uneven shotcrete thickness. Furthermore, over-excavation and under-excavation, common in drill-and-blast tunnel construction, further exacerbate construction difficulties and impact the overall support and stability of the tunnel.

[0003] In recent years, optimizing spraying parameters to achieve optimal spraying results has gained increasing attention. Some studies have experimentally analyzed the impact of spraying parameters on deposition patterns. However, traditional experimental methods lack flexibility and ease of use. Effectively modeling and optimizing the shotcrete construction process remains a pressing technical challenge in this field. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the purpose of this application is to effectively model and optimize the construction process of shotcrete.

[0005] To achieve the above objectives, in a first aspect, the present application provides a tunnel shotcrete construction method based on machine learning optimization, the method comprising:

[0006] Based on the tunnel design outline and tunnel point cloud data, quality deviation analysis is performed to determine the tunnel over-excavation and under-excavation amount;

[0007] A dynamic deposition model of shotcrete is constructed based on the nozzle's spray trajectory. The dynamic deposition model is used to determine the deposition thickness at any position on the spray plane based on the nozzle's spray parameters. The spray trajectory is the motion trajectory of the nozzle center.

[0008] The improved particle swarm optimization (IQPSO) algorithm is used to solve the objective function and obtain the target injection parameters of the nozzle. The objective function takes minimizing the difference between the deposition thickness and the over-excavation and under-excavation as the optimization goal. The improved particle swarm optimization algorithm is determined by introducing the quantum random jump mechanism into the particle swarm optimization algorithm.

[0009] The spray equipment is regulated based on the target spray parameters of the nozzle.

[0010] It can be understood that optimizing the spraying parameters through the IQPSO algorithm can minimize the difference between the deposition thickness and the over-excavation and under-excavation, significantly improve the accuracy of shotcrete construction, and reduce the deposition thickness error. The optimized spraying trajectory and parameters can improve construction efficiency, reduce manual intervention, improve the stability and consistency of spraying construction, shorten construction time, accurately control the spraying thickness and material usage, effectively reduce concrete waste, and effectively model and optimize the construction process of shotcrete.

[0011] In one possible implementation, the above-mentioned introduction of the quantum random jump mechanism includes updating the position of the particle by the following formula;

[0012] ;

[0013] ;

[0014] in, It is The particle in The optimal position of an individual in the iteration, is a hyperparameter, The particle is The best individual position in the iteration, It is in The global optimal position of the particle swarm in the iteration; For the In the iteration The position of the particle, For The location determined after the location update, represents the average value of the optimal solution of all particles at present, is uniformly distributed A random value of and are the upper and lower boundaries of the search space, respectively. Indicates the maximum number of iterations.

[0015] In one possible implementation, elastic rebound processing is also performed on out-of-bounds particles using the following formula:

[0016] ;

[0017] in, Indicates the In the iteration The position of the particle, and Respectively represent The upper and lower boundaries of each particle.

[0018] In one possible implementation, the spray parameters to be optimized include: the spacing between two adjacent control points, the distance between two adjacent paths (scanning distance), the nozzle's dwell time at each control point, and the nozzle's movement speed between two adjacent control points; a control point is a point on a spray trajectory, and a spray trajectory includes multiple paths arranged side by side;

[0019] Accordingly, the objective function is solved by the improved particle swarm optimization (IQPSO) algorithm to obtain the target injection parameters of the nozzle, including the following two-stage optimization:

[0020] In the first stage of optimization, the objective function is solved by improving the particle swarm optimization algorithm to optimize the interval distance between two adjacent control points and the distance between two adjacent paths;

[0021] In the second stage of optimization, the objective function is solved by improving the particle swarm optimization algorithm to optimize the nozzle's residence time at each control point and the nozzle's movement speed between two adjacent control points.

[0022] In one possible implementation, the objective function is specifically:

[0023] ;

[0024] in, Indicates the distance between two adjacent control points. Represents the distance between two adjacent paths. Indicates the time the nozzle stays at each control point, Indicates that the nozzle is The dwell time at each control point, Indicates the movement speed of the nozzle between two adjacent control points, Indicates that the nozzle is The control point and The speed of movement between control points, represents the number of control points, represents the number of evaluation points, Indicates the number of the evaluation point, Indicates the evaluation points The sediment thickness at Indicates the evaluation points The over-excavation and under-excavation at is a fixed value, is an integer, express The upper limit value of express The upper limit value of and They are The lower and upper limits of and They are lower and upper limits.

[0025] In one possible implementation, the quality deviation analysis based on the tunnel design profile and tunnel point cloud data to determine the tunnel overbreak and underbreak includes:

[0026] Reduce the density and noise of tunnel point cloud data through preprocessing;

[0027] Based on the pre-processed tunnel point cloud data, an irregular triangular mesh model is constructed;

[0028] Based on the irregular triangular mesh model, the intersection points of the irregular triangular mesh model and the target section are calculated, and the difference between the intersection points and the tunnel design contour is calculated to determine the over-excavation and under-excavation of the tunnel.

[0029] In a possible implementation, the dynamic deposition model of shotcrete is constructed based on the spray trajectory of the nozzle, specifically including determining the dynamic deposition model using the following formula:

[0030] ;

[0031] ;

[0032] in, Indicates the position on the injection plane The sediment thickness at Indicates the position on the jet plane in the case of a single path The sediment thickness at represents the radial distance of the jet at a specified axial distance, Represents the distance between two adjacent paths. For the path The control point and Any position between the control points, Indicates that the jet is at position The injection rate on middle Indicates the nozzle process parameters of shotcrete, middle represents any position on the jet plane, middle Indicates a point The normal vector of the jet plane at , Indicates the distance between two adjacent control points. Indicates the speed of the nozzle moving between two adjacent control points.

[0033] In a second aspect, the present application provides a tunnel shotcrete construction device based on machine learning optimization, comprising:

[0034] The overbreak and underbreak determination module is used to perform quality deviation analysis based on the tunnel design profile and tunnel point cloud data to determine the overbreak and underbreak of the tunnel;

[0035] A dynamic deposition model building module is used to build a dynamic deposition model of shotcrete based on the nozzle's spray trajectory. The dynamic deposition model is used to determine the deposition thickness at any position on the spray plane based on the nozzle's spray parameters. The spray trajectory is the motion trajectory of the nozzle center.

[0036] The injection parameter optimization module is used to solve the objective function through the improved particle swarm optimization algorithm to obtain the target injection parameters of the nozzle. The objective function takes minimizing the difference between the deposition thickness and the over-excavation and under-excavation as the optimization goal. The improved particle swarm optimization algorithm is determined by introducing the quantum random jump mechanism into the particle swarm optimization algorithm.

[0037] The control module is used to control the injection device based on the target injection parameters of the nozzle.

[0038] In a third aspect, the present application provides an electronic device comprising: at least one memory for storing programs; and at least one processor for executing the programs stored in the memory. When the program stored in the memory is executed, the processor is used to execute the method described in the first aspect or any possible implementation of the first aspect.

[0039] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method described in the first aspect or any possible implementation of the first aspect.

[0040] It can be understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here.

[0041] In general, the above technical solutions conceived by this application have the following beneficial effects compared with the existing technologies:

[0042] This method offers advantages such as high optimization accuracy, high construction efficiency, minimal material waste, and strong applicability. By optimizing shotcrete parameters using an improved particle swarm optimization (IQPSO) algorithm, the accuracy of shotcrete construction is significantly improved, and the error in deposition thickness is reduced. The optimized shotcrete trajectory and parameters enhance construction efficiency, reduce manual intervention, improve the stability and consistency of shotcrete construction, and shorten construction time. Precise control of shotcrete thickness and material usage effectively reduces concrete waste and construction costs. Furthermore, this method and device are adaptable to tunnel construction of varying scales and complexities, including railway, highway, and subway projects, demonstrating strong engineering applicability. Combining 3D laser scanning with a feedback adjustment mechanism enables intelligent control of shotcrete construction, minimizing the impact of manual experience on construction quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is one of the flow charts of the tunnel shotcrete construction method based on machine learning optimization provided in the embodiment of the present application;

[0044] Figure 2 This is the second flow chart of the tunnel shotcrete construction method based on machine learning optimization provided in the embodiment of the present application;

[0045] Figure 3 This is the third flow chart of the tunnel shotcrete construction method based on machine learning optimization provided in the embodiment of the present application;

[0046] Figure 4 is a schematic diagram of a projection model spray-deposited on a tunnel profile provided in an embodiment of the present application;

[0047] Figure 5 Schematic diagram of a projection model when an angle exists between the nozzle and the tunnel profile, provided by an embodiment of the present application;

[0048] Figure 6 is a schematic diagram of a dynamic spray deposition model on a single trajectory provided by an embodiment of the present application;

[0049] Figure 7 is a schematic diagram of a dynamic spray deposition model on multiple trajectories provided in an embodiment of the present application;

[0050] Figure 8 This is a schematic diagram of a vertical injection model provided in an embodiment of the present application;

[0051] Figure 9 This is a convergence process diagram of IQPSO and PSO provided in an embodiment of the present application;

[0052] Figure 10 This is a diagram of the convergence process of IQPSO under different hyperparameters provided in the embodiment of the present application;

[0053] Figure 11 This is a sensitivity analysis diagram of the parameter settings provided in the embodiment of the present application;

[0054] Figure 12 This is a schematic diagram of the locations of evaluation points and control points provided in an embodiment of the present application;

[0055] Figure 13 It is the spray deposition diagram of IQPSO and PSO provided in the embodiment of this application;

[0056] Figure 14 is a relative error diagram of IQPSO and PSO provided in the embodiment of the present application;

[0057] Figure 15 is a residence time diagram of a control point provided in an embodiment of the present application;

[0058] Figure 16 is a velocity diagram between control points provided by an embodiment of the present application;

[0059] Figure 17 Schematic diagram of the structure of a tunnel shotcrete construction device based on machine learning optimization provided in an embodiment of the present application;

[0060] Figure 18 It is a structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0062] In the embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0063] In the description of the embodiments of the present application, unless otherwise specified, "multiple" means two or more, for example, multiple processing units means two or more processing units, etc.; multiple elements means two or more elements, etc.

[0064] The embodiments of the present application are described below in conjunction with the drawings in the embodiments of the present application.

[0065] Figure 1This is one of the flow charts of the tunnel shotcrete construction method based on machine learning optimization provided in the embodiment of the present application, such as Figure 1 As shown, the method includes the following steps S101 to S104.

[0066] Step S101: performing quality deviation analysis based on the tunnel design profile and tunnel point cloud data to determine the amount of overbreak and underbreak of the tunnel;

[0067] Step S102: constructing a dynamic deposition model of shotcrete based on the spraying trajectory of the nozzle. The dynamic deposition model is used to determine the deposition thickness at any position on the spraying plane based on the spraying parameters of the nozzle. The spraying trajectory is the motion trajectory of the nozzle center.

[0068] Step S103, solving the objective function by using an improved particle swarm optimization (IQPSO) algorithm to obtain the target injection parameters of the nozzle. The objective function takes minimizing the difference between the deposition thickness and the over-excavation amount as the optimization goal. The improved particle swarm optimization algorithm is determined by introducing a quantum random jump mechanism into the particle swarm optimization algorithm.

[0069] Step S104: regulating the spraying device based on the target spraying parameters of the nozzle.

[0070] It can be understood that optimizing the spraying parameters through the IQPSO algorithm can minimize the difference between the deposition thickness and the over-excavation and under-excavation, significantly improve the accuracy of shotcrete construction, and reduce the deposition thickness error. The optimized spraying trajectory and parameters can improve construction efficiency, reduce manual intervention, improve the stability and consistency of spraying construction, shorten construction time, accurately control the spraying thickness and material usage, effectively reduce concrete waste, and effectively model and optimize the construction process of shotcrete.

[0071] The following is an illustrative example of the tunnel shotcrete construction method based on machine learning optimization provided by this application.

[0072] Figure 2 This is the second flow chart of the tunnel shotcrete construction method based on machine learning optimization provided in the embodiment of the present application, such as Figure 2 As shown, the method includes the following steps S201, S202, S203 and S204.

[0073] Step S201: Obtain three-dimensional contour data of the tunnel after construction through laser scanning and three-dimensional modeling technology, and parameterize the injection trajectory;

[0074] Step S202: establishing a dynamic deposition model of shotcrete based on the principles of fluid mechanics and pneumatic transport physics;

[0075] Step S203: optimizing injection parameters using an improved particle swarm optimization algorithm;

[0076] Step S204: Adjust the parameters of the spraying equipment in real time according to the optimization results to ensure uniform deposition of concrete and improve construction efficiency.

[0077] In step S201, a laser scanner is used to capture the tunnel outline and generate point cloud data, and its quality deviation is calculated. This process includes the following steps S2011-S2013:

[0078] S2011, Data Preprocessing,The raw point cloud from the complex construction site needs to be carefully preprocessed before reconstruction, including three tasks: (1) manually segmenting the tunnel outline; (2) filtering the point cloud using the random sampling consensus algorithm (RANSAC); (3) downsampling the point cloud using the voxel grid algorithm (VG). Through these steps, a moderately dense and less noisy tunnel outline point cloud can be obtained.

[0079] S2012: Irregular triangular mesh model construction. After removing outliers, the Bowyer-Watson (BW) algorithm is used for Delaunay triangulation. Each point is inserted into the triangulation and updated through local optimization to maintain the Delaunay property. Once all points have been analyzed, the irregular triangular mesh model (i.e., mesh) is complete.

[0080] S2013, Mass Deviation Calculation, developed an automated algorithm to accurately calculate mass deviation. This is achieved by fitting a surface function and calculating the deviation from the design surface function. The process is divided into two stages: (1) calculating the intersection of the irregular triangular mesh model and the cut surface (specific cross section); (2) calculating the difference between the cut surface intersection and the design contour and obtaining the overcut or undercut of the cut surface by integration.

[0081] In step S201, the injection trajectory is the motion trajectory of the nozzle center. The parameterized injection trajectory includes defining the injection trajectory through the following parameters: the tunnel contour grid and its normal vector, the control point and the scanning distance. Generally, the injection trajectory includes multiple paths that are arranged side by side and in order; a path is a straight line, and different paths remain parallel; for two adjacent paths, the nozzle moves from the end of the front path to the beginning of the rear path, and the line between the end of the front path and the beginning of the rear path is perpendicular to the path. Control points refer to discrete points with intervals on the injection trajectory map, at which the nozzle can adjust its kinematic parameters to produce different amounts of deposition. The scanning distance is the distance between two adjacent paths.

[0082] In step S202, the present application develops a static deposition model to describe the spraying behavior during the spraying process, which represents the deposition amount at any point on the spraying plane. To facilitate the description of the static deposition process of tilted spraying, the following assumptions are made during the spraying process:

[0083] (1) Ignore the air pressure loss in the nozzle;

[0084] (2) Ignoring the parallel flow of concrete;

[0085] (3) Ignoring the effect of steel mesh on deposition;

[0086] (4) The residual coefficient (indicating the proportion of concrete remaining adhered to the tunnel wall after rebound) is used to represent the rebound effect of concrete.

[0087] The deposition model is defined as ,in Indicates the nozzle process parameters of shotcrete, including air pressure, spray angle and spray distance; represents any position on the injection plane in the injection coordinate system; Indicates a point Given the complexity of parameterizing deposition models for arbitrary surfaces, this application chooses to simplify the problem into two independent components, as shown below:

[0088] ;

[0089] in, Indicates that the spray distance is In the case of vertical spray to point Virtual plane Deposition on. Indicates that the virtual plane The magnification factor of the projection of the deposition onto the jet plane.

[0090] In step S202, the behavior of shotcrete is described using the principles of fluid mechanics and pneumatic transport physics, and a vertical spray deposition model is established, such as Figure 4 As shown, Figure 4 The x-axis represents the normal direction of the vertical plane, the y-axis represents the horizontal direction on the spraying plane, and the z-axis represents the vertical direction on the spraying plane. It includes two main stages: (1) the acceleration process of concrete in the nozzle, and (2) the diffusion process of concrete in the air. Since the circular motion of particles in the nozzle does not affect the axial velocity, only the force acting in the axial direction is considered. The speed of concrete ejected from the nozzle is mainly determined by the pressure of the compressed air. Friction between nozzle and According to Newton's second law, the cross-sectional area of ​​the nozzle is The force analysis of the concrete cross section is carried out, and the force acting on the concrete aggregate can be expressed by the following formula:

[0091] ;

[0092] in, is the air-material resistance coefficient, , Distance from the jet nozzle The cross-sectional area and diameter of the nozzle at , denote the densities of compressed air and concrete respectively, is the flow resistance coefficient related to the Reynolds number of concrete, and represent the nozzle contraction angle and the incident angle of compressed air, respectively. , , They represent the air flow velocity, concrete transportation velocity and concrete suspension velocity respectively. is the acceleration due to gravity.

[0093] The motion pattern of the concrete in the nozzle is an accelerated motion with decreasing acceleration, followed by a stable uniform motion. When the acceleration is zero, the concrete reaches its maximum velocity, which is equivalent to the ideal injection velocity. , can be calculated using the following formula:

[0094] ;

[0095] in, is the air pressure, is the density of air. The suspension velocity of concrete particles is typically determined semi-empirically through experimental results. For shotcrete with a particle size of approximately 10 mm, this velocity is approximately 12.1 m / s. Based on the Reynolds number of shotcrete, the flow resistance coefficient is set to 0.082.

[0096] During the spraying process, the concrete flow continuously carries the surrounding air, forming a diffuse jet that expands along the spray path. When the concrete flow reaches the tunnel contour along the central axis, its velocity decreases as the spraying distance increases. The semi-empirical relationship between the axial velocity and the distance is described as follows:

[0097] ;

[0098] in, is the axial velocity, is the axial distance, is the nozzle radius.

[0099] In the spray diffusion area, the velocity of the concrete decreases as it moves away from the axis. At the spray boundary (i.e., the end of the diffusion radius), the velocity of the concrete becomes zero. This application assumes that the variances of the spray range in the y and z directions are equal and uncorrelated. Based on this assumption, the relationship between the velocity at any point in the diffusion range and the axis is derived as follows:

[0100] ;

[0101] in, represents the radial distance to the center of the jet in the y and z coordinate planes, and are the horizontal and vertical coordinates of the jet on the plane, is the radial distance of the jet at axial distance x, is located in concrete aggregate velocity.

[0102] During the spraying process, concrete follows the law of conservation of mass. Therefore, the spatial distribution of mass is directly related to the velocity distribution, and the deposition thickness per second at any point on the spraying plane can be expressed by the following formula:

[0103] ;

[0104] in, and represent the density and velocity of concrete respectively, is the gradient of mass flow rate; The residual coefficient of concrete refers to the fact that during the shotcrete process, the concrete sprayed from the nozzle is not completely deposited on the tunnel wall. Due to the influence of various factors such as rebound and gravity during the spraying process, part of the concrete is not completely deposited, but is wasted or lost. It is often used to measure these losses.

[0105] In step S202, in order to obtain the best molding quality, the nozzle should be perpendicular to the tunnel contour. However, due to the inherent limitation of the robot arm's range of motion in the tunnel, it becomes impractical to maintain a completely vertical spray direction. Therefore, when there is an angle between the nozzle and the tunnel contour, it becomes crucial to analyze the spray deposition. The projection model is as follows Figure 5 As shown. On a reference plane perpendicular to the nozzle, the deposition magnification factor is derived through a virtual vertical plane According to the geometric relationship between the three planes, the magnification factor can be expressed as follows:

[0106] ;

[0107] in, is the distance from the nozzle to the virtual vertical plane, is the angle between the nozzle and the contour, is the angle between the nozzle and the cross-section normal.

[0108] In step S202, the single path model mainly represents the deposition change along a continuous single path without considering the influence of surrounding paths. This process can be regarded as a variable speed motion, in which the speed In a certain range The dynamic spray deposition model scenario on a single path is as follows Figure 6 shown. and are two control points on the injection trajectory, and the nozzle residence time is and This application assumes that the nozzle is and Under this assumption, the deposition model at any point can be expressed as follows:

[0109] ;

[0110] in, Indicates a position between two adjacent control points The sediment depth on Indicates that the jet is at position The spray rate (deposition thickness per second) on The spacing between adjacent control points.

[0111] In order to achieve a consistent deposition thickness across the entire spray area, multiple paths need to be overlapped. Figure 7 As shown, the deposition model at any given point can be expressed as the following equation:

[0112] ;

[0113] in, In the multipath model, given location The total sediment thickness at In the single path model, given the location The sediment thickness at Indicates the distance between two adjacent paths (scan distance).

[0114] In step S203, in the mathematical model of the spraying process, there are three types of parameters that define the deposition on the tunnel profile:

[0115] (1) The position of the control point is determined by the spacing distance and the scanning distance;

[0116] (2) Kinematic parameters of the nozzle, including the nozzle's residence time at each control point and the nozzle's movement speed between two adjacent control points;

[0117] (3) Nozzle process parameters, such as air pressure and spray angle.

[0118] These parameters are tightly coupled and can compensate for each other to produce the desired deposition. For example, a smaller spray angle may reduce the distance between adjacent control points while increasing the spray time. Therefore, it is unnecessary to treat all parameters as optimization targets. In this application, we consider the nozzle's process parameters as preset inputs.

[0119] The optimization model is constructed as a two-stage framework. The first stage mainly optimizes the separation distance and scanning distance , determine the position of the control point. For each interval distance and scanning distance The second stage optimization will be carried out to obtain the residence time in this case and speed ,like Figure 8 Based on this, this application develops an IQPSO algorithm to solve this two-stage optimization task.

[0120] In order to evaluate the performance of the model, the injection area is further discretized to generate evaluation points that are denser than the control points. Then, the over-excavation and under-excavation at these evaluation points are obtained through the irregular triangular mesh model of the tunnel. The objective function of the IQPSO algorithm aims to minimize the mean absolute error, which is defined as the difference between the actual deposition and the over-excavation and under-excavation at all evaluation points. The objective function is shown in the following formula. In order to ensure the operability of the solution, this application uses the spacing distance and scanning distance Apply constraints to limit them to fixed values Integer multiples of .

[0121] ;

[0122] in, Indicates the The dwell time at each control point, Indicates that the nozzle is The control point and The speed of movement between control points, represents the number of control points, represents the number of evaluation points, Indicates the number of the evaluation point, Indicates the evaluation points The sediment thickness at Indicates the evaluation points The amount of over-excavation and under-excavation.

[0123] The implementation flow chart of the two-stage optimization model is as follows: Figure 3 As shown. The process first initializes the interval distance and scanning distance , and then initialize the dwell time for each control point and speed . Subsequently, the IQPSO algorithm is applied to iteratively update the dwell time and speed according to its optimization principle. If the maximum number of iterations is reached (i.e., the termination condition), IQPSO determines the optimal dwell time and speed solution corresponding to the current interval distance and scan distance. At this time, the interval distance and scan distance are updated, and a new round of optimization begins. After each iteration, the termination condition is checked: if the maximum number of iterations has not been reached, the interval distance and scan distance, as well as the dwell time and speed, are continued to be updated. Once the maximum number of iterations is reached in both stages, the optimization process ends and the optimal solution is output, minimizing the objective function value.

[0124] The basic particle swarm optimization (PSO) algorithm, developed from the predatory behavior of bird flocks, has two key components: cooperation between individuals and information sharing. Given a swarm of N particles, in a D-dimensional search space, each particle can update its velocity and position using the following formula. The optimal solution for the entire swarm can be determined by the optimal values ​​of all particles.

[0125] ;

[0126] in, , represent the velocity and position of the particle respectively; , They represent the inertia weight and learning factor, which are the hyperparameters of the PSO algorithm; r is a random number ranging from [0,1]; and They represent the individual optimal position and the global optimal position of the particle respectively.

[0127] In the IQPSO algorithm, particles simulate quantum behavior and adopt a new rule to update their positions and velocities. These mechanisms further introduce randomness and enhance the PSO algorithm's search capabilities. The state of a particle is described by a wave function in the Schrödinger equation, which represents the probability of the particle appearing at a given spatial location. Each particle will converge to a region pulled by an attractor, which is calculated as follows:

[0128] ;

[0129] in, The particle is The best individual position in the iteration, It is The particle in The optimal position of an individual in the iteration, is the global optimal position in the particle swarm, It is a parameter used to calculate the weighted average of local optimum and global optimum, which determines the direction of optimization. In order to better control the behavior of IQPSO algorithm, Consider it as a hyperparameter rather than a random value. Based on this concept, we can update the position of the particle as follows:

[0130] ;

[0131] in, It is The position of the particle, is uniformly distributed A random value of represents the average value of the optimal solution of all particles at present, is a control factor that affects the convergence speed. and are the upper and lower boundaries of the search space, respectively. Indicates the maximum number of iterations.

[0132] In order to avoid local optimal solutions, the iterative formula of the boundary control strategy is proposed as follows:

[0133] ;

[0134] in, and Respectively represent The upper and lower boundaries of a particle.

[0135] In step S204, the optimized injection parameters obtained through the above steps are transmitted in real time to the control system of the injection equipment. These equipment can automatically adjust their operating parameters based on the optimization results to ensure uniform concrete deposition on the tunnel surface.

[0136] The following describes a specific application of a tunnel shotcrete construction method based on machine learning optimization. The flowchart of the method in this specific application is as follows: Figure 3 shown.

[0137] 1. Data collection.

[0138] The case study is a single-track, double-track railway tunnel constructed using the drill-and-blast method, with a total length of approximately 2.4 km. Construction data was collected by contractors and researchers, and included three main steps: (1) manual segmentation of the tunnel contours, (2) further processing using the RANSAC and VG algorithms to obtain a sparse and clean point cloud, and (3) mesh reconstruction using the BW algorithm, and the quality deviation was calculated using the algorithm.

[0139] 2. Parameter setting and sensitivity analysis of IQPSO algorithm.

[0140] PSO was selected as the benchmark algorithm to demonstrate the effectiveness and superiority of the IQPSO algorithm. According to the Technical Specifications for Railway Tunnel Construction (TZ 204-2008) and engineering practice, the basic parameters of the injection optimization model are set as shown in Table 1.

[0141] Table 1 Basic parameter settings of the injection optimization model

[0142]

[0143] First, the locations for evaluating the injection effect are determined, which are called evaluation points. The evaluation points are set at fixed intervals in the selected area to ensure that the density of the evaluation points has the ability to fully evaluate the effect of the injection model filling the over-excavation area between the control points. The interval and scanning distance of the evaluation points are set to 0.1 meters and 0.15 meters respectively, and finally 78 evaluation points are set. Then, the IQPSO and PSO algorithms are used to simulate the optimization parameter combination under given conditions, and their performance is compared to demonstrate the superiority of the IQPSO algorithm. The hyperparameters of the two algorithms are determined through preliminary experiments, and their final values ​​are shown in Table 2. The performance of the IQPSO and PSO algorithms in the iterative process is shown in Table 2. Figure 9 As shown in Figure 3, the PSO's initial values ​​are close to the global optimum, but it quickly gets stuck in a local optimum and cannot identify the direction leading to the global optimum. In contrast, although the IQPSO's initial values ​​perform poorly, it continuously resets its position through randomness and a boundary control strategy, and does not get stuck in a local optimum during any iteration.

[0144] Table 2 Hyperparameter settings of IQPSO and PSO

[0145]

[0146] In addition, a univariate sensitivity analysis was conducted on the parameter settings of the IQPSO algorithm (i.e., the number of particles and the average weight). The number of particles was gradually increased from 5 to 25 when the average weight was 0.5, and the average weight was gradually increased from 0.3 to 0.7 when the number of particles was 20. The convergence process under different parameter settings is shown in Figure 2. Figure 10As shown in the figure, 20 particles and an average weight of 0.5 are the most suitable settings for quickly and accurately finding the optimal injection parameter solution. Furthermore, the convergence value of the worst parameter combination differs from the best result by less than 30%, indicating that IQPSO is less sensitive to hyperparameter settings and is able to converge to a neighborhood of the global optimal solution in the vast majority of cases.

[0147] A univariate sensitivity analysis was also performed on the PSO algorithm to further illustrate the superiority of IQPSO. To compare with IQPSO, the parameters of the PSO algorithm were the number of particles and the inertia factor (i.e., the parameter that determines the proportion of particles maintaining their current state, similar to the average weight). The ranges of variation for the number of particles and the inertia factor were [5, 25] and [0.3, 0.7], respectively. When performing the sensitivity analysis on the number of particles / inertia factor, the fixed values ​​of the inertia factor / number of particles were set to 0.7 and 20. Ten simulations were repeated for each parameter combination to reduce the influence of randomness. The results are shown in Figure 2. Figure 11 As shown in the figure, IQPSO achieves more satisfactory results for all parameter combinations. The smaller dispersion of IQPSO results indicates that the algorithm successfully approaches the global optimal solution in each experiment. Notably, the worst solution generated by IQPSO is still better than the best solution generated by PSO, demonstrating the reliability and superiority of IQPSO in optimizing injection models.

[0148] 3. Optimization result analysis.

[0149] The spraying effects after IQPSO and PSO optimization of the spraying parameters were evaluated. The evaluation points were still set at a 0.1-meter interval and a 0.15-meter scanning distance. Figure 12 The optimization results of IQPSO and PSO for determining the control point positions under a given distribution of evaluation points are shown in Table 3. The spacing and scanning distances are shown. Obviously, the evaluation points fully cover the area between the control points and effectively evaluate the performance of the injection model. At the same time, to facilitate the presentation of results, this application defines the naming rules for points and paths. The leftmost point of each group is marked as point 1, and the points on the right are numbered in increasing order. After marking all the points in the bottom row, mark the leftmost point in the second-to-last row, and so on, until all points are marked. Path marking follows the same principle.

[0150] Table 3 Control point position optimization results

[0151]

[0152] The injection performance of IQPSO and PSO is measured by four key performance indicators (KPIs), which can be calculated by the following formulas: (1) coefficient of determination; (2) root mean square error (RMSE); (3) mean absolute error (MAE); and (4) mean absolute percentage error (MAPE). Figure 13and Table 4 show the difference between deposition and overexcavation at each evaluation point, as well as the KPIs of IQPSO and PSO. Figure 13 The sub-graph (a) is the data sample. Figure 13 The sub-graph (b) is a scatter plot.

[0153] ;

[0154] Where n is the number of data points, is the predicted value, is the true value, is the mean of the true values.

[0155] Table 4 KPIs of IQPSO and PSO in injection simulation

[0156]

[0157] Obviously, IQPSO outperforms PSO and achieves satisfactory accuracy. The data points in the overexcavation-deposition coordinate system are close to the straight line y=x, indicating that the two are highly consistent. This is also reflected by Figure 14 The relative error in the results was demonstrated. Based on experience, manual jetting operations typically have an average thickness error of 30 mm, while the IQPSO solution (with an average thickness error of 6 mm) demonstrates its potential in guiding and improving manual jetting. The results demonstrate the reliability of the IQPSO in optimizing the jetting model and establishing effective primary support, thereby enhancing the structural stability of blasting tunnels.

[0158] The residence time of the control points and the speed between the control points are calculated by IQPSO and PSO respectively. The results are as follows: Figure 15 and Figure 16 shown. Figure 15 The subgraph (a) is the residence time graph of the control points corresponding to IQPSO. Figure 15 The subgraph (b) is the residence time graph of the control points corresponding to PSO. Figure 16 The subgraph (a) is the speed diagram between the control points corresponding to IQPSO. Figure 16The (b) subgraph is the velocity diagram between the control points corresponding to the PSO. The total injection time of the IQPSO and PSO schemes is 874 seconds and 814 seconds, respectively. PSO demonstrates superior performance in injection time by determining a smaller number of control points and increasing the scanning distance. However, these sparser control points also limit the optimization of injection parameters. In this case, the strategy of PSO focuses on prioritizing the residence time to match the over-excavation of the control points. When the over-excavation is large, PSO produces more deposition by increasing the residence time while moving at a higher speed on the path to prevent excessive deposition; on the contrary, when the over-excavation of the control point is small, PSO prioritizes reducing the residence time and compensates for the small amount of deposition caused by the shorter residence time by reducing the speed. Therefore, the speed optimization results of PSO are usually distributed near the extreme values ​​(see Figure 16 (b)). This shows that it is difficult for PSO to consider all variables when optimizing such a high-dimensional problem.

[0159] In contrast, IQPSO emphasizes the complementary relationship between injection time and velocity in the deposition process, providing an ideal solution. IQPSO generates 16 more control points than PSO, allowing for more flexible adjustment of injection time and velocity. IQPSO achieves a total absolute error (TAE) of 0.568 meters between injection deposition and overexcavation, surpassing PSO's 0.709 meters. Furthermore, with 16 more control points (and a minimum injection time of 160 seconds), IQPSO's total injection time is only 60 seconds longer than PSO's. This demonstrates that IQPSO effectively balances velocity and injection time, effectively optimizing injection models in blasting tunnels and offering advantages particularly in high-dimensional optimization problems.

[0160] In summary, the tunnel shotcrete construction method based on machine learning optimization provided in this application significantly improves construction quality, reduces material waste, and provides reliable support for intelligent tunnel construction.

[0161] The tunnel shotcrete construction device based on machine learning optimization provided in this application is described below. The tunnel shotcrete construction device based on machine learning optimization described below and the tunnel shotcrete construction method based on machine learning optimization described above can be referenced to each other.

[0162] Figure 17 Schematic diagram of the structure of the tunnel shotcrete construction device based on machine learning optimization provided in the embodiment of the present application, such as Figure 17 As shown, the device includes: an over-excavation and under-excavation determination module 10, a dynamic deposition model construction module 20, an injection parameter optimization module 30 and a control module 40. Among them:

[0163] An overbreak and underbreak determination module 10 is used to perform quality deviation analysis based on the tunnel design profile and tunnel point cloud data to determine the overbreak and underbreak of the tunnel;

[0164] A dynamic deposition model building module 20 is used to build a dynamic deposition model of shotcrete based on the spray trajectory of the nozzle. The dynamic deposition model is used to determine the deposition thickness at any position on the spray plane based on the spray parameters of the nozzle. The spray trajectory is the motion trajectory of the nozzle center.

[0165] The injection parameter optimization module 30 is used to solve the objective function by using the improved particle swarm optimization algorithm to obtain the target injection parameters of the nozzle. The objective function takes minimizing the difference between the deposition thickness and the over-excavation amount as the optimization goal. The improved particle swarm optimization algorithm is determined by introducing a quantum random jump mechanism into the particle swarm optimization algorithm.

[0166] The control module 40 is configured to control the spraying device based on the target spraying parameters of the nozzle.

[0167] It is understandable that the detailed functional implementation of each of the above units / modules can be found in the introduction of the aforementioned method embodiment, and will not be repeated here.

[0168] It should be understood that the above-mentioned device is used to execute the method in the above-mentioned embodiment. The implementation principle and technical effect of the corresponding program module in the device are similar to those described in the above-mentioned method. The working process of the device can refer to the corresponding process in the above-mentioned method and will not be repeated here.

[0169] Based on the method in the above embodiment, an embodiment of the present application provides an electronic device, Figure 18 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application, such as Figure 18 As shown, the electronic device may include: a processor (Processor) 810, a communication interface (Communications Interface) 820, a memory (Memory) 830, and a communication bus 840. The processor 810, the communication interface 820, and the memory 830 communicate with each other via the communication bus 840. The processor 810 may call the logic instructions in the memory 830 to execute the method in the above embodiment.

[0170] In addition, the logic instructions in the aforementioned memory 830 can be implemented in the form of a software functional unit and, when sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application.

[0171] Based on the method in the above embodiment, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method in the above embodiment.

[0172] Based on the method in the above embodiment, an embodiment of the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the method in the above embodiment.

[0173] It is understood that the processor in the embodiments of the present application may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0174] The method steps in the embodiments of the present application can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be an integral part of the processor. The processor and storage medium can be located in an ASIC.

[0175] The above embodiments can be implemented in whole or in part using software, hardware, firmware, or any combination thereof. When implemented using software, they can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions. When loaded and executed on a computer, the computer program instructions fully or partially produce the processes or functions described in the embodiments of this application. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted via the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be magnetic media (e.g., floppy disk, hard disk, tape), optical media (e.g., DVD), or semiconductor media (e.g., solid-state drive (SSD)).

[0176] It will be understood that the various numerical numbers involved in the embodiments of the present application are merely distinctions for the convenience of description and are not intended to limit the scope of the embodiments of the present application.

[0177] It is easy for those skilled in the art to understand that the above is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A tunnel shotcrete construction method based on machine learning optimization, characterized in that: include: Based on the tunnel design outline and tunnel point cloud data, quality deviation analysis is performed to determine the tunnel over-excavation and under-excavation amount; A dynamic deposition model of shotcrete is constructed based on the nozzle's spray trajectory. The dynamic deposition model is used to determine the deposition thickness at any position on the spray plane based on the nozzle's spray parameters. The spray trajectory is the motion trajectory of the nozzle center. The target injection parameters of the nozzle are obtained by solving the objective function through the improved particle swarm optimization algorithm. The objective function takes minimizing the difference between the deposition thickness and the over-excavation and under-excavation as the optimization goal. The improved particle swarm optimization algorithm is determined by introducing the quantum random jump mechanism into the particle swarm optimization algorithm. Control the spray equipment based on the target spray parameters of the nozzle; The introduction of the quantum random jump mechanism includes updating the position of the particle through the following formula; ； ； in, It is The particle in The optimal position of an individual in the iteration, is a hyperparameter, The particle is The best individual position in the iteration, It is in The global optimal position of the particle swarm in the iteration; For the In the iteration The position of the particle, For The location determined after the location update, represents the average value of the optimal solution of all particles at present, is uniformly distributed A random value of and are the upper and lower boundaries of the search space, respectively. represents the maximum number of iterations, is a control factor.

2. The tunnel shotcrete construction method based on machine learning optimization according to claim 1 is characterized in that: It also includes elastic rebound processing for out-of-bounds particles using the following formula: ； in, Indicates the In the iteration The position of the particle, and Respectively represent The upper and lower boundaries of each particle.

3. The tunnel shotcrete construction method based on machine learning optimization according to any one of claims 1-2 is characterized in that: The spray parameters to be optimized include: the distance between two adjacent control points, the distance between two adjacent paths, the nozzle's dwell time at each control point, and the nozzle's movement speed between two adjacent control points; a control point is a point on the spray trajectory, which includes multiple paths arranged side by side; Accordingly, the objective function is solved by improving the particle swarm optimization algorithm to obtain the target injection parameters of the nozzle, including the following two-stage optimization: In the first stage of optimization, the objective function is solved by improving the particle swarm optimization algorithm to optimize the interval distance between two adjacent control points and the distance between two adjacent paths; In the second stage of optimization, the objective function is solved by improving the particle swarm optimization algorithm to optimize the nozzle's residence time at each control point and the nozzle's movement speed between two adjacent control points.

4. The tunnel shotcrete construction method based on machine learning optimization according to claim 3 is characterized in that: The objective function is specifically: ； in, Indicates the distance between two adjacent control points. Represents the distance between two adjacent paths. Indicates the nozzle's residence time at each control point, Indicates that the nozzle is The dwell time at each control point, Indicates the movement speed of the nozzle between two adjacent control points, Indicates that the nozzle is The control point and The speed of movement between control points, represents the number of control points, represents the number of evaluation points, Indicates the number of the evaluation point, Indicates the evaluation points The sediment thickness at Indicates the evaluation points The over-excavation and under-excavation at is a fixed value, is an integer, express The upper limit value of express The upper limit value of and They are The lower and upper limits of and They are lower and upper limits.

5. The tunnel shotcrete construction method based on machine learning optimization according to claim 4 is characterized in that: The quality deviation analysis is performed based on the tunnel design profile and tunnel point cloud data to determine the tunnel overbreak and underbreak, including: Reduce the density and noise of tunnel point cloud data through preprocessing; Based on the pre-processed tunnel point cloud data, an irregular triangular mesh model is constructed; Based on the irregular triangular mesh model, the intersection points of the irregular triangular mesh model and the target section are calculated, and the difference between the intersection points and the tunnel design contour is calculated to determine the over-excavation and under-excavation of the tunnel.

6. The tunnel shotcrete construction method based on machine learning optimization according to claim 4 is characterized in that: The dynamic deposition model of shotcrete is constructed based on the spraying trajectory of the nozzle, specifically including determining the dynamic deposition model by the following formula: ； ； in, Indicates the position on the injection plane The sediment thickness at Indicates the position on the jet plane in the case of a single path The sediment thickness at represents the radial distance of the jet at a specified axial distance, Represents the distance between two adjacent paths. For the path The control point and Any position between the control points, Indicates that the jet is at position The injection rate on middle Indicates the nozzle process parameters of shotcrete, middle represents any position on the jet plane, middle Indicates a point The normal vector of the jet plane at , Indicates the distance between two adjacent control points. Indicates the speed of the nozzle moving between two adjacent control points.

7. A tunnel shotcrete construction device based on machine learning optimization, characterized in that: include: The overbreak and underbreak determination module is used to perform quality deviation analysis based on the tunnel design profile and tunnel point cloud data to determine the overbreak and underbreak of the tunnel; A dynamic deposition model building module is used to build a dynamic deposition model of shotcrete based on the nozzle's spray trajectory. The dynamic deposition model is used to determine the deposition thickness at any position on the spray plane based on the nozzle's spray parameters. The spray trajectory is the motion trajectory of the nozzle center. The injection parameter optimization module is used to solve the objective function through the improved particle swarm optimization algorithm to obtain the target injection parameters of the nozzle. The objective function takes minimizing the difference between the deposition thickness and the over-excavation and under-excavation as the optimization goal. The improved particle swarm optimization algorithm is determined by introducing the quantum random jump mechanism into the particle swarm optimization algorithm. A control module, for controlling the spraying device based on target spraying parameters of the nozzle; The introduction of the quantum random jump mechanism includes updating the position of the particle through the following formula; ； ； in, It is The particle in The optimal position of an individual in the iteration, is a hyperparameter, The particle is The best individual position in the iteration, It is in The global optimal position of the particle swarm in the iteration; For the In the iteration The position of the particle, For The location determined after the location update, represents the average value of the optimal solution of all particles at present, is uniformly distributed A random value of and are the upper and lower boundaries of the search space, respectively. represents the maximum number of iterations, is a control factor.

8. An electronic device, characterized in that: include: at least one memory for storing a computer program; At least one processor is used to execute the program stored in the memory. When the program stored in the memory is executed, the processor is used to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed on a processor, the processor is caused to execute the method according to any one of claims 1 to 6.

Images