Shield tunneling dynamic regulation and control method and system based on multi-source information fusion

Abstract

The invention provides a shield tunneling dynamic regulation and control method and system based on multi-source information fusion. The shield tunneling dynamic regulation and control method comprises the steps that shield state parameters, a ground surface settlement value, a cutterhead abrasion value and a tunneling specific energy value are obtained; the obtained data is preprocessed; based on the preprocessed data, utilizing a pre-trained prediction model to predict and obtain a ground surface settlement prediction value, a cutterhead wear prediction value and a tunneling specific energy prediction value; and by taking the shield construction parameters as decision variables and taking maximization of tunneling specific energy and minimization of ground surface settlement and cutter wear as optimization objectives, performing global optimization within an actual shield tunneling setting constraint range, and determining an optimal solution of the shield construction parameters. According to the invention, intelligent, precise and dynamic regulation and control of the shield tunneling process can be realized.

Description

Technical Field

[0001] The present invention belongs to the field of shield tunneling dynamic control, and in particular relates to a shield tunneling dynamic control method and system based on multi-source information fusion. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Shield machines have become a widely used large-scale construction equipment in urban tunnel construction due to their advantages, including no impact on ground traffic, strong technical and economic efficiency, high degree of automation, and independence from climatic conditions. However, due to the high complexity and variability of geological conditions and the uncertainty of structural equipment, shield machines often enter a "blind" state. During operation, shield machines are often affected by surface conditions, the geological conditions ahead, and the status of the cutterhead equipment, making it impossible to dynamically adjust excavation parameters based on actual conditions.

[0004] With the development of technology, artificial intelligence methods have been applied to the dynamic control of tunneling parameters. However, intelligent models still need to be trained with a large amount of high-quality data for optimization. The acquisition and processing of diverse data are limited by factors, and the generalization ability of the model is affected under different geological conditions and construction environments. It lacks real-time and dynamic characteristics, making it difficult to respond to various emergencies and changes in the tunneling process in a timely manner. Summary of the Invention

[0005] In order to solve the above problems, the present invention proposes a dynamic control method and system for shield tunneling based on multi-source information fusion. The present invention comprehensively considers various information such as geological conditions, cutterhead equipment status and construction operations, and realizes intelligent, precise and dynamic control of the shield tunneling process, so as to improve tunneling efficiency, reduce safety risks and optimize construction costs.

[0006] According to some embodiments, the present invention adopts the following technical solutions:

[0007] A shield tunneling dynamic control method based on multi-source information fusion includes the following steps:

[0008] Obtain shield status parameters, surface settlement values, cutterhead wear values, and tunneling specific energy values;

[0009] Preprocess the acquired data;

[0010] Based on the pre-processed data, the pre-trained prediction model is used to predict the surface settlement, cutterhead wear and driving energy.

[0011] Using shield construction parameters as decision variables, maximizing tunneling specific energy and minimizing surface settlement and tool wear as optimization objectives, a global optimization search is performed within the actual constraints of shield tunneling to determine the optimal solution for shield construction parameters.

[0012] Based on the optimal solution of shield construction parameters, the shield machine's excavation speed, total thrust and cutterhead speed are optimized and controlled.

[0013] As an optional implementation method, the process of obtaining the cutterhead wear value includes: obtaining the acoustic wave signal of the shield machine excavation process, preprocessing it, extracting and processing the acoustic wave signal for time-frequency analysis, extracting frequency characteristics, estimating the degree of cutterhead wear, correcting the estimated degree of cutterhead wear, and obtaining the cutterhead wear value.

[0014] As a further limited embodiment, the process of preprocessing the acoustic wave signal, extracting and performing time-frequency analysis on the processed acoustic wave signal, extracting frequency features, and estimating the degree of wear of the cutter head includes: using an ultrasonic microphone array installed at a key position of the cutter head to collect acoustic wave signals generated by the interaction between the tool and the workpiece during the machining process, using beamforming technology to weightedly sum the signals collected by multiple microphones, and in the weighted process, enhancing the sound signal from the direction of the tool and weakening interference from other directions;

[0015] The beamformed signal is subjected to wavelet transform and Wiener filtering to remove low-frequency and high-frequency noise, retaining the frequency components related to tool wear. Time-frequency analysis is then performed on the filtered signal, and the frequency features related to tool wear are extracted from the time-frequency spectrum. The frequency features are compared with the initial design frequency curve of the cutterhead, and the frequency offset is calculated to estimate the degree of cutterhead wear.

[0016] As a further limited embodiment, the process of correcting the estimated cutter head wear degree includes adding the tool wear empirical formula q for correction:

[0017]

[0018] Among them, n d is the dynamic coefficient, F n is the total thrust, n is the cutterhead speed, r is the cutterhead radius, l is the tunnel ring length, k is the empirical coefficient between friction energy and tool wear; σ c is the unconfined compressive strength, T is the cutterhead torque, and v is the tunneling speed;

[0019] The above data are normalized and the target data samples are scaled to the interval [-1, 1] to the same extent.

[0020] As an optional implementation method, the process of obtaining surface settlement values includes setting a plurality of monitoring points at set distances in the horizontal direction on the ground with the tunnel centerline as the center line, and setting a plurality of monitoring points at set distances in the longitudinal direction of the track surface with each horizontal monitoring point as the origin.

[0021] As an optional implementation, the prediction model is a CNN_LSTM_ATTENTION prediction model, comprising a connected CNN unit and an LSTM unit based on an attention mechanism, wherein the CNN unit based on the attention mechanism is based on the CNN module and adds a parallel attention mechanism module to extract significant features. The CNN module is composed of a stack of multiple layers of one-dimensional networks, each layer comprising a convolution layer, a batch normalization layer, a nonlinear layer and a pooling layer. The attention mechanism module includes feature aggregation and scale recovery parts. The feature aggregation part uses a stack of multiple layers of convolution and pooling layers to extract key features from cross-scale subsequences. The last layer uses a convolution kernel of size 1×1 to mine linear relationships. The scale recovery part restores the key features to the original data size, and then uses the sigmoid function to constrain the values between 0 and 1.

[0022] As an optional implementation manner, the actual setting constraint range of the shield tunneling is that the maximum value and minimum value of the shield construction parameter determined according to the actual construction environment are respectively used as the upper limit and lower limit of the value of the corresponding construction parameter.

[0023] As an optional implementation method, a global optimization process for determining the optimal solution for shield construction parameters includes: using the NSGA-III algorithm for global optimization, calculating the fitness value based on the predicted values of surface settlement, cutterhead wear, and tunneling specific energy, with the optimization objectives of maximizing tunneling specific energy and minimizing surface settlement and cutter wear;

[0024] Perform non-dominated sorting on the individuals in the population, determine the non-dominated frontier, and select some individuals to be retained in the next generation of the population based on the non-dominated sorting and crowding distance;

[0025] Perform crossover operations on individuals that meet the requirements to generate new parameter combinations, perform mutation operations on the individuals generated after crossover to introduce new genetic mutations, and add the newly generated individuals to the next generation population;

[0026] Check whether the preset maximum number of iterations has been reached. If so, stop the algorithm and output the optimal solution at this time; otherwise, return to the step of performing non-dominated sorting on the individuals in the population;

[0027] The Pareto optimal solution obtained from the NSGA-Ⅲ algorithm is used to select the optimal decision plan using the fuzzy decision method.

[0028] As an optional implementation method, the process of performing global optimization and determining the optimal solution for shield construction parameters includes: selecting the optimal decision plan from the Pareto optimal solution set obtained by the NSGA-III algorithm using a fuzzy decision method, defining a membership function for each solution in the Pareto optimal solution set, calculating the membership of each solution to each optimization objective, calculating the comprehensive dominance value of each Pareto optimal solution based on the membership of each solution to each optimization objective, and selecting the Pareto solution corresponding to the maximum dominance value as the optimal compromise solution based on the principle of maximum comprehensive dominance value.

[0029] A shield tunneling dynamic control system based on multi-source information fusion, including:

[0030] A data processing module is configured to obtain shield state parameters, surface settlement values, cutterhead wear values, and tunneling specific energy values, and pre-process the obtained data;

[0031] The data prediction module is configured to predict the surface settlement prediction value, the cutterhead wear prediction value and the tunneling specific energy prediction value based on the preprocessed data using the pretrained prediction model;

[0032] The optimization module is configured to use shield construction parameters as decision variables, with maximizing tunneling specific energy and minimizing surface settlement and tool wear as optimization objectives. Within the actual constraints of shield tunneling, it conducts a global optimization search to determine the optimal solution for shield construction parameters.

[0033] The control module is configured to optimize and control the tunneling speed, total thrust and cutterhead speed of the shield machine based on the optimal solution of the shield construction parameters.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] This invention addresses the difficulties in fusing multi-source information and supporting intelligent decision-making. It focuses on intelligent perception and prediction of geological-structural-equipment data, establishes the actual control range of engineering site control targets, and uses machine learning and deep learning algorithms to dynamically optimize excavation parameters, providing scientific advice for shield drivers to avoid safety risks and save project costs.

[0036] The present invention performs data preprocessing on the tunneling parameters of the shield machine, eliminates the interconnected relationships, takes surface settlement, tunneling specific energy and tool wear as control targets, selects decision variables, explores the nonlinear relationship between construction parameters and control targets, constructs a prediction model based on the attention mechanism, predicts the parameters related to the control targets, and can predict the next tunneling state of the shield machine in real time and dynamically, thereby ensuring the accuracy of the tunneling parameters.

[0037] The present invention constructs a fitness function of the control objective based on the prediction model, uses NSGA-III to perform multi-objective optimization design of construction parameters, and selects the best combination of construction parameters from multiple Pareto optimal solution sets through fuzzy decision-making method, providing a basis for intelligent excavation control and management decision-making of shield tunnels.

[0038] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0040] Figure 1 1 is a schematic diagram of shield status data cleaning results of an embodiment, wherein (a) is the original data, (b) is the data for monitoring abnormal values, (c) is the data after removing abnormal values and completing, and (d) is the data after smoothing;

[0041] Figure 2 This is a schematic diagram of the arrangement of land subsidence monitoring points according to an embodiment;

[0042] Figure 3 This is a schematic diagram of the network structure of the CNN_LSTM_ATTENTION prediction model of an embodiment;

[0043] Figure 4 The present invention is a flowchart of a method for dynamic control of shield tunneling based on multi-source information fusion according to an embodiment. DETAILED DESCRIPTION

[0044] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0045] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0046] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0047] In the absence of conflict, the embodiments and features in the embodiments of this application can be combined with each other.

[0048] Example 1

[0049] A shield tunneling dynamic control method based on multi-source information fusion, such as Figure 4 As shown, the following steps are included:

[0050] Obtain shield status parameters, surface settlement values, cutterhead wear values, and tunneling specific energy values;

[0051] Preprocess the acquired data;

[0052] Based on the pre-processed data, the pre-trained prediction model is used to predict the surface settlement, cutterhead wear and driving energy.

[0053] Using shield construction parameters as decision variables, maximizing tunneling specific energy and minimizing surface settlement and tool wear as optimization objectives, a global optimization search is performed within the actual constraints of shield tunneling to determine the optimal solution for shield construction parameters.

[0054] Based on the optimal solution of shield construction parameters, the shield machine's excavation speed, total thrust and cutterhead speed are optimized and controlled.

[0055] The following is a detailed introduction.

[0056] First, the shield machine status parameters are obtained, and the operating status of the entire excavation process is recorded using the sensors configured on the shield machine.

[0057] First, shield status parameters are obtained. This example utilizes the ultra-large diameter shield machine (TBM) in the Huanggang Road-Crossing Yellow River Tunnel. During on-site construction, the shield equipment generates massive amounts of data. This system automatically collects and stores various construction and excavation parameters at a 1 Hz frequency, primarily including basic data on geology, rock mass, and equipment. The data is presented as a CSV or TEXT file, with the recorded data corresponding to changes in the number of tunneling rings.

[0058] This embodiment uses the total thrust, propulsion speed, and cutterhead speed being non-zero as the normal excavation judgment standard, and the coding is removed as the shield shutdown parameter.

[0059] This example uses a boxplot model from statistics to detect and correct outliers, ensuring the accuracy of the final model. Outliers are replaced with null values, and a moving window algorithm is used to interpolate and fill in the missing values in the data. The moving window algorithm primarily involves the following steps: First, a window size is set. Within each window, data segments containing missing values are analyzed. The window is then slid across the time series, filling in missing values within each window one at a time. In this example, the window size is set to 15, and the median of all data within the window is used for interpolation.

[0060] In addition, to improve data stability, this embodiment performs smoothing after data completion. During the data cleaning process, samples are input in seconds, and the data collected during the shield tunneling process are cleaned ring by ring.

[0061] Taking advancement speed as an example, the data cleaning process for each ring of shield tunneling is as follows: Figure 1 shown.

[0062] By analyzing the correlation between parameters, we screen out those with a significant impact on tunneling efficiency and eliminate redundant information. Using the Pearson correlation coefficient matrix, we identify and retain parameters that are highly correlated with the target variable, simplifying model input and improving prediction accuracy.

[0063] In this embodiment, based on the actual excavation data of the engineering project, the Pearson correlation coefficient analysis is used to remove any parameter with an absolute value of the pairwise correlation coefficient ≥ 0.8, and the data with lower correlation is retained as input.

[0064] This embodiment uses an ultrasonic sensor to monitor the wear degree of the cutter disc in real time, estimates the wear amount by the change in ultrasonic signal intensity, and performs correction in combination with research empirical formulas to ensure the accuracy of the wear value.

[0065] Ultrasonic sensors are placed at key locations on the cutterhead to collect signals in real time. An ultrasonic microphone array captures the acoustic signals generated by the interaction between the tool and the workpiece during machining. Using beamforming technology, the signals collected by multiple microphones are weighted and summed, enhancing the acoustic signal from the direction of the tool while attenuating interference from other directions. The beamformed signal is then subjected to a wavelet-based Wiener filter to remove low- and high-frequency noise while retaining frequency components associated with tool wear. Time-frequency analysis is performed on the filtered signal, extracting frequency features associated with tool wear from the time-frequency spectrum. These frequency features are then compared with the cutterhead's originally designed frequency curve to calculate the frequency offset, enabling a precise estimate of cutterhead wear.

[0066] The sensor placement and its signal coverage directly affect the monitoring accuracy. Therefore, this embodiment adds an empirical formula for tool wear to perform correction based on previous research to ensure the accuracy of the wear value.

[0067] The specific correction method is to take the average value of wear:

[0068]

[0069] Among them, n d is the dynamic coefficient, F n is the total thrust, n is the cutterhead speed, r is the cutterhead radius, l is the tunnel ring length, k is the empirical coefficient between friction energy and tool wear; σ c is the unconfined compressive strength, T is the cutterhead torque, and v is the tunneling speed.

[0070] q is the numerical empirical formula for tool wear, which is mainly used to correct the collected tool wear degree. The correction adopts the linear correction method.

[0071] The linear correction process includes:

[0072] (1) Collect multiple sets of theoretical empirical values y1 and measured values y2;

[0073] (2) Fitting the linear relationship using the least squares method: y2 = ay1 + b + θ;

[0074] (3) Calculate the slope a and intercept b, and the correction formula is: y = (y2-b) / a;

[0075] Where y is the tool wear correction value.

[0076] In order to reduce the impact of dimensional differences between indicators on prediction efficiency and accuracy, the data in the sample set are normalized, and the target data samples are scaled to the interval [-1,1] to the same extent.

[0077] This example uses a total station to monitor the project's surface settlement values. Surface settlement monitoring benchmarks are set, and the monitoring method uses road and surface monitoring, integrating them into a closed-loop space and forming a node network composed of attached lines. Burial methods mainly include manual excavation and drilling.

[0078] First, in this embodiment, for soft soil, manual operation of the Luoyang shovel is used for excavation and burial. For stone or hard soil, drilling rigs and other machines are used for burial. The aperture size of the hole is 70 mm for both methods, and the depth is 1 meter below the ground surface. Subsequently, the bottom of the hole is compacted, the debris in the hole is removed and treated, and clean water is introduced for maintenance. Concrete is poured, and the concrete is kept 4 cm away from the ground surface. Then, steel bars are inserted into the center of the hole and the steel bars extend 20 cm beyond the concrete. A protective cover is set for maintenance for 15 days to keep the burial flat and stable.

[0079] The purpose of tamping the bottom of the tunnel and removing the debris is to strengthen the bearing capacity of the foundation and prevent sinking or tilting due to a soft bottom after concrete pouring. The purpose of introducing clean water for curing is to wash away floating dust and improve the contact density between the concrete and the soil.

[0080] like Figure 2 As shown in the figure, the horizontal monitoring point design: Due to the influence of the ground environment, 7 monitoring points are designed on each section, and each monitoring point is 5 meters apart. The numbering design of the horizontal monitoring points: the same horizontal surface is numbered from left to right 0, 1, 2, 3, 4, 5, 6.

[0081] Design of longitudinal detection points: With each horizontal detection point as the origin, a monitoring point is set at every 8m in the longitudinal direction to form a grid detection system.

[0082] Calculated specific driving energy Average quantitative representation model.

[0083]

[0084] Where, Indicates the average value of the degree of fragmentation; It represents the average value of uniaxial compressive strength of rock; Indicates the average torque of the cutter head; Average propulsion speed.

[0085] This embodiment considers the three aspects of shield tunneling geology, cost and safety, selects surface settlement, cutterhead wear and tunneling specific energy as optimization targets, establishes a multi-objective optimization function, and makes control decisions on shield machine construction parameters.

[0086] The nonlinear mapping relationship between multiple objectives such as surface settlement, cutterhead wear, and tunneling specific energy and decision variables is used to construct a CNN_LSTM_ATTENTION prediction model. Through real-time data feedback, the tunneling parameters are dynamically adjusted to ensure construction accuracy and safety.

[0087] The CNN-LSTM-ATTENTION network model of this embodiment is as follows: Figure 3 As shown in the figure, it consists of three parts: CNN unit based on attention mechanism, LSTM unit and output unit. The details are as follows:

[0088] (1) Data preprocessing: For multivariate heterogeneous data, data cleaning and redundant parameter filtering are used for preprocessing. The training set, test set, and validation set are randomly divided into a ratio of 7:2:1.

[0089] (2) Attention-based CNN unit: Multiple partially overlapping continuous subsequences are intercepted from the original data as the input of this unit. Each CNN input has a subsequence length of s (original scale). Based on the standard CNN module, a parallel attention mechanism branch is added to extract salient features.

[0090] The CNN module is composed of multiple layers of stacked one-dimensional networks. Each layer includes a convolutional layer, a batch normalization layer, and a nonlinear layer. Pooling layers are used to aggregate samples. The stacking of convolutional layers creates a hierarchical structure that extracts progressively more abstract features. The module outputs m feature sequences of length n, which can be expressed as (n × m).

[0091] The attention mechanism module consists of two parts: feature aggregation and scale recovery. The feature aggregation part uses a stack of multiple convolutional and pooling layers to extract key features from cross-scale subsequences. The last layer uses a convolution kernel of size 1×1 to mine linear relationships. The scale recovery part restores the key features to (n×m), that is, to keep the same size as the output features of the CNN module, and then uses the sigmoid function to constrain the values between 0 and 1. Based on the CNN structure of the attention mechanism, the attention module branch input sequence covers the input of the CNN, and stacks deep convolutional networks and pooling layers. The input receptive field corresponding to the features is expanded, and the model obtains more comprehensive contextual information, thereby learning the importance of the current local sequence features.

[0092] (3) LSTM unit: The output of the previous unit is used as the input of this unit to build a time series prediction model.

[0093] Multi-objective optimization of shield operating parameters is carried out from three aspects: surface settlement, tunneling specific energy and cutterhead wear. For high-dimensional multi-objective optimization problems, the CNN_LSTM_ATTENTION prediction regression fitting function is used as the fitness function of the optimization objective. The NSGAⅢ algorithm is used to achieve global optimization of decision-making and determine the optimal solution for the operating parameter combination, thereby improving the tunneling efficiency of tunnel shield construction and reducing the engineering cost and risk during the tunneling process.

[0094] (1) Determine the objective function and constraint range

[0095] Efficiency, cost, and safety during tunneling are key concerns when making construction parameter control decisions. However, it is often difficult to simultaneously optimize both tunneling efficiency and project cost and safety risk reduction. In order to consider multiple decision-making objectives simultaneously, this project considers efficiency, cost, and safety from the perspective of surface settlement, tunneling specific energy, and cutter wear during tunneling shield construction. Using shield construction parameters as decision variables, maximizing propulsion speed, minimizing tunneling specific energy, and minimizing cutter wear are selected as objectives to construct a multi-objective optimization problem. A trained regression function is introduced as the optimized fitness function, so the objective function is:

[0096]

[0097] Among them, f1, f2, and f3 represent the fitness functions of surface settlement, cutterhead wear, and tunneling specific energy, respectively. i Indicates the construction parameters of the shield.

[0098] The constraint range of each operation parameter is set according to the actual situation of the project, and the maximum and minimum values of the measured data of the existing ring of the project are used as the upper and lower limits of the value of each construction parameter respectively.

[0099] (2) Multi-objective optimization

[0100] With the optimization objectives of minimizing surface settlement and tool wear while maximizing specific excavation energy, the NSGA-III algorithm was used for global optimization to determine the optimal solution for the operating parameter combination. Before performing NSGA-III multi-objective optimization, it was necessary to first determine the number of genetic algorithm objectives, population size, crossover and mutation operator values, and optimization stopping criteria. Considering that an appropriate population size and number of iterations can promote convergence in multi-objective optimization, this example selected a crossover operator of 0.7, a mutation operator of 0.01, a population size of 100, and a maximum evolution and stopping generations of 60 for the NSGA-III algorithm. After completing the parameter settings, the NSGA-III algorithm was run in Python to obtain a Pareto-optimal solution set.

[0101] The specific steps of multi-objective optimization algorithm are as follows:

[0102] ① Evaluation and fitness calculation. For each individual in the population, a regression model is trained using this set of parameters. The trained model is used to predict surface settlement, tool wear, and specific excavation energy on the validation set. Fitness is calculated, with the optimization objectives of minimizing surface settlement and tool wear while maximizing specific excavation energy.

[0103] ② Non-dominated sorting and selection. Perform non-dominated sorting on the individuals in the population and determine the non-dominated frontier. Based on the non-dominated sorting and crowding distance, select some individuals to be retained in the next generation of the population.

[0104] ③Genetic operation. Perform a crossover operation on individuals that meet the requirements to generate new parameter combinations. Perform a mutation operation on the individuals generated after crossover to introduce new genetic mutations. The newly generated individuals are added to the next generation population.

[0105] ④ Termination condition determination. Check whether the preset maximum number of iterations has been reached. If the termination condition is met, the algorithm stops; otherwise, return to step ② to continue iteration.

[0106] The non-dominated solution sets found during the NSGA-III algorithm search process represent different trade-offs between surface settlement, cutterhead wear, and tunneling specific energy. Multiple sets of parameters are given as the final optimization results, corresponding to multiple control parameter decision schemes, and the decision range of active control parameters is given for actual engineering.

[0107] The Pareto optimal solution obtained using the NSGA-III algorithm is a set of solutions that meet multi-objective optimization requirements and correspond to multiple decision-making schemes for operating parameters. However, in actual tunnel construction, only one decision-making scheme is needed to guide on-site tunnel excavation. Therefore, a method is required to select the optimal solution from the Pareto solution set. This embodiment uses a fuzzy decision-making method to select the optimal decision-making scheme. The implementation process is as follows:

[0108] Define a membership function for each solution in the Pareto solution set Calculate the membership of each solution to each optimization objective. According to the optimization direction of the optimization objective, the membership calculation is divided into two cases.

[0109] (1) When the optimization objective is minimization:

[0110]

[0111] (2) When the optimization goal is maximization:

[0112]

[0113] in, represents the membership of the j-th solution in the Pareto solution set to the i-th target, f i max represents the maximum value of the target in the solution set, f i min represents the minimum value of the target in the solution set, f i j It represents the value of the i-th objective corresponding to the j-th solution in the Pareto solution set.

[0114] According to the membership of each solution to each optimization objective, the comprehensive dominance value of each Pareto optimal solution is calculated, and the dominance value of the jth solution is The calculation of is as follows:

[0115]

[0116] Where m represents the number of optimization objectives, and n represents the number of solutions in the Pareto solution set obtained by the NSGA-III algorithm.

[0117] According to the principle of maximum comprehensive dominance value, the Pareto solution corresponding to the maximum dominance value is selected as the optimal compromise solution; the larger the calculated comprehensive dominance value, the greater the comprehensive membership of the solution to each optimization objective, that is, the best comprehensive performance of the solution.

[0118] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0119] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0120] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0121] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0122] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made by those skilled in the art that fall within the spirit and principles of the present invention and do not require creative effort are intended to be within the scope of protection of the present invention.

Claims

1. A shield tunneling dynamic control method based on multi-source information fusion, characterized by: The following steps are involved: Obtain shield status parameters, surface settlement values, cutterhead wear values, and tunneling specific energy values; Preprocess the acquired data; Based on the pre-processed data, the pre-trained prediction model is used to predict the surface settlement, cutterhead wear and driving energy. Using shield construction parameters as decision variables, maximizing tunneling specific energy and minimizing surface settlement and tool wear as optimization objectives, a global optimization search is performed within the actual constraints of shield tunneling to determine the optimal solution for shield construction parameters. Based on the optimal solution of shield construction parameters, the shield machine's excavation speed, total thrust and cutterhead speed are optimized and controlled.

2. The shield tunneling dynamic control method based on multi-source information fusion according to claim 1 is characterized in that: The process of obtaining the cutterhead wear value includes: obtaining the acoustic wave signal of the shield machine during the excavation process, preprocessing it, extracting and processing the acoustic wave signal for time-frequency analysis, extracting frequency characteristics, estimating the degree of cutterhead wear, correcting the estimated degree of cutterhead wear, and obtaining the cutterhead wear value.

3. The shield tunneling dynamic control method based on multi-source information fusion according to claim 2 is characterized in that: The process of preprocessing the acoustic wave signal, extracting and performing time-frequency analysis on the processed acoustic wave signal, extracting frequency features, and estimating the degree of cutterhead wear includes: using an ultrasonic microphone array installed at key locations on the cutterhead to collect acoustic wave signals generated by the interaction between the tool and the workpiece during machining; using beamforming technology to weight and sum the signals collected by multiple microphones, thereby enhancing the sound signal from the tool direction and weakening interference from other directions; The beamformed signal is subjected to wavelet transform and Wiener filtering to remove low-frequency and high-frequency noise, retaining the frequency components related to tool wear. Time-frequency analysis is then performed on the filtered signal, and the frequency features related to tool wear are extracted from the time-frequency spectrum. The frequency features are compared with the initial design frequency curve of the cutterhead, and the frequency offset is calculated to estimate the degree of cutterhead wear.

4. The shield tunneling dynamic control method based on multi-source information fusion according to claim 1 is characterized in that: The process of correcting the estimated cutter wear includes adding the empirical formula q for tool wear: Among them, n d is the dynamic coefficient, F n is the total thrust, n is the cutterhead speed, r is the cutterhead radius, l is the tunnel ring length, k is the empirical coefficient between friction energy and tool wear; σ c is the unconfined compressive strength, T is the cutterhead torque, and v is the tunneling speed; The above data are normalized and the target data samples are scaled to the interval [-1, 1] to the same extent.

5. The shield tunneling dynamic control method based on multi-source information fusion according to claim 1 is characterized in that: The process of obtaining the surface settlement value includes setting a plurality of monitoring points at set intervals horizontally on the ground with the tunnel centerline as the center line, and setting a plurality of monitoring points at set intervals longitudinally on the track surface with each horizontal monitoring point as the origin.

6. The shield tunneling dynamic control method based on multi-source information fusion according to claim 1 is characterized in that: The prediction model is a CNN_LSTM_ATTENTION prediction model, which includes a connected CNN unit and an LSTM unit based on an attention mechanism. The CNN unit based on the attention mechanism is based on the CNN module and adds a parallel attention mechanism module to extract significant features. The CNN module is composed of a stack of multiple layers of one-dimensional networks, each layer of which includes a convolution layer, a batch normalization layer, a nonlinear layer and a pooling layer. The attention mechanism module includes feature aggregation and scale recovery parts. The feature aggregation part uses a stack of multiple layers of convolution and pooling layers to extract key features from cross-scale subsequences. The last layer uses a convolution kernel of size 1×1 to mine linear relationships. The scale recovery part restores the key features to the original data size, and then uses the sigmoid function to constrain the values between 0 and 1.

7. The shield tunneling dynamic control method based on multi-source information fusion according to claim 1 is characterized in that: The actual setting constraint range of the shield tunneling is that the maximum value and minimum value of the shield construction parameters determined according to the actual construction environment are respectively used as the upper limit and lower limit of the corresponding construction parameters.

8. The shield tunneling dynamic control method based on multi-source information fusion according to claim 1 is characterized in that: The process of performing global optimization to determine the optimal solution for shield construction parameters includes: using the NSGA-III algorithm for global optimization, calculating the fitness value based on the predicted values of surface settlement, cutterhead wear, and tunneling specific energy, with the optimization goals of maximizing tunneling specific energy and minimizing surface settlement and cutter wear; Perform non-dominated sorting on the individuals in the population, determine the non-dominated frontier, and select some individuals to be retained in the next generation of the population based on the non-dominated sorting and crowding distance; Perform crossover operations on individuals that meet the requirements to generate new parameter combinations, perform mutation operations on the individuals generated after crossover to introduce new genetic mutations, and add the newly generated individuals to the next generation population; Check whether the preset maximum number of iterations has been reached. If so, stop the algorithm and output the optimal solution at this time; otherwise, return to the step of performing non-dominated sorting on the individuals in the population.

9. The shield tunneling dynamic control method based on multi-source information fusion according to claim 1, characterized in that: The process of global optimization and determining the optimal solution for shield construction parameters includes: selecting the optimal decision plan from the Pareto optimal solution set obtained by the NSGA-III algorithm using the fuzzy decision method, defining a membership function for each solution in the Pareto optimal solution set, calculating the membership of each solution to each optimization objective, calculating the comprehensive dominance value of each Pareto optimal solution based on the membership of each solution to each optimization objective, and selecting the Pareto solution corresponding to the maximum dominance value as the optimal compromise solution based on the principle of maximum comprehensive dominance value.

10. A shield tunneling dynamic control system based on multi-source information fusion, characterized by: include: A data processing module is configured to obtain shield state parameters, surface settlement values, cutterhead wear values, and tunneling specific energy values, and pre-process the obtained data; The data prediction module is configured to predict the surface settlement prediction value, the cutterhead wear prediction value and the tunneling specific energy prediction value based on the preprocessed data using the pretrained prediction model; The optimization module is configured to use shield construction parameters as decision variables, with maximizing tunneling specific energy and minimizing surface settlement and tool wear as optimization objectives. Within the actual constraints of shield tunneling, it conducts a global optimization search to determine the optimal solution for shield construction parameters. The control module is configured to optimize and control the tunneling speed, total thrust and cutterhead speed of the shield machine based on the optimal solution of the shield construction parameters.

Images