Intelligent direction adjusting method for shield tunneling of volcanic ash stratum

Abstract

The invention relates to the technical field of shield tunneling, and discloses an intelligent direction adjusting method for shield tunneling of a volcanic ash stratum. The method comprises the following steps: collecting construction data of a shield tunneling machine and a target area through a multi-source sensor, preprocessing the construction data, and fusing the preprocessed construction data to obtain real-time monitoring data; acquiring historical tunneling error data of the target area, analyzing the historical tunneling error data and the real-time monitoring data by using a deep learning model, and generating a tunneling direction adjustment instruction; according to the tunneling direction adjusting instruction, the propelling speed and the steering angle of the shield tunneling machine and thrust distribution of a jacking oil cylinder are adjusted; and generating a mud cake control instruction according to the real-time monitoring data, adjusting the cleaning parameters of the mud cake cleaning device according to the mud cake control instruction, and displaying the adjusted operation state and tunneling direction of the shield tunneling machine through a human-computer interaction interface. The problem of shield tunneling direction control is solved, and real-time regulation and control of the propelling speed and the steering angle are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of shield tunneling, and in particular to an intelligent direction adjustment method for shield tunneling in volcanic ash strata. Background Art

[0002] With the rapid development of urban infrastructure, the demand for large-scale underground projects such as subways and tunnels continues to increase. The shield tunneling method has become one of the main tunneling methods due to its advantages such as construction safety, high efficiency, and minimal ground disturbance.

[0003] However, in complex geological conditions such as volcanic ash formations, traditional shield tunneling methods face numerous challenges due to the uneven structural signals, fine, loose particles, and high moisture content of the formation. In particular, directional control is difficult and mud cake formation is prone to occur, which affects tunneling efficiency and construction safety. Current intelligent directional control technologies are primarily designed for general formations and lack specific optimization for the unique properties of volcanic ash formations. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to overcome the deficiencies in the prior art and provide an intelligent direction adjustment method for shield tunneling in volcanic ash strata.

[0005] The present invention provides the following technical solutions: In a first aspect, the present invention provides an intelligent direction adjustment method for shield tunneling in volcanic ash formations, the method comprising: Collecting construction data of the shield machine and the target area through multi-source sensors, preprocessing the construction data, and fusing the preprocessed construction data to obtain real-time monitoring data; Acquire historical excavation error data of the target area, analyze the historical excavation error data and the real-time monitoring data using a deep learning model, and generate an excavation direction adjustment instruction; adjusting the propulsion speed and steering angle of the shield machine according to the tunneling direction adjustment instruction, and adjusting the thrust distribution of multiple jacking cylinders in the shield machine according to the tunneling direction adjustment instruction; A mud cake prevention and control instruction is generated according to the real-time monitoring data, and the removal parameters of the mud cake removal device are adjusted according to the mud cake prevention and control instruction. The adjusted operating status and excavation direction of the shield machine are displayed in real time through the human-computer interaction interface.

[0006] In an optional embodiment, the multi-source sensor includes a posture sensor, a geological radar, a camera, and a mud parameter sensor. The construction data includes tilt angle, acceleration, stratum structure signal, tunneling face image, pressure data, flow data, viscosity data, and stratum moisture data. The construction data of the shield machine and the target area collected by the multi-source sensor includes: Optimizing the installation position of the multi-source sensor on the shield machine using a genetic algorithm; The inclination angle and acceleration of the shield machine are collected by the attitude sensor, and the stratum structure signal of the target area is collected by the geological radar; The camera collects the excavation face image of the target area, and the mud parameter sensor collects the mud parameters of the target area, including at least one of pressure data, flow data, viscosity data and formation moisture data.

[0007] In an optional embodiment, the preprocessing of the construction data and fusing the preprocessed construction data to obtain real-time monitoring data includes: Using a Kalman filter algorithm to smooth the inclination angle and acceleration of the shield machine to obtain a preprocessed inclination angle and a preprocessed acceleration; Performing multi-resolution analysis on the stratum structure signal of the target area by using a wavelet transform algorithm to obtain a pre-processed stratum structure signal; Performing image analysis on the tunneling face image of the target area using a convolutional neural network to obtain a preprocessed tunneling face image; Using a support vector machine to classify and predict the mud parameters of the target area to obtain preprocessed mud parameters; The extended Kalman filter algorithm is used to fuse the preprocessed inclination angle, the preprocessed acceleration, the preprocessed stratum structure signal, the preprocessed tunneling face image and the preprocessed mud parameters to generate the real-time monitoring data.

[0008] In an optional embodiment, the pre-processed stratum structure signal includes rock, pores, moisture and stratum heterogeneity, and the multi-resolution analysis of the stratum structure signal of the target area using a wavelet transform algorithm to obtain the pre-processed stratum structure signal includes: determining a target wavelet function according to the frequency characteristics of the stratum structure signal and the target resolution; Performing wavelet decomposition on the stratum structure signal using the target wavelet function to obtain wavelet components, including low-frequency components and high-frequency components; Noise signals in the high-frequency component that are higher than a preset frequency threshold are filtered out, and the rocks, pores, and moisture in the target area are identified by analyzing the mutation points and intensities of the denoised high-frequency component. The stratigraphic heterogeneity in the target area is identified by analyzing the fluctuations of the low-frequency component.

[0009] In an optional embodiment, the analyzing the historical excavation error data and the real-time monitoring data using a deep learning model to generate an excavation direction adjustment instruction further includes: Performing regression analysis on the historical excavation error data using a support vector regression model to determine the excavation direction deviation trend and excavation direction deviation factors; Using a Bayesian network to comprehensively analyze the geological parameters of the stratum structure signal to determine the influence weight of each geological stratum segment in the target area on the tunneling direction; Determining the thrust distribution of each jacking cylinder in the shield machine by using a fuzzy adaptive control algorithm and based on the historical excavation error data and the stratum structure signal; The deep learning model is used to process the real-time monitoring data, the excavation direction deviation trend, the excavation direction deviation factor, the influence weight of each geological segment in the target area on the excavation direction, and the thrust distribution of each jacking cylinder in the shield machine to generate the excavation direction adjustment instruction.

[0010] In an optional embodiment, the use of a support vector regression model to perform regression analysis on the historical excavation error data to determine the excavation direction deviation trend and excavation direction deviation factors includes: Performing regression analysis on the historical tunneling error data using the support vector regression model to generate an error prediction curve; Identifying the linear trend, periodic fluctuation and mutation point of the tunneling displacement error based on the error prediction curve to obtain the tunneling direction deviation trend; Acquire additional data collected by additional sensors, perform correlation analysis on the additional data and the historical excavation error data, determine basic factors that are correlated with the excavation displacement error, and use the support vector regression model to determine target factors among the basic factors whose correlation with the excavation displacement error is greater than a preset correlation threshold, thereby obtaining the excavation direction deviation factor.

[0011] In an optional embodiment, the comprehensive analysis of the geological parameters of the stratum structure signal using a Bayesian network to determine the influence weight of each geological stratum segment in the target area on the tunneling direction includes: Defining nodes of the Bayesian network according to the geological parameters, and defining conditional dependencies and a conditional probability table between the nodes according to the historical tunneling error data, wherein the conditional probability table is used to describe the probability distribution of a node under given conditions of its parent node; Inputting each of the geological parameters into the Bayesian network to calculate the probability of the excavation direction deviation of each of the geological parameters under given conditions; According to the probability of the excavation direction deviation of each geological parameter under given conditions and the degree of influence of each excavation direction deviation on the excavation direction, the influence weight of each geological layer section in the target area on the excavation direction is calculated.

[0012] In an optional embodiment, the determining the thrust distribution of each jacking cylinder in the shield machine by using a fuzzy adaptive control algorithm and based on the historical tunneling error data and the stratum structure signal includes: Converting the geological parameters of the stratum structure signal, the historical excavation error data, and the real-time excavation error data into fuzzy sets, and defining the range of each fuzzy set using a preset membership function; Determining fuzzy rules based on the historical tunneling error data, applying the fuzzy rules to each of the fuzzy sets using the fuzzy adaptive control algorithm, and calculating the fuzzy thrust distribution of each of the jacking cylinders; The fuzzy thrust distribution of each jacking cylinder is defuzzified by using a defuzzification algorithm to obtain the thrust distribution of each jacking cylinder.

[0013] In an optional embodiment, the deep learning model includes a convolutional neural network module, a long short-term memory network module, and a comprehensive output module. The deep learning model is used to process the real-time monitoring data, the tunneling direction deviation trend, the tunneling direction deviation factor, the influence weight of each geological layer segment in the target area on the tunneling direction, and the thrust distribution of each jacking cylinder in the shield machine to generate the tunneling direction adjustment instruction, including: Extracting spatial feature data of the real-time monitoring data through the convolutional neural network module; Analyzing the time correlation between the historical excavation error data and the real-time monitoring data through the long short-term memory network module to obtain time series data; Combining the spatial feature data with the time series data through the comprehensive output module to generate a prediction result, wherein the prediction result is used to represent a prediction of formation change trends and migration risks; Based on the prediction results, the target excavation direction adjustment amount is calculated, and the excavation direction adjustment instruction is generated by combining the target excavation direction adjustment amount, the excavation direction deviation trend, the excavation direction deviation factor, the influence weight of each geological segment in the target area on the excavation direction, and the thrust distribution of each jacking cylinder in the shield machine using a PID control algorithm.

[0014] In an optional embodiment, generating a mud cake prevention instruction based on the real-time monitoring data, and adjusting a removal parameter of a mud cake removal device according to the mud cake prevention instruction, includes: Identifying the mud cake location, mud cake thickness, mud cake distribution range, and mud cake physical properties based on the tunneling face image; generating the mud cake prevention and control instruction according to the mud cake position, the mud cake thickness, the mud cake distribution range, the mud cake physical properties and the mud parameters; The cleaning force and frequency of the mud cake removal device are adjusted according to the mud cake prevention and control instructions.

[0015] Beneficial effects of this application: The intelligent direction adjustment method for shield tunneling in volcanic ash formations provided in the embodiment of the present application comprehensively utilizes multi-source sensor data, historical tunneling error analysis, uneven geological distribution in volcanic ash formations, thrust distribution of different jacking cylinders and advanced intelligent algorithms to solve the problem of shield tunneling direction control, realize real-time regulation of propulsion speed and steering angle, and ensure the precise tunneling and construction safety of the shield machine under complex geological conditions.

[0016] 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

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be considered as limiting the scope. A person of ordinary skill in the art can also derive other relevant drawings based on these drawings without inventive effort. Similar components are numbered similarly in the various drawings.

[0018] Figure 1 A flow chart of an intelligent direction adjustment method for shield tunneling in volcanic ash strata provided by an embodiment of the present application is shown; Figure 2 A flowchart of another intelligent direction adjustment method for shield tunneling in volcanic ash formations provided by an embodiment of the present application is shown; Figure 3 A flow chart of another intelligent direction adjustment method for shield tunneling in volcanic ash formations provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0019] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0020] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature specified as "first" or "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs. The terms used in the template description herein are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0022] Example 1 like Figure 1 FIG. 1 is a flow chart of a method for intelligent direction adjustment of shield tunneling in volcanic ash formations according to an embodiment of the present application. The method for intelligent direction adjustment of shield tunneling in volcanic ash formations provided by the embodiment of the present application includes the following steps: Step S110 , collecting construction data of the shield machine and the target area through multi-source sensors, preprocessing the construction data, and fusing the preprocessed construction data to obtain real-time monitoring data.

[0023] In this embodiment, multi-source sensors are installed at the front end and key locations of the shield machine to monitor construction data from the shield machine and the target area in real time. These sensors include, but are not limited to, attitude sensors, geological radar, cameras, and mud parameter sensors. Construction data includes, but is not limited to, tilt angle, acceleration, stratum structure signals, tunneling face images, pressure data, flow data, viscosity data, and stratum moisture data.

[0024] Preferably, a genetic algorithm (GA) can be used to optimize the arrangement of multi-source sensors at the front end and key positions of the shield machine to ensure the comprehensiveness and real-time nature of sensor data and improve data acquisition efficiency.

[0025] Specifically, attitude sensors (gyroscopes, accelerometers) are used to collect the inclination angle and acceleration of the shield machine to ensure stability during the excavation process; geological radar is used to collect the stratum structure signals of the target area, including rock, pores, moisture and stratum heterogeneity, to provide accurate geological basis for intelligent direction adjustment; cameras are used to collect excavation surface images of the target area in real time to assist in identifying stratum changes, potential obstacles and mud cake formation, thereby improving the accuracy and efficiency of image recognition; mud parameter sensors are used to collect mud parameters in the target area, including at least one of pressure data, flow data, viscosity data and stratum humidity data, so as to promptly detect signs of mud cake formation and ensure the normal operation of the slurry shield excavation system.

[0026] Optionally, in one embodiment, each sensor may be calibrated regularly to ensure data accuracy and reliability. The calibration process may include static calibration and dynamic calibration to eliminate errors and drifts of the sensor itself.

[0027] This process uses genetic algorithms to optimize sensor placement, ensuring comprehensive and real-time data collection, reducing blind spots, and improving data collection efficiency. Construction data encompasses multiple dimensions, including shield machine posture, geological structure, tunneling face images, and mud parameters, providing a high-quality foundation for subsequent analysis.

[0028] Furthermore, the Kalman filter algorithm is used to smooth the inclination angle and acceleration of the shield machine to obtain the pre-processed inclination angle and pre-processed acceleration, eliminate noise interference, and improve the stability monitoring accuracy of the shield machine; the wavelet transform algorithm is used to perform multi-resolution analysis on the stratum structure signal of the target area to obtain the pre-processed stratum structure signal; the convolutional neural network (CNN) is used to perform image analysis on the excavation face image of the target area to obtain the pre-processed excavation face image, thereby improving the accuracy and efficiency of image recognition; the support vector machine (SVM) is used to classify and predict the mud parameters of the target area to obtain the pre-processed mud parameters to ensure the normal operation of the mud system.

[0029] In an optional embodiment, as Figure 2 As shown, a multi-resolution analysis is performed on the stratum structure signal of the target area using a wavelet transform algorithm to obtain a pre-processed stratum structure signal, including steps S111 to S113: Step S111, determining a target wavelet function according to the frequency characteristics of the stratum structure signal and the target resolution; Step S112, performing wavelet decomposition on the stratum structure signal using the target wavelet function to obtain wavelet components, including low-frequency components and high-frequency components; Step S113: Filter out the noise signals in the high-frequency component that are higher than a preset frequency threshold, identify the rocks, pores, and moisture in the target area by analyzing the mutation points and intensities of the denoised high-frequency component, and identify the stratigraphic heterogeneity in the target area by analyzing the fluctuations of the low-frequency component.

[0030] Understandably, the target wavelet function (such as Daubechies wavelet, Haar wavelet, etc.) is first determined according to the frequency characteristics and target resolution of the stratum structure signal, and then the target wavelet function is used to perform wavelet decomposition on the stratum structure signal. Through continuous or discrete wavelet transform, the stratum structure signal is decomposed into wavelet components of different scales, including low-frequency components and high-frequency components. The low-frequency component usually contains the overall trend information of the signal and can be used to identify large-scale stratum heterogeneity, while the high-frequency component reflects local changes and is suitable for identifying the presence of rocks, pores or water.

[0031] A threshold filtering method is then used to remove noise signals above a preset frequency threshold from the high-frequency components, retaining only the significant geological signatures. Further, by analyzing the mutation points and intensity of the denoised high-frequency components, pattern recognition algorithms (such as support vector machines and neural networks) are used to accurately identify rock, pore, and moisture characteristics in the target area. By analyzing the fluctuations of the low-frequency components, pattern recognition algorithms are used to accurately identify the heterogeneity of the stratigraphic features in the target area.

[0032] It should be noted that the preset frequency threshold can be adjusted according to experiments and actual formation conditions, and the parameters of each multi-source sensor and the parameters of each algorithm model can also be calibrated regularly to adapt to different geological conditions. The embodiments of this application do not limit this.

[0033] The wavelet transform (steps S111-S113) in the above process extracts low-frequency and high-frequency characteristics of the stratigraphic signal through multi-resolution analysis, accurately identifying large-scale stratigraphic heterogeneity (low-frequency) and local geological details (high-frequency). It also filters high-frequency noise and detects sudden changes, enhancing the ability to identify geological features such as rock, pores, and moisture, providing a high-precision geological basis for direction adjustment.

[0034] Finally, the Extended Kalman Filter (EKF) algorithm is used to fuse the preprocessed inclination angle, preprocessed acceleration, preprocessed stratum structure signal, preprocessed tunneling face image and preprocessed mud parameters. The stratum structure, image, parameters, and current status of the shield machine are integrated to generate unified real-time monitoring data, enhancing the system's comprehensive perception of complex geological environments.

[0035] In step S110, multi-source sensors are used to comprehensively collect construction data of the shield machine and the target area, ensuring the comprehensiveness and real-time nature of the data. Genetic algorithms are used to optimize sensor layout to improve data collection efficiency. At the same time, algorithms such as Kalman filtering, wavelet transform, convolutional neural network, and support vector machine are used to pre-process the data to improve data accuracy and reliability. The extended Kalman filtering algorithm is used to fuse the pre-processed data to generate unified real-time monitoring data, providing a basis for subsequent analysis.

[0036] Step S120: Obtain historical excavation error data of the target area, analyze the historical excavation error data and the real-time monitoring data using a deep learning model, and generate an excavation direction adjustment instruction.

[0037] In an optional embodiment, as Figure 3 As shown, step S120 includes steps S121 to S124: Step S121 , performing regression analysis on the historical excavation error data using a support vector regression model to determine the excavation direction deviation trend and excavation direction deviation factors.

[0038] As can be understood, the Support Vector Regression (SVR) model is first used to perform regression analysis on historical tunneling error data to generate an error prediction curve. By observing the changing trends of the error prediction curve, the linear trend, periodic fluctuations, and mutation points of the tunneling displacement error are identified, and the tunneling direction deviation trend is obtained. If the tunneling displacement error shows a linear increase over time, it may indicate a persistent source of deviation in the system, such as equipment wear or uneven thrust distribution. If the tunneling displacement error shows periodic changes, it may be related to periodic formation characteristics (such as changes in formation density) or the equipment operation cycle. If there is a mutation point in the tunneling displacement error on the error prediction curve, it may be due to a sudden change in geological conditions (such as entering hard rock formations) or equipment failure.

[0039] Next, additional data collected by additional sensors (such as thrust, vibration, stratum density, etc.) are obtained, and correlation analysis is performed on the additional data combined with historical tunneling error data to identify basic factors such as geology or equipment related to changes in tunneling displacement errors. Then, through feature weight analysis of the support vector regression model, target factors among the basic factors whose correlation with the tunneling displacement error is greater than the preset correlation threshold are determined, that is, the main factors that have a greater impact on the tunneling displacement error. For example, changes in stratum properties may be the main factor leading to an increase in tunneling displacement errors.

[0040] Optionally, this embodiment can use charts (such as error trend charts, deviation heat maps) to intuitively display the changes in excavation displacement errors and possible factors, and evaluate the prediction accuracy of the support vector regression model and the actual impact of directional deviation by comparing the differences between the actual excavation path and the planned path.

[0041] Step S121 above effectively identifies linear growth, periodic fluctuations, and sudden changes in tunneling direction deviation, revealing potential issues such as equipment wear and stratum changes. Feature weight analysis pinpoints the primary sources of error (such as changes in stratum properties), providing a targeted basis for directional adjustment strategies.

[0042] Step S122: using a Bayesian network to comprehensively analyze the geological parameters of the stratum structure signal to determine the influence weight of each geological stratum segment in the target area on the tunneling direction.

[0043] Understandably, in this embodiment, the geological parameters of the stratum structure signal include density, porosity, water content, compressive strength, and stratum thickness. First, a Bayesian network is constructed, with each geological parameter representing a node. Based on historical excavation error data and geological knowledge, conditional dependencies between nodes are defined. For example, water content affects porosity, which in turn affects the compressive strength of the stratum. Based on historical excavation error data and expert knowledge, a conditional probability table is defined to describe the probability distribution of a node under the given conditions of its parent node.

[0044] Next, the real-time monitored geological parameters are input into the Bayesian network, and a Bayesian inference algorithm (such as Gibbs sampling and variational inference) is used to calculate the probability of tunneling direction deviation for each parameter under given conditions. Furthermore, the probability of tunneling direction deviation for each geological parameter under given conditions is combined with the degree of influence of each tunneling direction deviation on the tunneling direction. The weighted average or other statistical methods are used to calculate the influence weight of each geological layer segment on the tunneling direction. The formula is as follows: ; Where, is the influence weight of the i-th geological layer on the tunneling direction, For geological parameters Deviation of the excavation direction under given conditions The probability of Deviation in the excavation direction The preset degree of influence on the driving direction adjustment command.

[0045] Optionally, in this embodiment, the Bayesian network structure and conditional probability table can be regularly updated to adapt to new geological information and excavation experience, thereby gradually improving the prediction accuracy and applicability of the model through a feedback mechanism.

[0046] Step S122 quantifies the influence of each geological segment (density, porosity, etc.) on tunneling direction based on the conditional probability table and real-time geological parameters. Dynamically updating the network with expert knowledge improves the model's adaptability to complex geological conditions and enhances the scientific nature of directional adjustments.

[0047] Step S123 , using a fuzzy adaptive control algorithm and based on the historical tunneling error data and the stratum structure signal, determining the thrust distribution of each jacking cylinder in the shield machine.

[0048] Understandably, geological parameters, historical tunneling error data, and real-time tunneling error data are converted into fuzzy sets. For example, the geological parameter ground density can be divided into three fuzzy sets: "low," "medium," and "high." Pre-set membership functions (such as triangular or trapezoidal functions) are then used to define the range of each fuzzy set. Fuzzy rules are then formulated based on expert experience and historical tunneling error data. For example, if the ground density is high and the tunneling error is large, the thrust is increased; if the ground density is low and the tunneling error is small, the thrust is reduced.

[0049] Then, a fuzzy adaptive control algorithm (such as the Mamdani inference algorithm) is used to apply fuzzy rules to each fuzzy set, and the fuzzy thrust distribution of each jacking cylinder is calculated. The fuzzy thrust distribution of each jacking cylinder is defuzzified using a defuzzification algorithm, and the fuzzy thrust distribution is converted into a specific numerical output to obtain the thrust distribution of each jacking cylinder.

[0050] Optionally, in this embodiment, thrust feedback during tunneling can be monitored in real time to assess the effectiveness of the current thrust distribution. This feedback mechanism adjusts the preset membership functions and fuzzy rules, improving the adaptability of the control system. Based on this feedback data, the parameters of the fuzzy adaptive control algorithm can be dynamically adjusted to suit varying geological conditions and tunneling requirements. For example, in high-density formations, the threshold of the preset membership function can be appropriately increased to provide greater thrust.

[0051] Step S123 converts geological parameters and historical errors into fuzzy rules to dynamically adjust the thrust of each jacking cylinder. The feedback mechanism optimizes membership functions and rules to adapt to different geological conditions and ensure balanced thrust distribution and fast response.

[0052] Step S124, using the deep learning model to process the real-time monitoring data, the excavation direction deviation trend, the excavation direction deviation factor, the influence weight of each geological segment in the target area on the excavation direction, and the thrust distribution of each jacking cylinder in the shield machine, to generate the excavation direction adjustment instruction.

[0053] The deep learning model consists of a convolutional neural network (CNN) module, a long-short-term memory (LSTM) network module, and an integrated output module. First, the convolutional, pooling, and fully connected layers within the CNN module extract spatial feature data (such as ground texture and density variations) from real-time monitoring data. The LSTM module then analyzes the temporal correlation between historical tunneling error data and real-time monitoring data. The LSTM module captures long-term dependencies and dynamic trends in data, making it suitable for predicting ground change trends and tunneling deviation risks, effectively capturing time series data. The integrated output module then combines spatial feature data with time series data, using fully connected layers and activation functions to output predictions of future ground changes and deviation risks.

[0054] Furthermore, the target tunneling direction adjustment is calculated based on the prediction results. For example, if the prediction results indicate a high risk of excursion in a certain stratum, the shield machine's propulsion direction needs to be adjusted. A PID control algorithm is then used to convert the target tunneling direction adjustment into a specific tunneling direction adjustment instruction to control the shield machine's propulsion angle. Based on the tunneling direction deviation trend and tunneling direction deviation factors, targeted tunneling direction adjustment instructions are formulated, such as modifying the thrust distribution strategy or adjusting the tunneling speed. Based on the influence weight of each geological stratum segment on the tunneling direction within the target area, tunneling direction adjustment instructions are adjusted in real time, such as changing the shield machine's thrust force or correcting its direction. Based on the thrust distribution of each jacking cylinder in the shield machine, tunneling direction adjustment instructions are adjusted in real time to ensure the balance and effectiveness of the overall thrust.

[0055] Optionally, in this embodiment, the actual performance of the tunneling process can be monitored in real time, and the feedback data can be used to continuously optimize various models and algorithms and adjust instructions. By continuously updating the model algorithms and adjustment instructions, the accuracy and response speed of the direction adjustment instructions can be improved.

[0056] The above step S124 predicts stratum changes and migration risks by fusing spatial features (stratum texture) with time series (historical error trends), and converts the prediction results into precise tunneling direction adjustment instructions (such as propulsion angle and thrust distribution) to achieve closed-loop control.

[0057] Step S130 , adjusting the propulsion speed and steering angle of the shield machine according to the tunneling direction adjustment instruction, and adjusting the thrust distribution of multiple jacking cylinders in the shield machine according to the tunneling direction adjustment instruction.

[0058] Understandably, the PID control algorithm adjusts the instructions according to the excavation direction, accurately adjusts the propulsion speed and steering angle of the shield machine, realizes intelligent direction adjustment, ensures that the excavation direction remains within the predetermined trajectory, and dynamically adjusts the thrust distribution of each jacking cylinder to optimize the overall thrust configuration and achieve stable propulsion of the shield machine.

[0059] Step S130 uses a PID control algorithm to precisely adjust propulsion parameters, ensuring the shield machine advances along the planned trajectory and minimizing human intervention errors. Fuzzy adaptive control results are used to optimize the thrust of each cylinder, balancing the forces acting on the equipment, avoiding local overloads or deviations, and improving tunneling stability.

[0060] Step S140: Generate mud cake prevention and control instructions based on the real-time monitoring data, adjust the removal parameters of the mud cake removal device according to the mud cake prevention and control instructions, and display the adjusted operating status and excavation direction of the shield machine in real time through the human-computer interaction interface.

[0061] Understandably, during shield tunneling, mud cake refers to slurry blockages formed in the cutterhead, conveying pipelines, and other locations due to improper soil conditions and mud-water ratios, which can affect the normal operation of the equipment. Mud parameters collected by the mud parameter sensor and tunneling face images captured by the camera can be used to identify the location, thickness, distribution, and physical properties (such as viscosity and hardness) of the mud cake within the shield machine's cutterhead and conveying pipelines, and then generate mud cake prevention instructions. Based on these instructions, the cleaning force and frequency of the mud cake removal device are adjusted, and the operation of mud cake removal devices (such as agitators and mud pumps) is controlled to promptly remove mud cake, ensure smooth mud flow, and prevent mud cake blockages from affecting shield machine operation.

[0062] Optionally, in this embodiment, the adjusted shield machine status and mud cake removal effect can be monitored in real time, and a deep reinforcement learning (Deep Q-Network, DQN) model can be used for feedback learning to continuously optimize the direction adjustment and removal strategies, thereby improving the system's adaptability and decision-making accuracy.

[0063] The above process combines mud parameters and image data to identify the location, thickness and physical properties of the mud cake, and dynamically adjusts the force and frequency of the cleaning device to effectively prevent blockage.

[0064] The remote control terminal displays the shield machine's operating status, location, ground moisture, and mud cake conditions in real time through a human-machine interface, enhancing the operator's monitoring experience. A heat map algorithm is also used to display tunneling direction adjustment instructions, execution results, and the activation of mud cake prevention measures in real time, helping operators quickly understand the shield machine's operating status. An isolation forest algorithm is used to analyze monitoring data in real time to detect system anomalies (such as severe mud cake accumulation and sudden ground changes). A multi-level alarm mechanism (audio, visual, SMS / email notifications) issues timely alerts to ensure construction safety and enable timely response measures. When necessary, manual intervention is performed through the AI-assisted decision-making system to adjust tunneling parameters or activate backup mud cake prevention measures.

[0065] This process uses a human-machine interface to display real-time tunneling status and anomalies (such as sudden ground changes), assisting operators in making quick decisions. High-risk events (such as severe mud cake accumulation) are promptly warned through audio, visual, and remote notifications, ensuring construction safety. Intuitive displays of equipment status, ground moisture, and the results of directional command execution enhance monitoring efficiency and the operator experience.

[0066] The intelligent direction adjustment method for shield tunneling in volcanic ash formations provided in the embodiment of the present application comprehensively utilizes multi-source sensor data, historical tunneling error analysis, uneven geological distribution in volcanic ash formations, thrust distribution of different jacking cylinders and advanced intelligent algorithms to solve the problem of shield tunneling direction control, realize real-time regulation of propulsion speed and steering angle, and ensure the precise tunneling and construction safety of the shield machine under complex geological conditions.

[0067] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and structure diagrams in the accompanying drawings show the possible architectures, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in an alternative implementation, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the structure diagram and / or flowchart, and the combination of boxes in the structure diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions.

[0068] In addition, the functional modules or units in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0069] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part 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 several instructions for enabling a computer device (which can be a smart phone, 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 invention. The aforementioned storage medium can be a non-volatile storage medium or a volatile storage medium. For example, the storage medium can be: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., various media that can store program code.

[0070] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the scope of protection of the present invention.

Claims

1. An intelligent direction adjustment method for shield tunneling in volcanic ash strata, characterized in that: The method comprises: Collecting construction data of the shield machine and the target area through multi-source sensors, preprocessing the construction data, and fusing the preprocessed construction data to obtain real-time monitoring data; Acquire historical excavation error data of the target area, analyze the historical excavation error data and the real-time monitoring data using a deep learning model, and generate an excavation direction adjustment instruction; adjusting the propulsion speed and steering angle of the shield machine according to the tunneling direction adjustment instruction, and adjusting the thrust distribution of multiple jacking cylinders in the shield machine according to the tunneling direction adjustment instruction; A mud cake prevention and control instruction is generated according to the real-time monitoring data, and the removal parameters of the mud cake removal device are adjusted according to the mud cake prevention and control instruction. The adjusted operating status and excavation direction of the shield machine are displayed in real time through the human-computer interaction interface.

2. The intelligent direction adjustment method for shield tunneling in volcanic ash strata according to claim 1 is characterized in that: The multi-source sensors include attitude sensors, geological radars, cameras, and mud parameter sensors. The construction data include tilt angles, accelerations, stratum structure signals, tunneling face images, pressure data, flow data, viscosity data, and stratum moisture data. The construction data of the shield machine and the target area collected by the multi-source sensors include: Optimizing the installation position of the multi-source sensor on the shield machine using a genetic algorithm; The inclination angle and acceleration of the shield machine are collected by the attitude sensor, and the stratum structure signal of the target area is collected by the geological radar; The camera collects the excavation face image of the target area, and the mud parameter sensor collects the mud parameters of the target area, including at least one of pressure data, flow data, viscosity data and formation moisture data.

3. The intelligent direction adjustment method for shield tunneling in volcanic ash strata according to claim 2 is characterized in that: The preprocessing of the construction data and fusing the preprocessed construction data to obtain real-time monitoring data includes: Using a Kalman filter algorithm to smooth the inclination angle and acceleration of the shield machine to obtain a preprocessed inclination angle and a preprocessed acceleration; Performing multi-resolution analysis on the stratum structure signal of the target area by using a wavelet transform algorithm to obtain a pre-processed stratum structure signal; Performing image analysis on the tunneling face image of the target area using a convolutional neural network to obtain a preprocessed tunneling face image; Using a support vector machine to classify and predict the mud parameters of the target area to obtain preprocessed mud parameters; The extended Kalman filter algorithm is used to fuse the preprocessed inclination angle, the preprocessed acceleration, the preprocessed stratum structure signal, the preprocessed tunneling face image and the preprocessed mud parameters to generate the real-time monitoring data.

4. The intelligent direction adjustment method for shield tunneling in volcanic ash strata according to claim 3 is characterized in that: The pre-processed stratum structure signal includes rock, pores, water and stratum heterogeneity. The multi-resolution analysis of the stratum structure signal of the target area using the wavelet transform algorithm to obtain the pre-processed stratum structure signal includes: determining a target wavelet function according to the frequency characteristics of the stratum structure signal and the target resolution; Performing wavelet decomposition on the stratum structure signal using the target wavelet function to obtain wavelet components, including low-frequency components and high-frequency components; Noise signals in the high-frequency component that are higher than a preset frequency threshold are filtered out, and the rocks, pores, and moisture in the target area are identified by analyzing the mutation points and intensities of the denoised high-frequency component. The stratigraphic heterogeneity in the target area is identified by analyzing the fluctuations of the low-frequency component.

5. The intelligent direction adjustment method for shield tunneling in volcanic ash strata according to claim 2 is characterized in that: The analyzing the historical excavation error data and the real-time monitoring data using a deep learning model to generate an excavation direction adjustment instruction further includes: Performing regression analysis on the historical excavation error data using a support vector regression model to determine the excavation direction deviation trend and excavation direction deviation factors; Using a Bayesian network to comprehensively analyze the geological parameters of the stratum structure signal to determine the influence weight of each geological stratum segment in the target area on the tunneling direction; Determining the thrust distribution of each jacking cylinder in the shield machine by using a fuzzy adaptive control algorithm and based on the historical excavation error data and the stratum structure signal; The deep learning model is used to process the real-time monitoring data, the excavation direction deviation trend, the excavation direction deviation factor, the influence weight of each geological segment in the target area on the excavation direction, and the thrust distribution of each jacking cylinder in the shield machine to generate the excavation direction adjustment instruction.

6. The intelligent direction adjustment method for shield tunneling in volcanic ash strata according to claim 5 is characterized in that: The support vector regression model is used to perform regression analysis on the historical excavation error data to determine the excavation direction deviation trend and the excavation direction deviation factors, including: Performing regression analysis on the historical tunneling error data using the support vector regression model to generate an error prediction curve; Identifying the linear trend, periodic fluctuation and mutation point of the tunneling displacement error based on the error prediction curve to obtain the tunneling direction deviation trend; Acquire additional data collected by additional sensors, perform correlation analysis on the additional data and the historical excavation error data, determine basic factors that are correlated with the excavation displacement error, and use the support vector regression model to determine target factors among the basic factors whose correlation with the excavation displacement error is greater than a preset correlation threshold, thereby obtaining the excavation direction deviation factor.

7. The intelligent direction adjustment method for shield tunneling in volcanic ash strata according to claim 5 is characterized in that: The method of using the Bayesian network to comprehensively analyze the geological parameters of the stratum structure signal to determine the influence weight of each geological stratum segment in the target area on the tunneling direction includes: Defining nodes of the Bayesian network according to the geological parameters, and defining conditional dependencies and a conditional probability table between the nodes according to the historical tunneling error data, wherein the conditional probability table is used to describe the probability distribution of a node under given conditions of its parent node; Inputting each of the geological parameters into the Bayesian network to calculate the probability of the excavation direction deviation of each of the geological parameters under given conditions; According to the probability of the excavation direction deviation of each geological parameter under given conditions and the degree of influence of each excavation direction deviation on the excavation direction, the influence weight of each geological layer section in the target area on the excavation direction is calculated.

8. The intelligent direction adjustment method for shield tunneling in volcanic ash strata according to claim 5 is characterized in that: The method of using a fuzzy adaptive control algorithm and determining the thrust distribution of each jacking cylinder in the shield machine according to the historical tunneling error data and the stratum structure signal includes: Converting the geological parameters of the stratum structure signal, the historical excavation error data, and the real-time excavation error data into fuzzy sets, and defining the range of each fuzzy set using a preset membership function; Determining fuzzy rules based on the historical tunneling error data, applying the fuzzy rules to each of the fuzzy sets using the fuzzy adaptive control algorithm, and calculating the fuzzy thrust distribution of each of the jacking cylinders; The fuzzy thrust distribution of each jacking cylinder is defuzzified by using a defuzzification algorithm to obtain the thrust distribution of each jacking cylinder.

9. The intelligent direction adjustment method for shield tunneling in volcanic ash strata according to claim 5, characterized in that: The deep learning model includes a convolutional neural network module, a long short-term memory network module, and a comprehensive output module. The deep learning model is used to process the real-time monitoring data, the tunneling direction deviation trend, the tunneling direction deviation factor, the influence weight of each geological layer segment in the target area on the tunneling direction, and the thrust distribution of each jacking cylinder in the shield machine to generate the tunneling direction adjustment instruction, including: Extracting spatial feature data of the real-time monitoring data through the convolutional neural network module; Analyzing the time correlation between the historical excavation error data and the real-time monitoring data through the long short-term memory network module to obtain time series data; Combining the spatial feature data with the time series data through the comprehensive output module to generate a prediction result, wherein the prediction result is used to represent a prediction of formation change trends and migration risks; Based on the prediction results, the target excavation direction adjustment amount is calculated, and the excavation direction adjustment instruction is generated by combining the target excavation direction adjustment amount, the excavation direction deviation trend, the excavation direction deviation factor, the influence weight of each geological segment in the target area on the excavation direction, and the thrust distribution of each jacking cylinder in the shield machine using a PID control algorithm.

10. The intelligent direction adjustment method for shield tunneling in volcanic ash strata according to claim 2, characterized in that: Generating a mud cake prevention instruction according to the real-time monitoring data, and adjusting the removal parameters of the mud cake removal device according to the mud cake prevention instruction, includes: Identifying the mud cake location, mud cake thickness, mud cake distribution range, and mud cake physical properties based on the tunneling face image; generating the mud cake prevention and control instruction according to the mud cake position, the mud cake thickness, the mud cake distribution range, the mud cake physical properties and the mud parameters; The cleaning force and frequency of the mud cake removal device are adjusted according to the mud cake prevention and control instructions.

Images