Ground subsidence self-adaptive regulation and control method for double control of shield deslagging volume and soil bin pressure

Abstract

The invention discloses a shield deslagging volume and soil bin pressure double-controlled land subsidence self-adaptive regulation and control method, and belongs to the technical field of underground engineering construction. The method comprises the following steps: S1, data acquisition of monitoring equipment and various sensors; s2, constructing and calibrating an initial model; s3, setting and executing an initial regulation and control strategy; s4, adaptive regulation and control and model optimization are carried out; s5, automatic early warning and manual intervention are carried out; and S6, regulation and control effect verification and construction adjustment are carried out. According to the method, by monitoring the deslagging volume and the soil bin pressure in real time and combining multiple factors such as geological conditions and shield tunneling machine parameters, the deslagging speed and the soil bin pressure in the construction process are automatically adjusted, so that ground subsidence is reduced, and the construction safety and precision are improved. The device is used in the shield construction process, and self-adaptive regulation and control of ground subsidence are achieved through dual control over the deslagging volume and the soil bin pressure.

Description

Technical Field

[0001] The present invention belongs to the technical field of underground engineering construction, and in particular relates to a method for realizing adaptive control of ground settlement by dual control of slag discharge volume and soil bin pressure during shield construction. Background Art

[0002] Shield tunnel construction is a widely used excavation method in current underground engineering, and is particularly suitable for long-distance, deep-buried and complex geological conditions. The shield machine cuts the soil by rotating the cutterhead and transports the resulting slag to the surface to form a tunnel structure. However, the ground subsidence problem during shield construction is a common and challenging engineering problem, especially in soft soil, sand layers or areas with abundant groundwater. Ground subsidence may pose serious safety hazards to surrounding buildings, roads, underground pipelines, etc. Therefore, how to effectively control ground subsidence during shield construction has become an important technical issue of concern to the engineering community.

[0003] In the construction of earth pressure balance shield, the slag discharge volume and soil bin pressure are two key parameters. Among them, the slag discharge volume refers to the volume of slag discharged by the shield machine during the excavation process. By adjusting the slag discharge volume, the amount of soil in the soil bin can be controlled, thereby affecting the ground stress distribution in front of the shield machine. However, there are certain limitations in the control of the slag discharge volume. If the slag discharge volume is too large, it may lead to insufficient soil in the soil bin, weakening the support for the excavation surface, and causing excessive ground settlement; and insufficient slag discharge volume will cause soil accumulation in the soil bin, increase the pressure in the soil bin (soil bin pressure refers to the pressure state in the soil bin of the shield machine). Although it can reduce ground settlement to a certain extent, excessive soil bin pressure may cause instability of the excavation surface or damage to the shield machine equipment.

[0004] On the other hand, by adjusting the soil bin pressure, the ground stress field in front of the shield machine can be balanced, thereby controlling ground settlement. A higher soil bin pressure can enhance the support for the excavation face and reduce the risk of ground settlement; however, too high a soil bin pressure may lead to poor slag discharge, excessive torque on the shield machine cutterhead, and even dangerous situations such as sudden surges of soil on the excavation face; while too low a soil bin pressure may lead to insufficient support for the excavation face, resulting in soil collapse and causing greater ground settlement.

[0005] Although existing technologies have made certain progress in controlling the volume of slag discharge and the pressure of the earth bin, since the two are usually controlled independently, they cannot form an effective linkage and cannot flexibly respond to the needs of ground subsidence control under complex and changeable geological conditions. Especially when encountering complex geological conditions such as alternation of soft soil and hard rock, and changes in aquifer thickness, the control of a single parameter is often difficult to achieve accurate control of ground subsidence, and there are certain engineering risks. Therefore, how to comprehensively consider the volume of slag discharge and the pressure of the earth bin, and propose a more accurate and efficient ground subsidence control method to adapt to shield construction under complex geological conditions, has become a technical problem that urgently needs to be solved.

[0006] Based on this background, the present invention proposes a method for adaptively controlling ground subsidence by dual-controlling the shield slag discharge volume and soil bin pressure. Through real-time monitoring and comprehensive control of the slag discharge volume and soil bin pressure, it can better adapt to different geological conditions, achieve precise control of ground subsidence, reduce construction risks, and improve project safety and efficiency. Summary of the invention

[0007] The purpose of the present invention is to provide a method for adaptively controlling ground subsidence by dual control of slag discharge volume and soil bin pressure during shield construction. The method automatically adjusts the slag discharge speed and soil bin pressure during construction by real-time monitoring of slag discharge volume and soil bin pressure, combined with multiple factors such as geological conditions and shield machine parameters, to reduce the occurrence of ground subsidence and improve construction safety and accuracy.

[0008] To achieve the above object, the technical solution adopted by the present invention is:

[0009] A method for adaptively controlling ground settlement by controlling shield slag discharge volume and soil bin pressure, the method comprising the following steps:

[0010] Step S1: Data collection of monitoring equipment and various sensors;

[0011] Use various sensors and monitoring equipment to collect various data including shield machine working status, geological environment, slag output and surface settlement, and transmit the data to the central control system;

[0012] Step S2: initial model construction and calibration;

[0013] Step S21: cleaning, standardizing and intelligently filtering the various data collected in step S1;

[0014] Step S22: initial model construction;

[0015] Step S23: calibrating the initial model;

[0016] Step S3: initial control strategy setting and execution;

[0017] Step S31: generating a parameter control strategy;

[0018] Step S32: Automatically perform regulation;

[0019] Step S33: continuous regulation and data storage;

[0020] Step S4: Adaptive regulation and model optimization;

[0021] Step S41: adaptive learning;

[0022] Step S42: model parameter optimization;

[0023] The specific process of model parameter optimization includes gradient optimization, genetic algorithm improvement and multi-objective optimization;

[0024] Step S5: automatic warning and manual intervention;

[0025] Step S6: Verification of control effect and construction adjustment.

[0026] Furthermore, in step S1, the data collection of the monitoring equipment and various sensors includes the following steps:

[0027] Step S11: Deployment of monitoring equipment and various sensors; specifically:

[0028] The soil bin pressure sensor is arranged inside the soil bin of the shield machine;

[0029] Install geological parameter sensors at the front and periphery of the shield machine;

[0030] Deploy land subsidence monitoring equipment on buildings above the construction site or in sensitive areas;

[0031] Install the propulsion speed sensor on the propulsion cylinder or hydraulic system of the shield machine, install the cutterhead speed sensor on the cutterhead drive system of the shield machine, and install the thrust sensor on the thrust mechanism of the shield machine to monitor the propulsion speed, cutterhead speed, and thrust parameters of the shield machine respectively;

[0032] The slag volume sensor is installed on the conveyor belt of the shield machine slag discharge system;

[0033] Step S12: data collection;

[0034] The ground subsidence monitoring equipment and various sensors are connected to the central control system signal, and the data collected by the ground subsidence monitoring equipment and various sensors are transmitted to the central control system in real time.

[0035] Furthermore, in step S21, the collected multiple data are cleaned, standardized and intelligently filtered, and the specific steps are as follows:

[0036] Step S211: In the data preprocessing stage, the central control system screens the data collected by the sensor and removes abnormal values ​​and noise in the sensor data through the Kalman filter algorithm. The Kalman filter estimates the real data using the following formula:

[0037]

[0038] in: is the estimated value of the current state, is the estimated value of the previous state, z k is the current observation value, H is the observation matrix, K k is the Kalman gain, defined as:

[0039]

[0040] Where: P k-1 is the error covariance matrix of the previous moment, R is the observation noise covariance;

[0041] Step S212: Standardize the preprocessed data to convert data from different sources into a unified standard unit. The standardization method uses the Z-Score standardization method:

[0042]

[0043] Where: X is the original data, μ is the mean, σ is the standard deviation;

[0044] Step S213: Intelligently filter the standardized data. The central control system uses a low-pass filter to remove high-frequency noise and retain low-frequency signals. The formula is:

[0045] y[n]=αy[n-1]+(1-α)x[n]

[0046] Where: y[n] is the output after filtering, y[n-1] is the filter output value at the previous moment, x[n] is the current input, and α is the filter coefficient, which controls the smoothness and has a value between 0 and 1.

[0047] In step S22, the specific process of constructing the initial model is as follows:

[0048] The central control system uses a neural network algorithm to establish an initial model; each layer of the neural network model is calculated as follows:

[0049] h (l) =f(W (l) h (l-1) +b (l) )

[0050] Where: h(l) is the output of the lth layer, h (l-1) is the output of the l-1th layer, W (l) is the weight matrix of the lth layer, b (l) is the bias of the lth layer, f is the activation function, and the ReLU function is used: f(x) = max(0,x);

[0051] During the optimization of the initial model, the mean square error loss function is used to measure the difference between the model's predicted value and the true value:

[0052]

[0053] Where: L is the loss function, N is the number of samples, y i is the true value of the i-th sample, is the model prediction value;

[0054] Update the model's weights and biases using the back-propagation algorithm and gradient descent:

[0055]

[0056] Where: η is the learning rate, is the gradient of the loss function with respect to the weights.

[0057] Furthermore, in step S31, the specific process of generating the parameter control strategy is:

[0058] The central control system defines a multi-objective optimization function including the slag discharge volume and the soil bin pressure; the multi-objective optimization function is expressed as:

[0059]

[0060] Where: S(x) represents the ground settlement, M(x) and P(x) represent the current slag discharge and soil bin pressure respectively, M E and P E They represent the expected slag discharge volume and soil bin pressure value respectively, w1, w2, and w3 are S(x), The weight coefficient of .

[0061] Furthermore, in step S32, the specific process of automatically executing the control is:

[0062] The generated control strategy is converted into the operating instructions of the shield machine through the central control system. The operating instructions include specific slag discharge speed settings, soil bin pressure adjustment values ​​and propulsion speed settings. The shield machine automatically adjusts the corresponding operating parameters according to the received operating instructions to achieve the control objectives.

[0063] Furthermore, in step S33, the specific process of continuous regulation and data storage is:

[0064] In the early stage of excavation, the real-time data of slag volume, soil bin pressure and ground settlement are continuously obtained through the slag volume sensor, soil bin pressure sensor and ground settlement monitoring sensor, and continuous optimization is carried out according to steps S31 and S32 until the settlement control requirements are met by only one parameter adjustment. At the same time, the monitoring data is stored in the cloud and local database; the monitoring data includes slag volume, soil bin pressure, advancement speed, ground settlement changes, and control strategies.

[0065] Furthermore, in step S41, the adaptive learning includes the following steps:

[0066] Step S411: preprocessing storage data;

[0067] First, the central control system obtains raw data from ground subsidence monitoring equipment and various sensors, including slag discharge, soil bin pressure, and ground subsidence, which contain abnormal values. By setting statistical thresholds, obviously abnormal data points are eliminated; the formula is as follows:

[0068] IQR=Q3-Q1

[0069] δ∈[Q1-1.5×IQR,Q3+1.5×IQR]

[0070] Where: Q1 and Q3 are the first and third quartiles of the data respectively, and δ is the outlier threshold;

[0071] Subsequently, the central control system normalizes the slag discharge volume and soil bin pressure data; the details are as follows:

[0072]

[0073] Among them: x is the original data, x′ is the normalized data;

[0074] Step S412: feature extraction;

[0075] During the construction process, the time series features are extracted through the sliding window method to capture the temporal dynamic characteristics of the data; the details are as follows:

[0076]

[0077] In addition, the central control system will also extract the interaction characteristics between the slag volume and the soil bin pressure;

[0078] Step S413: model training and verification;

[0079] The random forest algorithm is selected as the initial model, and each tree is trained with different subsamples and feature subsets; the formula is as follows:

[0080]

[0081] Where: T i (x) is the predicted value of the i-th decision tree for input x;

[0082] In order to evaluate the generalization ability of the model, the central control system uses the K-fold cross-validation method to divide the data set into k subsets, train with k-1 subsets each time, and test the model with the remaining subset to calculate the average error; the details are as follows:

[0083]

[0084] Where: y ij is the actual observed value of the jth data point in the ith subset, is the predicted value of the jth data point in the i-th subset.

[0085] Furthermore, in step S42, the gradient optimization, genetic algorithm improvement and multi-objective optimization include the following steps:

[0086] Step S421: Gradient optimization; the central control system sets a loss function based on the error between the actual ground subsidence and the predicted value, and optimizes it through a gradient descent algorithm:

[0087]

[0088] By calculating the gradient of the loss function, the central control system gradually adjusts the model parameters to make the predicted results closer to the actual values:

[0089]

[0090] Where: θ is the model parameter, η is the learning rate;

[0091] Step S422: Genetic algorithm improvement; the central control system further optimizes the model parameters through the genetic algorithm, first generates multiple parameter combinations, then selects the combination with the highest fitness based on the loss function, performs crossover operations between the parameter combinations with high fitness, and introduces random mutations;.

[0092] θ new =α·θ1+(1-α)·θ2+variation term

[0093] Among them: α is the cross coefficient, and the variation term is the random disturbance;

[0094] Step S423: multi-objective optimization; the central control system considers multiple objectives at the same time and constructs a multi-objective optimization model; the objective function is as follows:

[0095] F(θ)=[J1(θ),J2(θ),...,J k (θ)]

[0096] Where: F(θ) is the objective function vector in multi-objective optimization, each J k (θ) represents an optimization objective.

[0097] Furthermore, in step S5, the specific steps of automatic warning and manual intervention are:

[0098] Step S51: setting warning thresholds; setting safety thresholds for slag discharge volume, soil bin pressure, and ground settlement, and adjusting them when geological conditions change according to geological exploration reports to reflect the risk levels at different construction stages;

[0099] Step S52: Early warning triggering and operator response; when the central control system detects that a certain parameter exceeds a preset threshold, the early warning information is transmitted to the operator's terminal device and a warning is issued at the same time;

[0100] Step S53: Operator response and manual control; after receiving the warning signal, the operator selects a response measure according to the risk analysis report displayed by the central control system;

[0101] Step S54: Recording and feedback: Record the process and results of manual intervention, and feed this information back to the central control system for subsequent analysis and optimization of the central control system.

[0102] Furthermore, in step S6, the specific steps of the control effect verification and construction adjustment are:

[0103] Step S61: Effect evaluation; the central control system automatically generates a parameter change report, compares and analyzes key parameters such as slag discharge volume, soil bin pressure, and ground settlement before and after regulation to determine whether the regulation measures are effective; focuses on evaluating the changing trend of ground settlement, and through dynamic analysis of ground settlement monitoring data, the central control system generates a settlement trend chart to evaluate the changes in ground settlement and compare it with the expected target;

[0104] Step S62: Optimize the excavation plan.

[0105] The beneficial effects of the present invention relative to the prior art are:

[0106] 1. Precise control and real-time dynamic regulation. By comprehensively monitoring the key parameters in the shield construction process, the central control system can obtain accurate data in real time, thereby better predicting and controlling the risk of ground subsidence. Through the dynamic feedback mechanism, the central control system can immediately adjust the construction parameters when the ground subsidence exceeds expectations, ensuring the flexibility and responsiveness of the construction, and effectively preventing serious ground subsidence problems.

[0107] 2. Multi-parameter collaborative control and comprehensive analysis. The present invention adopts a multi-parameter collaborative control model (initial model) to combine the slag discharge volume, soil bin pressure, geological conditions, shield machine operation status and ground subsidence monitoring data for comprehensive analysis and regulation. This multi-dimensional data analysis method enhances the adaptability of the central control system to complex geological conditions and significantly improves the accuracy and reliability of the regulation effect.

[0108] 3. Adaptive learning and model optimization. The central control system continuously accumulates historical data generated during the construction process, uses machine learning algorithms for adaptive learning, and continuously optimizes the control model. As the construction progresses, the central control system can automatically adjust parameters under different geological conditions, gradually improve the control accuracy of ground subsidence and construction efficiency, making the central control system more intelligent and efficient in long-term operation.

[0109] 4. Automatic early warning and manual intervention guarantee. The present invention sets a safety threshold. When the central control system detects that ground subsidence or other key parameters exceed the safety range, it will immediately and automatically send out an early warning signal. At the same time, the central control system also provides a manual intervention option, combining the dual guarantee mechanism of automatic control and manual intervention to further enhance the safety of the construction process and prevent accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0110] Figure 1 This is the control flow chart for the early stage of excavation;

[0111] Figure 2 It is a flowchart for adaptive learning;

[0112] Figure 3 This is the control flow chart after adaptive learning;

[0113] Figure 4 This is a block diagram of the signal connection between the ground subsidence monitoring equipment and various sensors and the central control system.

[0114] The names and reference numerals of the components involved in the above drawings are as follows:

[0115] Soil bin pressure sensor 101, geological parameter sensor 102, ground subsidence monitoring equipment 103, propulsion speed sensor 104, cutter head rotation speed sensor 105, thrust sensor 106, slag volume sensor 107, central control system 108. DETAILED DESCRIPTION

[0116] Specific implementation method 1: Figure 1-Figure 4 As shown, this embodiment discloses a method for adaptively controlling ground settlement by controlling shield slag discharge volume and soil bin pressure, the method comprising the following steps:

[0117] Step S1: Data collection of monitoring equipment and various sensors;

[0118] Soil bin pressure sensors 101 and geological parameter sensors 102 are placed at key locations of the shield machine, and slag volume sensors 107 are installed on the slag belt of the shield machine; ground subsidence monitoring equipment 103 is arranged in key monitoring areas on the surface. Afterwards, the central control system 108 begins to collect a variety of data including the working status of the shield machine, geological environment, slag volume, and surface subsidence to ensure comprehensive and real-time monitoring of key parameters during the construction process.

[0119] Using various sensors and monitoring equipment to collect various data including shield machine working status, geological environment, slag output and surface settlement, and transmit the various data to the central control system 108;

[0120] Step S2: constructing and calibrating an initial model (i.e., an initial multi-parameter collaborative control model);

[0121] Before the start of shield construction, the initial model (initial multi-parameter collaborative control model) in the central control system 108 is constructed by integrating historical data, theoretical research and expert experience. The model is based on historical data of similar projects in the past, combined with empirical formulas in the fields of soil mechanics, rheology, fluid mechanics, etc., to determine the range of key parameters in shield construction. In order to ensure the accuracy of the data and the reliability of the model, the central control system 108 performs a series of preprocessing, standardization and intelligent filtering on the collected data, and constructs an initial model in combination with neural network algorithms and optimization strategies. Subsequently, on-site surveys and data collection are carried out, and detailed analysis of geological parameters, groundwater levels and stratum composition is carried out. Through data processing and real-time analysis, the initial model is calibrated and optimized to ensure that it adapts to actual geological conditions.

[0122] Step S21: cleaning, standardizing and intelligently filtering the various data collected in step S1;

[0123] Step S22: initial model construction;

[0124] Step S23: calibrating the initial model (in the early stage of construction, the initial model constructed in the above step S22 is calibrated in combination with the data collected by the on-site survey to make it better match the on-site data and ensure the actual application effect of the initial model);

[0125] Step S3: initial control strategy setting and execution;

[0126] According to the initial model constructed in step 2, the central control system 108 generates a preliminary parameter control strategy, covering the set values ​​of the shield machine's propulsion speed, cutterhead speed, slag discharge volume, and soil bin pressure. These set values ​​are based on the prediction results of the initial model, aiming to balance the stratum pressure in front of the excavation and ensure that the ground settlement is controlled within a safe range. In the initial control, the central control system 108 sets a certain safety factor to cope with the possible uncertainty of the initial model. As the shield machine starts, the central control system 108 monitors the data of sensors and monitoring equipment in real time, compares the real-time data with the initial model prediction, and executes initial control measures.

[0127] Step S31: generating a parameter control strategy;

[0128] Step S32: Automatically perform regulation;

[0129] Step S33: continuous regulation and data storage;

[0130] Step S4: Adaptive regulation and model optimization;

[0131] During the initial construction process, the central control system 108 continuously monitors various key parameters, including slag volume, soil bin pressure, ground settlement and shield machine operation status, and performs dynamic feedback control. Based on the changing trend of real-time data, the central control system 108 automatically adjusts the control measures to ensure that the ground settlement during the construction process remains within a safe range. At the same time, the central control system 108 will accumulate and analyze historical data and continuously optimize the control model to make it more accurately adapt to the construction needs under different geological conditions. Through adaptive learning, the central control system 108 gradually improves the accuracy of the control strategy and achieves more effective ground settlement control.

[0132] Step S41: Adaptive learning (using a machine learning algorithm, taking historical data as a training set, regularly training and updating the model to improve its prediction accuracy of ground subsidence);

[0133] Step S42: model parameter optimization;

[0134] Model parameter optimization aims to make the initial model more accurate in predicting and controlling ground subsidence through continuous algorithm optimization and parameter adjustment.

[0135] The specific process of model parameter optimization includes gradient optimization, genetic algorithm improvement and multi-objective optimization;

[0136] Step S5: automatic warning and manual intervention;

[0137] In complex or unexpected situations, the central control system 108 will detect that the risk of ground subsidence exceeds the capacity of automatic control (set threshold). At this time, the central control system 108 will issue a warning signal to remind the operator to intervene manually. The central control system 108 will provide detailed warning information and recommended operating steps. The operator can manually adjust the construction parameters or take emergency measures based on this information to avoid safety hazards caused by excessive ground subsidence.

[0138] Step S6: Verification of control effect and construction adjustment.

[0139] After implementing the control plan, the central control system 108 will verify the actual control effect of the ground subsidence and evaluate the effectiveness of the control measures. If it is found that the control effect does not meet expectations, the central control system 108 will automatically adjust the parameters or generate a new control plan based on real-time data to ensure that the ground subsidence is always within the controllable range during the construction process. In this process, the central control system 108 will also optimize the parameter settings of subsequent construction to improve construction efficiency and safety.

[0140] Furthermore, in step S1, the data collection of the monitoring equipment and various sensors includes the following steps:

[0141] Step S11: Deployment of monitoring equipment and various sensors; specifically:

[0142] The soil bin pressure sensor 101 is arranged inside the soil bin of the shield machine (to ensure that the pressure data inside the soil bin can be obtained in real time);

[0143] Install the geological parameter sensor 102 at the front end and periphery of the shield machine (for real-time monitoring of the geological conditions ahead, such as soil density and moisture content);

[0144] Deploy a ground subsidence monitoring device 103 (such as a laser rangefinder, an inclinometer or a GNSS device) on a building above the construction site or in a sensitive area (for monitoring ground subsidence);

[0145] The propulsion speed sensor 104 is installed on the propulsion cylinder or hydraulic system of the shield machine, the cutterhead speed sensor 105 is installed on the cutterhead drive system of the shield machine, and the thrust sensor 106 is installed on the thrust mechanism of the shield machine, which are used to monitor the propulsion speed, cutterhead speed, and thrust parameters of the shield machine respectively (the propulsion speed sensor 104, the cutterhead speed sensor 105, and the thrust sensor 106 constitute a shield machine operation parameter sensor);

[0146] The slag volume sensor 107 is installed at the conveyor belt of the shield machine slag discharge system (for real-time monitoring of the discharged slag volume);

[0147] Step S12: data collection;

[0148] Connect the ground subsidence monitoring equipment 103 and various sensors to the central control system 108, and transmit the data collected by the ground subsidence monitoring equipment 103 and various sensors to the central control system 108 in real time (via a wired or wireless network) (the central control system 108 starts to collect a variety of data including the working status of the shield machine, geological environment, slag output, and surface subsidence to ensure comprehensive and real-time monitoring of key parameters in the construction process);

[0149] Preferably, the central control system 108 is provided with data caching and backup functions (to ensure the security and integrity of the data).

[0150] Furthermore, in step S21, the collected multiple data are cleaned, standardized and intelligently filtered, and the specific steps are as follows:

[0151] Step S211: In the data preprocessing stage, the central control system 108 screens the data collected by the sensor and removes abnormal values ​​and noise in the sensor data through the Kalman filter algorithm. The Kalman filter estimates the real data using the following formula:

[0152]

[0153] in: is the estimated value of the current state, is the estimated value of the previous state, z k is the current observation value, H is the observation matrix, K k is the Kalman gain, defined as:

[0154]

[0155] Where: P k-1 is the error covariance matrix of the previous moment, R is the observation noise covariance;

[0156] Step S212: Standardize the preprocessed data to convert data from different sources into a unified standard unit. The standardization method uses the Z-Score standardization method:

[0157]

[0158] Where: X is the original data, μ is the mean, σ is the standard deviation;

[0159] Step S213: Intelligently filter the standardized data. The central control system 108 uses a low-pass filter to remove high-frequency noise and retain low-frequency signals (frequency is not greater than 0.5 Hz). The formula is:

[0160] y[n]=αy[n-1]+(1-α)x[n]

[0161] Where: y[n] is the output after filtering, y[n-1] is the filter output value at the previous moment, x[n] is the current input, and α is the filter coefficient, which controls the smoothness and has a value between 0 and 1.

[0162] In step S22, the specific process of constructing the initial model is as follows:

[0163] After data preprocessing, the central control system 108 compares the processed data with historical engineering data and sets key parameters in the initial model based on theoretical research and expert experience;

[0164] The central control system 108 establishes an initial model (i.e., an initial multi-parameter collaborative control model) through a neural network algorithm; each layer of the neural network model is calculated as follows:

[0165] h (l) =f(W (l) h (l-1) +b (l) )

[0166] Where: h (l) is the output of the lth layer, h (l-1) is the output of the l-1th layer, W (l) is the weight matrix of the lth layer, b (l) is the bias of the lth layer, f is the activation function, and the ReLU function is used: f(x) = max(0,x);

[0167] During the optimization of the initial model, the mean square error loss function (MSE) is used to measure the difference between the model's predicted value and the true value:

[0168]

[0169] Where: L is the loss function, N is the number of samples, y i is the true value of the i-th sample, is the model prediction value;

[0170] Update the model's weights and biases using the back-propagation algorithm and gradient descent:

[0171]

[0172] Where: η is the learning rate, is the gradient of the loss function with respect to the weights.

[0173] Furthermore, in step S31, the specific process of generating the parameter control strategy is:

[0174] The central control system 108 defines a multi-objective optimization function including the slag discharge amount and the soil bin pressure (to simultaneously control the ground settlement and construction efficiency); the multi-objective optimization function is expressed as:

[0175]

[0176] Where: S(x) represents the ground settlement, M(x) and P(x) represent the current slag discharge and soil bin pressure respectively, M E and P E They represent the expected slag discharge volume and soil bin pressure value respectively, w1, w2, and w3 are S(x), The weight coefficient of (to ensure that the optimization objective function can balance the influence of different factors);

[0177] Preferably, in the early optimization process, in order to ensure safety, the central control system 108 preferably sets a larger safety factor when adjusting M(x) and P(x), such as setting the safety factor to 1.5. In the initial regulation, the central control system 108 sets a certain safety factor to cope with the uncertainty that may exist in the initial (multi-parameter collaborative control) model.

[0178] Furthermore, in step S32, the specific process of automatically executing the control is:

[0179] The generated control strategy is converted into the operating instructions of the shield machine through the central control system 108. The operating instructions include specific slag discharge speed settings, soil bin pressure adjustment values ​​and propulsion speed settings. The shield machine automatically adjusts the corresponding operating parameters according to the received operating instructions to achieve the control target.

[0180] Furthermore, in step S33, the specific process of continuous regulation and data storage is:

[0181] At the initial stage of excavation, the real-time data of slag discharge volume, soil bin pressure and ground settlement are continuously obtained through the slag volume sensor 107, the soil bin pressure sensor 101 and the ground settlement monitoring sensor, and the optimization is continuously carried out according to steps S31 and S32 until the settlement control requirements are met by only one parameter adjustment. At the same time, the monitoring data is stored in the cloud and the local database (the storage period can be set to once a day or once at the end of each construction stage); the monitoring data includes slag discharge volume, soil bin pressure, advancement speed, ground settlement changes, and control strategies (to facilitate subsequent adaptive learning and model optimization).

[0182] Furthermore, in step S41, the adaptive learning includes the following steps:

[0183] Step S411: preprocessing storage data;

[0184] First, the central control system 108 obtains raw data from the ground subsidence monitoring equipment 103 and various sensors, including slag discharge, soil bin pressure, and ground subsidence, which contain abnormal values. By setting statistical thresholds (such as standard deviation method or IQR (interquartile range) method), obviously abnormal data points are eliminated; the formula is as follows:

[0185] IQR=Q3-Q1

[0186] δ∈[Q1-1.5×IQR,Q3+1.5×IQR]

[0187] Where: Q1 and Q3 are the first and third quartiles of the data respectively, and δ is the outlier threshold;

[0188] Subsequently, in order to unify data of different dimensions, the central control system 108 normalizes the slag discharge volume and soil bin pressure data to ensure the stability of model training; the details are as follows:

[0189]

[0190] Among them: x is the original data, x′ is the normalized data;

[0191] Step S412: feature extraction;

[0192] During the construction process, ground settlement, slag discharge, excavation speed, soil bin pressure, and shield machine thrust have time correlation. The sliding window method is used to extract time series features, such as mean, variance, and trend, to capture the time dynamic characteristics of the data; the details are as follows:

[0193]

[0194] In addition, the central control system 108 will also extract the interaction characteristics between the slag volume and the soil bin pressure, such as their product, ratio, etc., to help the model better understand the relationship between the two and the ground settlement;

[0195] Step S413: model training and verification;

[0196] The Random Forest algorithm is selected as the initial model. The Random Forest uses multiple decision trees to vote to reduce the risk of overfitting. Each tree is trained with different subsamples and feature subsets. The formula is as follows:

[0197]

[0198] Where: T i (x) is the predicted value of the i-th decision tree for input x;

[0199] In order to evaluate the generalization ability of the model, the central control system 108 uses the K-fold cross-validation method to divide the data set into k subsets, and uses k-1 subsets for training each time, and the remaining subset to test the model and calculate the average error; the details are as follows:

[0200]

[0201] Where: y ij is the actual observed value of the jth data point in the ith subset, is the predicted value of the jth data point in the i-th subset.

[0202] Furthermore, in step S42, the gradient optimization, genetic algorithm improvement and multi-objective optimization include the following steps:

[0203] Step S421: Gradient optimization: When the initial model is constructed, the central control system 108 first sets the key parameters in the model based on historical data and expert experience. As the excavation progresses, the central control system 108 sets a loss function, such as mean square error (MSE), based on the error between the actual ground settlement and the predicted value, and optimizes it through a gradient descent algorithm:

[0204]

[0205] By calculating the gradient of the loss function, the central control system 108 gradually adjusts the model parameters to make the prediction results closer to the actual values:

[0206]

[0207] Where: θ is the model parameter, η is the learning rate;

[0208] Step S422: Genetic algorithm improvement; the central control system 108 further optimizes the model parameters through the genetic algorithm, first generates multiple parameter combinations (populations), then selects the combination with the highest fitness based on the loss function, performs crossover operations between the parameter combinations with high fitness, and introduces random mutations (to explore better parameter combinations);.

[0209] θ new =α·θ1+(1-α)·θ2+variation term

[0210] Among them: α is the cross coefficient, and the variation term is the random disturbance;

[0211] Step S423: multi-objective optimization; the central control system 108 considers multiple objectives (such as minimizing ground settlement and optimizing construction efficiency) at the same time and constructs a multi-objective optimization model; the objective function is as follows:

[0212] F(θ)=[J1(θ),J2(θ),...,J k (θ)]

[0213] Where: F(θ) is the objective function vector in multi-objective optimization, each J k (θ) represents an optimization objective;

[0214] The central control system 108 uses the Pareto optimization method to find a parameter combination that cannot be further improved on all objectives. The Pareto optimal solution θ * The choice ensures that the central control system 108 minimizes the risk of ground subsidence while taking into account construction efficiency.

[0215] Furthermore, in step S5, the specific steps of automatic warning and manual intervention are:

[0216] Step S51: setting warning thresholds; according to the adaptive learning results in step 4 and the historical data in step 3, setting safety thresholds for slag discharge volume, soil bin pressure, and ground settlement, and adjusting them according to geological exploration reports when geological conditions change (such as changes in the softness and hardness of the stratum) to reflect the risk levels of different construction stages;

[0217] Step S52: Warning triggering and operator response: When the central control system 108 detects that a parameter exceeds a preset threshold, the warning information is transmitted to the operator's terminal device (such as a smart phone or a console) (via a wireless signal), and a warning (including an audible warning (siren) and a visual warning (warning light)) is issued at the same time;

[0218] Step S53: Operator response and manual control; after receiving the warning signal, the operator selects a response measure according to the risk analysis report displayed by the central control system 108, such as manually adjusting the advancement speed, slag discharge, soil bin pressure or advancement direction of the shield machine; if necessary, the construction needs to be suspended and the risk needs to be further assessed to deal with complex construction situations;

[0219] Step S54: Recording and feedback; recording the process and results of manual intervention, and feeding back this information to the central control system 108 for subsequent analysis and optimization of the central control system 108 .

[0220] Furthermore, in step S6, the specific steps of the control effect verification and construction adjustment are:

[0221] Step S61: Effect evaluation; the central control system 108 automatically generates a parameter change report, compares and analyzes key parameters such as the slag discharge volume, soil bin pressure, and ground settlement before and after regulation, and determines whether the regulation measures are effective; focuses on evaluating the change trend of ground settlement, and through dynamic analysis of ground settlement monitoring data, the central control system 108 generates a settlement trend chart to evaluate the change of ground settlement and compare it with the expected target;

[0222] Step S62: Optimize the excavation plan (i.e., adjust the construction plan); based on the evaluation results of step S61 above, combined with historical data and current control strategies, update the construction plan, including the shield machine advancement path, speed control, and slag discharge plan, and apply the adjusted strategy to subsequent construction to further reduce ground settlement and ensure smooth construction. The new construction plan can also be used in the initial model construction (step S22 above) in the new working condition.

[0223] The present invention provides a ground settlement adaptive control method for dual-controlling shield slag discharge volume and soil bin pressure, aiming at optimizing ground settlement control in shield construction.

[0224] The method of the present invention first collects soil bin pressure, slag discharge, geological conditions and ground subsidence data in real time by deploying sensors and ground subsidence monitoring equipment at key positions of the shield machine and buildings or sensitive areas above the construction. Based on these data, an initial multi-parameter collaborative control model is established, and on-site calibration is performed to ensure the accuracy of the model. During the construction process, the central control system 108 automatically adjusts the operating parameters of the shield machine, including the propulsion speed and slag discharge amount, according to real-time data to achieve accurate control of ground subsidence. In addition, the central control system 108 adopts adaptive learning technology to continuously optimize the initial multi-parameter collaborative control model by analyzing historical data to improve the control accuracy and construction efficiency. In order to further improve safety, the central control system 108 sets a safety threshold. When the parameters are monitored to exceed the predetermined range, an early warning signal will be automatically issued, and the operator can perform manual intervention based on the early warning information. After the control effect is verified, the central control system 108 will also make necessary adjustments according to the actual construction effect to ensure that the ground subsidence is controlled within a safe range. The method of the present invention integrates dynamic control, adaptive optimization and early warning intervention, and provides comprehensive and intelligent construction safety protection.

[0225] The above are only preferred specific implementation modes of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical solutions and inventive concepts of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for adaptively controlling ground settlement by controlling both shield slag discharge volume and soil bin pressure, characterized in that: The method comprises the following steps: Step S1: Data collection of monitoring equipment and various sensors; Using various sensors and monitoring equipment to collect various data including shield machine working status, geological environment, slag output and surface settlement, and transmitting the various data to a central control system (108); Step S2: initial model construction and calibration; Step S21: cleaning, standardizing and intelligently filtering the various data collected in step S1; Step S22: initial model construction; Step S23: calibrating the initial model; Step S3: initial control strategy setting and execution; Step S31: generating a parameter control strategy; Step S32: Automatically perform regulation; Step S33: continuous regulation and data storage; Step S4: Adaptive regulation and model optimization; Step S41: adaptive learning; Step S42: model parameter optimization; The specific process of model parameter optimization includes gradient optimization, genetic algorithm improvement and multi-objective optimization; Step S5: automatic warning and manual intervention; Step S6: Verification of control effect and construction adjustment.

2. The method for adaptively controlling ground settlement by controlling shield slag discharge volume and soil bin pressure according to claim 1 is characterized in that: In step S1, the data collection of the monitoring equipment and various sensors includes the following steps: Step S11: Deployment of monitoring equipment and various sensors; specifically: Arranging a soil bin pressure sensor (101) inside the soil bin of the shield machine; Installing geological parameter sensors (102) at the front end and periphery of the shield machine; Deploying a ground subsidence monitoring device (103) on a building above the construction site or in a sensitive area; The propulsion speed sensor (104) is installed on the propulsion cylinder or hydraulic system of the shield machine, the cutter head speed sensor (105) is installed on the cutter head drive system of the shield machine, and the thrust sensor (106) is installed on the thrust mechanism of the shield machine, so as to monitor the propulsion speed, cutter head speed and thrust parameters of the shield machine respectively; The slag volume sensor (107) is installed on the conveyor belt of the shield machine slag discharge system; Step S12: data collection; connecting the ground subsidence monitoring equipment (103) and various sensors to the central control system (108) by signal, and transmitting the data collected by the ground subsidence monitoring equipment (103) and various sensors to the central control system (108) in real time.

3. The method for adaptively controlling ground settlement by controlling shield slag discharge volume and soil bin pressure according to claim 1 is characterized in that: In step S21, the collected data are cleaned, standardized and intelligently filtered. The specific steps are as follows: Step S211: In the data preprocessing stage, the central control system (108) screens the data collected by the sensor and removes abnormal values ​​and noise in the sensor data through the Kalman filter algorithm. The Kalman filter estimates the real data using the following formula: in: is the estimated value of the current state, is the estimated value of the previous state, z k is the current observation value, H is the observation matrix, K k is the Kalman gain, defined as: Where: P k-1 is the error covariance matrix of the previous moment, R is the observation noise covariance; Step S212: Standardize the preprocessed data to convert data from different sources into a unified standard unit. The standardization method uses the Z-Score standardization method: Where: X is the original data, μ is the mean, σ is the standard deviation; Step S213: Intelligently filter the standardized data. The central control system (108) uses a low-pass filter to remove high-frequency noise and retain low-frequency signals. The formula is: y[n]=αy[n-1]+(1-α)x[n] Where: y[n] is the output after filtering, y[n-1] is the filter output value at the previous moment, x[n] is the current input, and α is the filter coefficient, which controls the smoothness and has a value between 0 and 1. In step S22, the specific process of constructing the initial model is as follows: The central control system (108) establishes an initial model through a neural network algorithm; each layer of the neural network model is calculated as follows: h (l) =f(W (l) h (l-1) +b (l) ) Where: h (l) is the output of the lth layer, h (l-1) is the output of the l-1th layer, W (l) is the weight matrix of the lth layer, b (l) is the bias of the lth layer, f is the activation function, and the ReLU function is used: f(x) = max(0,x); During the optimization of the initial model, the mean square error loss function is used to measure the difference between the model's predicted value and the true value: Where: L is the loss function, N is the number of samples, y i is the true value of the i-th sample, is the model prediction value; Update the model's weights and biases using the back-propagation algorithm and gradient descent: Where: η is the learning rate, is the gradient of the loss function with respect to the weights.

4. The method for adaptively controlling ground settlement by controlling shield slag discharge volume and soil bin pressure according to claim 1 is characterized in that: In step S31, the specific process of generating the parameter control strategy is: The central control system (108) defines a multi-objective optimization function including the slag discharge amount and the soil bin pressure; the multi-objective optimization function is expressed as: Where: S(x) represents the ground settlement, M(x) and P(x) represent the current slag discharge and soil bin pressure respectively, M E and P E They represent the expected slag discharge volume and soil bin pressure value respectively, w1, w2, and w3 are S(x), The weight coefficient of .

5. The method for adaptively controlling ground settlement by controlling shield slag discharge volume and soil bin pressure according to claim 4 is characterized in that: In step S32, the specific process of automatically executing the control is as follows: The generated control strategy is converted into an operating instruction of the shield machine through the central control system (108), wherein the operating instruction includes a specific slag discharge speed setting, a soil bin pressure adjustment value and a propulsion speed setting value. The shield machine automatically adjusts the corresponding operating parameters according to the received operating instructions to achieve the control target.

6. The method for adaptively controlling ground settlement by controlling shield slag discharge volume and soil bin pressure according to claim 5 is characterized in that: In step S33, the specific process of continuous regulation and data storage is as follows: At the initial stage of excavation, the real-time data of the slag volume, soil bin pressure and ground settlement are continuously obtained through the slag volume sensor (107), the soil bin pressure sensor (101) and the ground settlement monitoring sensor, and the optimization is continuously carried out according to steps S31 and S32 until the settlement control requirements are met by only one parameter adjustment, and the monitoring data is stored in the cloud and the local database at the same time; the monitoring data includes the slag volume, soil bin pressure, advancement speed, ground settlement changes and control strategies.

7. The method for adaptively controlling ground settlement by controlling shield slag discharge volume and soil bin pressure according to claim 6 is characterized in that: In step S41, the adaptive learning includes the following steps: Step S411: preprocessing storage data; First, the central control system (108) obtains raw data from the ground subsidence monitoring equipment (103) and various sensors, including slag discharge, soil bin pressure, and ground subsidence, which contain abnormal values. By setting a statistical threshold, obviously abnormal data points are eliminated; the formula is as follows: IQR=Q3-Q1 δ∈[Q1-1.5×IQR,Q3+1.5×IQR] Where: Q1 and Q3 are the first and third quartiles of the data respectively, and δ is the outlier threshold; Then, the central control system (108) normalizes the slag discharge volume and soil bin pressure data as follows: Among them: x is the original data, x′ is the normalized data; Step S412: feature extraction; During the construction process, the time series features are extracted through the sliding window method to capture the temporal dynamic characteristics of the data; the details are as follows: In addition, the central control system (108) also extracts the interaction characteristics between the slag volume and the soil bin pressure; Step S413: model training and verification; The random forest algorithm is selected as the initial model, and each tree is trained with different subsamples and feature subsets; the formula is as follows: Where: T i (x) is the predicted value of the i-th decision tree for input x; In order to evaluate the generalization ability of the model, the central control system (108) uses the K-fold cross-validation method to divide the data set into k subsets, and uses k-1 subsets for training each time, and the remaining subset to test the model and calculate the average error; the details are as follows: Where: y ij is the actual observed value of the jth data point in the ith subset, is the predicted value of the jth data point in the i-th subset.

8. The method for adaptively controlling ground settlement by controlling shield slag discharge volume and soil bin pressure according to claim 7 is characterized in that: In step S42, the gradient optimization, genetic algorithm improvement and multi-objective optimization include the following steps: Step S421: Gradient optimization; the central control system (108) sets a loss function based on the error between the actual ground subsidence and the predicted value, and optimizes it through a gradient descent algorithm: By calculating the gradient of the loss function, the central control system (108) gradually adjusts the model parameters to make the predicted results closer to the actual values: Where: θ is the model parameter, η is the learning rate; Step S422: Genetic algorithm improvement; the central control system (108) further optimizes the model parameters through the genetic algorithm, first generates multiple parameter combinations, then selects the combination with the highest fitness based on the loss function, performs crossover operations between the parameter combinations with high fitness, and introduces random mutations;. i new =α·θ1+(1-α)·θ2+variance term Among them: α is the cross coefficient, and the variation term is the random disturbance; Step S423: multi-objective optimization; the central control system (108) considers multiple objectives at the same time and constructs a multi-objective optimization model; the objective function is as follows: F(θ)=[J1(θ),J2(θ),…,J k (i)] Where: F(θ) is the objective function vector in multi-objective optimization, each J k (θ) represents an optimization objective.

9. The method for adaptively controlling ground settlement by controlling shield slag discharge volume and soil bin pressure according to claim 1 is characterized in that: In step S5, the specific steps of automatic warning and manual intervention are: Step S51: setting warning thresholds; setting safety thresholds for slag discharge volume, soil bin pressure, and ground settlement, and adjusting them when geological conditions change according to geological exploration reports to reflect the risk levels at different construction stages; Step S52: early warning triggering and operator response; when the central control system (108) detects that a certain parameter exceeds a preset threshold, the early warning information is transmitted to the operator's terminal device and a warning is issued at the same time; Step S53: Operator response and manual control; after receiving the warning signal, the operator selects a response measure according to the risk analysis report displayed by the central control system (108); Step S54: Recording and feedback: Record the process and results of manual intervention, and feed this information back to the central control system (108) for subsequent analysis and optimization of the central control system (108).

10. The method for adaptively controlling ground settlement by controlling shield slag discharge volume and soil bin pressure according to claim 1, characterized in that: In step S6, the specific steps of the control effect verification and construction adjustment are: Step S61: effect evaluation; the central control system (108) automatically generates a parameter change report, compares and analyzes key parameters such as the slag discharge volume, soil bin pressure, and ground settlement before and after the regulation, and determines whether the regulation measures are effective; focuses on evaluating the change trend of ground settlement, and through dynamic analysis of ground settlement monitoring data, the central control system (108) generates a settlement trend chart to evaluate the change of ground settlement and compare it with the expected target; Step S62: Optimize the excavation plan.

Images