Tunnel rich water sand layer tunnel hole intelligent control grouting control method

Abstract

The application discloses a kind of tunnel water-rich sand layer tunnel hole intelligent control grouting control methods, it is related to tunnel engineering technical field, the method includes: through sensor network synchronous acquisition grouting, drilling tool and formation response parameter and carry out check;Utilize self-adapting filter to clean data, output situation assessment report in combination with expert rule base;According to the assessment result, switch between master-slave cooperation, safety priority and global optimization mode, wherein, master-slave cooperation mode is according to real-time formation hardness index dynamically adjusts grouting flow proportionality coefficient;Safety priority mode executes valve limiting amplitude, pump control switching and drilling tool pause linkage instruction and carries out ladder type recovery when identifying risk;Global optimization mode utilizes pressure mutual correlation coefficient to avoid slurry channeling.The application also realizes parameter closed-loop optimization by cloud edge cooperation, solves the problem of water-rich sand layer grouting control precision deficiency and drilling and injection cooperation difference, realizes self-adapting accurate control to complex geological working condition.

Description

Technical Field

[0001] This invention relates to the field of tunnel engineering technology, specifically to a control method for intelligent grouting in tunnels with water-rich sand layers. Background Technology

[0002] In tunnel construction in water-rich sandy geological environments, grouting reinforcement is a crucial process for improving soil properties, blocking groundwater channels, and ensuring excavation safety. Current grouting operations typically employ advanced small-diameter guide pipes or deep-hole grouting techniques, relying heavily on PLC control systems or manual experience. In conventional control logic, the system monitors the operational status by collecting pressure and flow values ​​from the grouting pump outlet, as well as the drilling rig's operating parameters. It then adjusts the grouting speed and pressure based on pre-set constant thresholds or a simple proportional-integral-derivative (PID) algorithm to complete grout injection and soil reinforcement during drilling.

[0003] However, water-rich sand layers exhibit strong heterogeneity and rheological properties, and even minor changes in geological conditions can trigger significant random noise and transient fluctuations at the sensor acquisition end. Existing control methods typically lack in-depth processing of the spatiotemporal consistency of multi-source data and adaptive suppression mechanisms for non-steady-state noise. When faced with feedback from complex formations, they struggle to accurately distinguish between genuine sudden changes in operating conditions (such as sudden pressure increases or formation hardening) and transient signal interference (such as sensor jitter or pipeline pulsation). This lack of signal recognition capability leads the control system to easily misinterpret disturbed fluctuation signals as valid state changes, resulting in frequent jumps in control commands or erroneous mode switching. Consequently, it becomes impossible to maintain the continuity of the grouting process and the accuracy of situation assessment under strong interference environments. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a control method for intelligent grouting in tunnels with water-rich sand layers, which solves the problems of insufficient control precision, poor drilling-grouting coordination, and lack of adaptability to complex geological conditions during the grouting process in water-rich sand layers.

[0005] To achieve the above objectives, the present invention provides a control method for intelligent grouting in tunnels with water-rich sand layers, comprising the following steps:

[0006] Step S100: Multi-source information synchronous sensing and acquisition step; collect grouting process parameters, drilling tool operation parameters and formation response parameters through sensor network, and use hardware clock signal to stamp the collected grouting process parameters, drilling tool operation parameters and formation response parameters with a unified timestamp, and verify the validity of the timestamped data;

[0007] Step S200: Multi-level information fusion and real-time situation assessment step; The verified grouting process parameters, drilling tool operation parameters and formation response parameters are aligned in time and space. The aligned data is cleaned based on a preset adaptive filtering algorithm, key feature parameters are calculated, and a situation assessment report is output in combination with a preset expert rule base and fuzzy logic. The situation assessment report includes qualitative diagnostic conclusions and comprehensive risk level.

[0008] Step S300: Intelligent collaborative control decision-making and instruction generation step; Based on the results of the situation assessment report, determine the current control mode from the preset master-slave collaborative mode, security-priority collaborative mode, and global optimization collaborative mode, and generate a set of control instructions;

[0009] Step S400: Command distribution and actuator coordinated execution step; The control command set is distributed to the drill string driver, the grouting pump driver, and the regulating valve driver to drive the drill string, the grouting pump, and the regulating valve to coordinate their actions, and the execution deviation is monitored in real time;

[0010] Step S500: Closed-loop feedback and system optimization step; Upload all process data generated during the execution of steps S100 to S400 to the cloud platform, receive the parameter combination and expert rule base optimized based on the full process data from the cloud platform, and apply them to steps S200 and S300.

[0011] Preferably, in step S200, cleaning the aligned data based on a preset adaptive filtering algorithm specifically includes: monitoring the observation residual sequence; when the weighted norm of the observation residual exceeds a preset threshold, it is determined that a data mutation has occurred; the adaptive filtering algorithm automatically reduces the weight of the measurement noise covariance matrix and increases the weight of the process noise covariance matrix.

[0012] In step S200, before outputting the situation assessment report, situation timing locking logic is executed: when a new operating condition is identified, the characteristic signal of the new operating condition is required to exist continuously for a period of time exceeding the confirmation time window before the situation assessment report is updated and the control mode is allowed to switch. The control strategy of the previous moment remains unchanged within the confirmation time window.

[0013] Preferably, in step S300, when the result of the situation assessment report is normal drilling or formation change, the current control mode is determined to be the master-slave collaborative mode. The control logic of the master-slave collaborative mode is: using the drill bit feed rate as the main command signal, the target grouting flow rate is calculated through a linear relationship, and the linear relationship includes a dynamic adaptive proportional coefficient and a base flow rate.

[0014] The dynamic adaptive proportional coefficient is dynamically adjusted online based on the formation diagnosis results: First, the real-time formation hardness index is calculated based on the ratio of drill bit torque to drill string feed speed. Then, the dynamic adaptive proportional coefficient is updated based on the deviation between the real-time formation hardness index and the regional historical average hardness. When the real-time formation hardness index increases, the dynamic adaptive proportional coefficient is increased proportionally to increase the grouting volume per unit depth.

[0015] Furthermore, the update logic of the dynamic adaptive proportional coefficient follows the following relationship: based on the benchmark proportional coefficient, it is adjusted according to the adjustment gain factor and the rate of change of the real-time formation hardness index relative to the regional historical average hardness; when the real-time formation hardness index is higher than the regional historical average hardness, the dynamic adaptive proportional coefficient is increased based on the benchmark proportional coefficient.

[0016] Preferably, in step S300, when the situation assessment report indicates a high risk of splitting or an excessive disturbance, the current control mode is determined to be the safety-priority collaborative mode, which has the highest priority and executes the following collaborative instruction sequence:

[0017] First sequence: Send a rapid closing command to the actuator of the regulating valve to achieve hard limiting;

[0018] Second sequence: Send a command to the driver of the grouting pump to forcibly switch the operating mode of the grouting pump from flow control to pressure control, and set the target pressure to a safe value to reduce the pump speed;

[0019] The third sequence: issue a command to the drill bit's driver to pause feed or slightly retract.

[0020] After executing the cooperative instruction sequence of the security-priority cooperative mode, a tentative recovery phase is entered, which specifically includes:

[0021] Pressure maintenance monitoring: Maintain a preset pressure holding time at the target pressure. If the grouting flow rate remains zero, it is determined to be a complete blockage and the machine will be shut down and an alarm will be triggered. If the grouting flow rate gradually recovers, it is determined to be a clearing of the channel.

[0022] Stepwise pressurization: After determining that the channel is clear, the output frequency of the grouting pump is gradually increased in a preset step size, while monitoring the response of the soil pressure.

[0023] Mode switching: If the dynamic pressure gradient remains below the safety threshold and the duration exceeds the stable window period during the step-by-step pressurization process, the control mode will be switched from pressure control back to flow control, and the master-slave cooperative mode will be reactivated.

[0024] Preferably, in step S300, when multiple grouting points are controlled to perform cluster operations, the current control mode is determined to be the global optimization and coordination mode, which includes a conflict avoidance strategy: determining whether the grout path is short-circuited by analyzing the data correlation of the soil pressure sensors at each grouting point.

[0025] Calculate the cross-correlation coefficient between grouting pressure and soil pressure at different grouting points within the sliding time window. If the cross-correlation coefficient is greater than the short-circuit determination threshold and there is a time delay within a preset range, it is determined that there is a grout flow channel between the two grouting points, and an instruction is generated to pause the grouting at one of the grouting points.

[0026] Specifically, the calculation logic of the cross-correlation coefficient is as follows: calculate the covariance between the grouting pressure at the first grouting point and the soil pressure at the second grouting point, calculate the product of the standard deviations of the grouting pressure and the soil pressure, and divide the covariance by the product of the standard deviations to obtain the cross-correlation coefficient.

[0027] Preferably, step S400 includes dead zone compensation logic during the collaborative execution process: the command value in the control command set is compared with the actual value fed back by the actuator in real time. If the actual value does not change after a small action command is issued, it is determined that the actuator is in a mechanical dead zone, and a compensation bias is automatically added to the control command set. If the deviation between the command value and the actual value continues to exceed the safety tolerance within a set time, an emergency stop logic is triggered to cut off the power supply to the main oil circuit hydraulic pump.

[0028] Preferably, in step S500, receiving the parameter combination and expert rule base optimized based on the full process data mining from the cloud platform specifically includes: before formally applying the optimized parameter combination and expert rule base, enabling shadow mode verification at the edge; the shadow mode verification is configured to: run the algorithm model containing the optimized weights in parallel in the background without outputting control commands, compare the difference between the output of the algorithm model and the current actual command, and automatically switch the algorithm model to the active state after the verification is passed.

[0029] Preferably, in step S100, the collected grouting process parameters include grouting pressure and grouting flow rate; the collected drilling tool operating parameters include drilling tool feed rate, drilling depth and drill bit torque; and the collected formation response parameters include soil pressure and surface settlement value. In step S200, the constructed unified state vector contains the real-time values ​​of all the above-mentioned collected parameters.

[0030] This invention provides a method for intelligent grouting control inside tunnels with water-rich sand layers. It has the following beneficial effects:

[0031] 1. This invention uses a multi-source information synchronous sensing and multi-level fusion processing mechanism to perform spatiotemporal alignment of the collected parameters. It uses an adaptive filtering algorithm based on the observation residual sequence to automatically adjust the weight of the noise covariance matrix to retain the true mutation characteristics. Combined with the situation timing lock logic, it forces the new working condition signal to continuously meet the confirmation time window requirements, thereby effectively shielding random noise interference and transient fluctuations, avoiding frequent oscillations in the control system caused by signal misjudgment, and realizing the robustness of grouting situation assessment and the stability of the control process in complex environments.

[0032] 2. This invention constructs a master-slave collaborative control mode based on real-time formation hardness feedback. The drill bit feed rate is used as the main command signal. The formation hardness index is calculated using the ratio of drill bit torque to feed rate. Based on the deviation of this index from the historical average, the ratio coefficient of grouting flow rate to drilling speed is dynamically adjusted online. In hard rock areas, the grouting volume is automatically increased to fill the fractures, while in soft soil areas, the flow rate is limited to prevent grout leakage. This achieves real-time dynamic matching between the grouting strategy and the geological environment, effectively ensuring the density and uniformity of grout filling in each section of the water-rich sand layer.

[0033] 3. By configuring a safety-first collaborative mode and its supporting tentative recovery logic, this invention automatically triggers a series of linked commands, including rapid limiting of the regulating valve, pressure control of the grouting pump switching, and suspension of drill bit feed, when the risk of splitting or disturbance exceeding the limit is identified, in order to block the risk. After the pressure stabilizes, it executes an automatic reset process including flow monitoring and step-by-step pressure increase, thereby realizing rapid response to sudden high-risk working conditions and smooth repair without human intervention. This reduces the loss of construction efficiency caused by abnormal downtime while preventing damage to the formation structure. Attached Figure Description

[0034] Figure 1 This is a flowchart illustrating the overall process of the intelligent control grouting method of the present invention.

[0035] Figure 2 This is a detailed flowchart of the multi-level information fusion and real-time situation assessment steps of the present invention;

[0036] Figure 3 This is a logical block diagram of the intelligent collaborative control decision-making and instruction generation steps of the present invention. Detailed Implementation

[0037] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] This invention provides a control method for intelligent grouting in tunnels with water-rich sand layers. This method deeply integrates modern sensor technology, edge computing, information fusion theory, adaptive control algorithms and cloud big data analysis. Through a closed-loop process of perception, fusion, decision-making, execution and optimization, it achieves intelligent collaborative control of the drilling and grouting processes.

[0039] This method is implemented by program logic deployed in edge computing devices (including industrial ARM core boards or industrial control computers) and interacts with the cloud platform via the Internet of Things, such as... Figure 1 As shown, the overall process includes the following steps:

[0040] Step S100: Multi-source information synchronous sensing and acquisition step;

[0041] This step is the starting point for intelligent control, aiming to provide the system with high-quality, unbiased raw data. Its key lies in synchronization and comprehensiveness, and it specifically includes the following sub-steps:

[0042] S110. Data source definition and acquisition:

[0043] The system automatically and continuously collects the following three types of physical quantities through a pre-set sensor network:

[0044] Grouting process parameters:

[0045] Grouting pressure ( ): Measured by a high-frequency response pressure sensor installed near the drill pipe inlet, it directly reflects the force of the slurry on the formation.

[0046] Grouting flow rate ( ): Measured by an electromagnetic flowmeter or mass flowmeter for accurate measurement of slurry output.

[0047] Drill string operating parameters:

[0048] Drill feed rate ( ) and drilling depth ( ): The signal is directly measured by a high-precision magnetostrictive displacement sensor built into the hydraulic cylinder of the driving drill bit, or obtained through differential calculation, so as to achieve direct and accurate measurement.

[0049] Drill bit torque ( ): This is measured by a torque sensor mounted on the power head to sense the formation's resistance to drilling.

[0050] Formation response parameters:

[0051] Earth pressure ( ): This is measured by earth pressure cells embedded in the soil in the grouting influence zone, and is used to directly monitor the changes in soil stress caused by grouting.

[0052] Surface subsidence value ( ): Data is obtained from automated monitoring systems (including hydrostatic levels and total stations) via interfaces and is used to assess the final effect and environmental impact of grouting.

[0053] S120, High-precision time synchronization:

[0054] All sensor data is stamped with a uniform timestamp with an accuracy of no less than 1 millisecond by a hardware clock signal when it is collected at the acquisition node or transmitted to the central controller. This step eliminates the problem of data timing disorder caused by communication delay and different sampling periods, and provides timing consistency guarantee for multi-source data fusion.

[0055] S130, Data Preprocessing and Caching:

[0056] The collected raw data undergoes preliminary format conversion and caching to form a multi-source data stream with timestamps, which is then transmitted to the data processing center in real time.

[0057] S140, Data Validation:

[0058] Before the data is transmitted to the processing center, the system performs integrity and validity checks on the cached data stream.

[0059] Extreme value verification: Set physical limit thresholds for each physical quantity. If the collected value exceeds the range, mark it as an anomaly and remove it.

[0060] Trend verification: The rate of change of data is calculated based on a sliding time window. If the rate of change of data at a certain moment exceeds the maximum response speed allowed by the physical system (e.g., pressure instantly returns to zero or jumps), it is determined to be a sensor malfunction or signal loss. The system automatically calls up the valid value from the previous moment for interpolation and fills in the gap, and at the same time reports a sensor malfunction alarm.

[0061] Step S200: Multi-level information fusion and real-time situation assessment steps;

[0062] This step is responsible for intelligent decision support, transforming raw data into actionable situational information. The processing is as follows: Figure 2 As shown, it is divided into three levels:

[0063] S210, Data-level Fusion:

[0064] Time alignment: The received multiple data streams are aligned to a unified timeline according to their timestamps, ensuring the accuracy of the data used for analysis. , , , The data belongs to the same moment.

[0065] Spatial alignment: aligning grouting parameters ( , (This is related to the borehole location, forming a depth) At this location, the pressure is Traffic is Spatiotemporal related data packets.

[0066] Data cleaning and filtering: An adaptive Kalman filter algorithm is used to smooth the signal. This algorithm is used to suppress on-site noise while preserving the true abrupt changes in the signal (such as sudden increases in pressure), providing input for risk identification.

[0067] Specifically, the adaptive Kalman filter algorithm adjusts the measurement noise covariance matrix in real time by monitoring the innovation sequence.

[0068] When the weighted norm of the observed residual exceeds a preset threshold, the system is determined to have experienced a sudden change (such as a sudden increase in pressure). At this time, the algorithm automatically reduces the weight of the measurement noise covariance matrix and increases the weight of the process noise covariance matrix, thereby improving the filter's sensitivity to tracking sudden change signals, avoiding misjudging real pressure sudden changes as noise for smoothing, and ensuring that splitting risk signals are not filtered out.

[0069] S220, Feature-level Fusion

[0070] Constructing a unified state vector: Combining the cleaned data into a feature vector describing the instantaneous state of the system.

[0071]

[0072] Real-time calculation of key feature parameters:

[0073] Drilling efficiency ratio ( ): Reflects the amount of grout injected per unit depth. By comparing it with the theoretical value or historical average, it is determined whether the filling is full.

[0074] Dynamic pressure gradient ( ): Calculate the instantaneous rate of change of pressure as a precursor indicator for early warning of splitting risk.

[0075] Formation impedance characteristics: Comprehensive analysis of torque and pressure Cooperative change patterns, for example, when Significantly increased and When rising slowly, it indicates a hard rock layer pattern; when Decrease Rapid ascents indicate weak layers or precursory splitting patterns.

[0076] Disturbance propagation factor: association and The relationship between the changes was investigated to assess the transmission and cumulative effects of grouting disturbances.

[0077] S230, Decision-level Integration

[0078] Expert rule base reasoning: Input feature parameters into a pre-built generative rule base constructed based on domain expert knowledge for matching and reasoning.

[0079] Rule Example 1: IF AND THEN condition = high risk of splitting, confidence level = 0.95.

[0080] Rule Example 2: IF consistently higher AND consistently below THEN Condition = Insufficient grouting in hard rock layer, confidence level = 0.80.

[0081] Rule Example 3: IF Continue to increase AND The condition "THEN" indicates that the disturbance exceeds the limit and the warning is in effect. The confidence level is 0.85.

[0082] Fuzzy logic processing: For working conditions with ambiguous boundaries, a fuzzy logic system is used to calculate the confidence level of different working condition attributes through membership functions to handle the uncertainty of information.

[0083] Situational timing lock: To avoid frequent high-frequency switching (oscillation) of control modes between S310 and S320 due to local heterogeneity of the strata, the system introduces timing lock logic. When the system identifies a new working condition (such as transitioning from soft soil to hard rock), it must require that the characteristic signal exist continuously for more than the confirmation time window (e.g., 3 seconds) before officially updating the situation assessment report and allowing the switching of control modes. Within the confirmation time window, the system maintains the control strategy of the previous moment unchanged, thereby filtering transient interference and ensuring the smoothness of the control process.

[0084] Situational output: Outputs a structured situational assessment report, including qualitative diagnostic conclusions (such as normal drilling, hard rock formations, and fracturing risk) and a comprehensive risk level (such as low, medium, and high) and its quantitative confidence level. This report provides a basis for control decisions.

[0085] Step S300: Intelligent collaborative control decision-making and instruction generation steps;

[0086] This step receives a situation report and formulates a control plan accordingly. This step includes three control modes that can be dynamically and seamlessly switched, such as... Figure 3 As shown:

[0087] S310, Master-Slave Collaborative Mode (Applicable to normal and optimized drilling conditions)

[0088] Triggering condition: The situation assessment result is normal drilling or formation change.

[0089] Core algorithm: using formula To take control.

[0090] in, Target grouting flow rate; Real-time drilling speed (master signal); This is a dynamic adaptive scaling factor; It serves as the base flow rate for lubrication and pressure maintenance.

[0091] Innovative Adaptive Mechanism: Coefficient This is a non-fixed value; the system dynamically adjusts it online based on the formation diagnosis results output in step S230. The value is automatically increased if the identification result is hard rock. Increase the grouting volume per unit depth to ensure compaction; if the soil identification result indicates soft soil, appropriately reduce the grouting volume. To prevent slurry waste.

[0092] Output command: Generates a flow control command for the grouting pump frequency converter, with a target value of .

[0093] coefficient The dynamic adjustment adopts a method based on the formation hardness index. The mapping function is implemented, and the system is first based on torque. With drilling speed Calculate the real-time formation hardness index Then, update according to the following logic value:

[0094]

[0095] in, As the benchmark scaling factor, This represents the historical average hardness of the region. To adjust the gain factor, this formula ensures that when the stratum hardens ( When (increases), The value automatically increases proportionally, thereby increasing the amount of grout injected per unit depth to fill any tiny cracks that may exist in the hard rock layer.

[0096] S320, Safety-First Collaborative Mode (Suitable for high-risk operating conditions)

[0097] Triggering conditions: The situation assessment result is a high-confidence risk such as high splitting risk or disturbance exceeding the limit.

[0098] Override control and coordination mechanism: This mode, once activated, has the highest priority and overrides other modes. The system generates a coordination instruction packet containing three sequences:

[0099] First sequence (millisecond response): Sends a rapid closing command to the electric pressure / flow control valve to achieve hard limiting.

[0100] The second sequence (second-level response): Simultaneously, a command is sent to the grouting pump frequency converter to forcibly switch its operating mode from flow control to pressure control, and the target pressure is set to a safe value. This allows the pump speed to decrease smoothly.

[0101] The third sequence (linkage control): issues a command to the drill string control system to pause feed or slightly retract, reducing disturbance at the source.

[0102] Effect: Enables rapid suppression and smooth transition of anomalies, protecting the formation structure.

[0103] Reset and probing mechanisms:

[0104] After executing the above three sequences of coordinated instructions, the system enters a tentative recovery phase. In this phase, instead of directly switching back to the S310 master-slave coordinated mode, the following logic is executed:

[0105] Pressure maintenance monitoring: At target pressure Maintain the preset pressure holding time (e.g., 30 seconds), and continuously monitor the grouting flow rate during this period. If the flow rate remains at zero, it is determined to be a complete blockage, and the system will remain shut down and trigger an alarm; if the flow rate gradually recovers, it is determined to be a clearing of the channel.

[0106] Step-by-step pressurization: After determining that the channel is clear, the system gradually increases the output frequency of the grouting pump in preset small steps (Step), while monitoring the soil pressure. The response.

[0107] Mode switchback: If during the pressurization process, the pressure dynamic gradient... If the temperature remains below the safety threshold and the duration exceeds the stable window period, the system will smoothly switch the control mode from pressure control back to flow control and reactivate the S310 master-slave collaborative mode, thereby achieving automatic closed-loop repair from fault conditions to normal conditions.

[0108] S330, Global Optimized Collaborative Mode (Suitable for multi-grouting point cluster operations)

[0109] Triggering condition: The system controls multiple grouting points (hydraulic tube insertion machine).

[0110] Decision-making logic:

[0111] Real-time monitoring of the status of all grouting points ( ).

[0112] Load balancing algorithm: If it is found that point A has low pressure and high flow (easy to inject), while point B has high pressure and low flow (difficult to inject), then an instruction is generated to reduce the pumping frequency of point B, while maintaining or increasing the pumping frequency of point A, so as to maximize the overall grouting efficiency of the cluster.

[0113] Conflict avoidance strategy: By analyzing the correlation of data from earth pressure sensors at each point, it is determined whether the grout path is short-circuited. If grouting at point A causes a sharp increase in earth pressure at point B, an instruction is generated to suspend grouting at either point A or point B to avoid ineffective grout loss and ensure the effectiveness of the filling range.

[0114] Analyzing data correlation specifically refers to: calculating the grouting pressure at point A. Earth pressure at point B Cross-correlation coefficients within the sliding time window The calculation formula is:

[0115]

[0116] in, For covariance, Set the short-circuit detection threshold to the standard deviation. (For example, 0.8). If the calculated value is... > If there is a time delay within the preset range, it is determined that there is a slurry flow channel (short circuit) between point A and point B, and an avoidance command is immediately triggered.

[0117] Step S400: Instruction issuance and executor coordination steps;

[0118] This step is responsible for translating the decision into physical actions.

[0119] S410, Command Distribution:

[0120] The control command set generated by the control decision module is sent to the corresponding actuator driver separately and simultaneously through a highly reliable industrial real-time network (including EtherCAT or Profinet).

[0121] S420, Collaborative Execution

[0122] Drilling tool system: The hydraulic servo drive controls the opening of the servo valve according to the received speed / displacement command, and drives the hydraulic cylinder to advance or retract at the set speed.

[0123] Grouting system: The frequency converter adjusts the speed of the grouting pump motor according to the received flow or pressure command.

[0124] Control valve: Receives on / off or opening commands and executes rapid actions.

[0125] All actuators work together under a unified clock schedule to complete the coordinated actions planned in step S300.

[0126] Deviation monitoring and compensation:

[0127] During collaborative execution, edge computing devices compare control command values ​​with actual values ​​fed back by sensors in real time (e.g., comparing commanded flow rate with flow meter readings, commanded valve position with actual valve opening).

[0128] Dead zone compensation:

[0129] If the system detects that the feedback value does not change after issuing a minor motion command, it determines that the actuator is in the mechanical dead zone. The system automatically adds a compensation bias to the output command to counteract the effects of mechanical backlash or static friction.

[0130] Abnormal Interruption: If the deviation between the command value and the feedback value continues to exceed the safety tolerance within the set time (e.g., the valve is stuck and cannot be closed), the system will immediately trigger the highest level of emergency stop logic, cut off the power supply to the main oil circuit hydraulic pump, and physically terminate the operation to prevent grouting accidents caused by actuator malfunction.

[0131] Step S500: Closed-loop feedback and system optimization steps;

[0132] This step involves closed loops at two spatiotemporal scales to enable continuous improvement of the system.

[0133] S510, Real-time Control Closed Loop (seconds / milliseconds):

[0134] The actuator's actions alter the grouting process (generating new pressure, flow, and depth data). These changes are captured in real time by the sensing system in step S100 and used as a new round of data input. This feedback mechanism forms a real-time control closed loop, enabling the system to continuously adjust to changes in operating conditions and ensure process stability and accuracy.

[0135] S520, Long-term Optimization Closed Loop (Daily / Weekly Level):

[0136] Data Upload: The edge controller uploads all data from the entire construction process (including state vectors, feature parameters, control commands, and situation reports) to the cloud platform.

[0137] Cloud-based big data analytics: Machine learning engines on cloud platforms (including deep learning networks and reinforcement learning algorithms) mine massive amounts of successful case data to find optimal solutions under different geological conditions. The parameters are combined, and the thresholds and logic in the expert rule base are optimized.

[0138] Strategy Update: The cloud platform will distribute the optimized algorithm model or parameter set to all field devices on the edge side.

[0139] The strategy update specifically includes the distribution of two types of data:

[0140] Parameter set update: Issue the updated control parameter table, including the threshold values ​​from step S230. , and the reference coefficient in step S310 , .

[0141] Shadow Mode Verification: Before issuing new algorithm model weights, the Shadow Mode is enabled on the edge device. The new model is computed in parallel in the background but does not output control commands. The system compares the output of the new model with the current actual commands. After the consistency or superiority verification is passed, the system automatically switches to the active state.

[0142] Improved performance: The optimized system is applied in subsequent construction to achieve continuous performance improvement on a larger temporal and spatial scale.

[0143] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A control method for intelligent grouting in water-rich sandy tunnels, characterized in that, Includes the following steps: Step S100: Multi-source information synchronous sensing and acquisition step; collect grouting process parameters, drilling tool operation parameters and formation response parameters through sensor network, and use hardware clock signal to stamp the collected grouting process parameters, drilling tool operation parameters and formation response parameters with a unified timestamp, and verify the validity of the timestamped data; Step S200: Multi-level information fusion and real-time situation assessment step; The grouting process parameters, drilling tool operation parameters and formation response parameters verified by S100 are aligned in time and space. The aligned data is cleaned based on a preset adaptive filtering algorithm, key feature parameters are calculated, and a situation assessment report is output in combination with a preset expert rule base and fuzzy logic. The situation assessment report includes qualitative diagnostic conclusions and comprehensive risk level. Step S300: Intelligent collaborative control decision-making and instruction generation step; Based on the results of the situation assessment report, determine the current control mode from the preset master-slave collaborative mode, security-priority collaborative mode, and global optimization collaborative mode, and generate a set of control instructions; Step S400: Instruction distribution and executor collaborative execution steps; The control command set is distributed to the drill string driver, the grouting pump driver, and the regulating valve driver to drive the drill string, the grouting pump, and the regulating valve to work together, and to monitor the execution deviation in real time. Step S500: Closed-loop feedback and system optimization step; Upload all process data generated during the execution of steps S100 to S400 to the cloud platform, receive the parameter combination and expert rule base optimized based on the full process data from the cloud platform, and apply them to steps S200 and S300.

2. The intelligent grouting control method for tunnels with water-rich sand layers according to claim 1, characterized in that, In step S100, the collected grouting process parameters include grouting pressure and grouting flow rate; the collected drilling tool operation parameters include drilling tool feed rate, drilling depth and drill bit torque; and the collected formation response parameters include soil pressure and surface settlement. In step S200, the constructed unified state vector contains the real-time values ​​of all the above-mentioned collected parameters.

3. The control method for intelligent grouting in a water-rich sandy layer tunnel according to claim 1, characterized in that, In step S200, cleaning and aligning the data based on a preset adaptive filtering algorithm specifically includes: The monitoring observation residual sequence is used. When the weighted norm of the observation residual exceeds a preset threshold, a data mutation is determined. The adaptive filtering algorithm automatically reduces the weight of the measurement noise covariance matrix and increases the weight of the process noise covariance matrix. In step S200, before outputting the situation assessment report, situation timing locking logic is executed: when a new operating condition is identified, the characteristic signal of the new operating condition is required to exist continuously for a period of time exceeding the confirmation time window before the situation assessment report is updated and the control mode is allowed to switch. The control strategy of the previous moment remains unchanged within the confirmation time window.

4. The intelligent grouting control method for tunnels with water-rich sand layers according to claim 1, characterized in that, In step S300, when the situation assessment report indicates normal drilling or formation change, the current control mode is determined to be the master-slave collaborative mode, and the control logic of the master-slave collaborative mode is as follows: The drill bit feed rate is used as the main command signal, and the target grouting flow rate is calculated through a linear relationship, which includes a dynamic adaptive proportional coefficient and a base flow rate. The dynamic adaptive proportional coefficient is dynamically adjusted online based on the formation diagnosis results: First, the real-time formation hardness index is calculated based on the ratio of drill bit torque to drill string feed speed. Then, the dynamic adaptive proportional coefficient is updated based on the deviation between the real-time formation hardness index and the regional historical average hardness. When the real-time formation hardness index increases, the dynamic adaptive proportional coefficient is increased proportionally to increase the grouting volume per unit depth.

5. The intelligent grouting control method for tunnels in water-rich sand layers according to claim 1, characterized in that, In step S300, when the situation assessment report indicates a high risk of splitting or an excessive disturbance, the current control mode is determined to be the safety-priority cooperative mode. This safety-priority cooperative mode has the highest priority and executes the following cooperative instruction sequence: First sequence: Send a rapid closing command to the actuator of the regulating valve to achieve hard limiting; Second sequence: Send a command to the driver of the grouting pump to forcibly switch the operating mode of the grouting pump from flow control to pressure control, and set the target pressure to a safe value to reduce the pump speed; The third sequence: issue a command to the drill bit's driver to pause feed or slightly retract.

6. The intelligent grouting control method for tunnels in water-rich sand layers according to claim 5, characterized in that, After executing the cooperative instruction sequence of the security-priority cooperative mode, a tentative recovery phase is entered, which specifically includes: Pressure maintenance monitoring: Maintain a preset pressure holding time at the target pressure. If the grouting flow rate remains zero, it is determined to be a complete blockage and the machine will be shut down and an alarm will be triggered. If the grouting flow rate gradually recovers, it is determined to be a clearing of the channel. Stepwise pressurization: After determining that the channel is clear, the output frequency of the grouting pump is gradually increased in a preset step size, while monitoring the response of the soil pressure. Mode switching: If the dynamic pressure gradient remains below the safety threshold and the duration exceeds the stable window period during the step-by-step pressurization process, the control mode will be switched from pressure control back to flow control, and the master-slave cooperative mode will be reactivated.

7. The intelligent grouting control method for tunnels with water-rich sand layers according to claim 1, characterized in that, In step S300, when multiple grouting points are controlled to perform cluster operations, the determined current control mode is the global optimization and coordination mode, which includes a conflict avoidance strategy: By analyzing the correlation of data from soil pressure sensors at each grouting point, it can be determined whether the grout path is short-circuited. Calculate the cross-correlation coefficient between grouting pressure and soil pressure at different grouting points within the sliding time window. If the cross-correlation coefficient is greater than the short-circuit determination threshold and there is a time delay within a preset range, it is determined that there is a grout flow channel between the two grouting points, and an instruction is generated to pause the grouting at one of the grouting points.

8. The intelligent grouting control method for tunnels with water-rich sand layers according to claim 1, characterized in that, In step S400, dead-zone compensation logic is included during the collaborative execution process: The instruction values ​​in the control instruction set are compared with the actual values ​​fed back by the actuator in real time. If it is detected that the actual value does not change after issuing a small action instruction, it is determined that the actuator is in the mechanical dead zone, and a compensation bias is automatically added to the control instruction set. If the deviation between the command value and the actual value continues to exceed the safety tolerance within the set time, the emergency stop logic is triggered, and the power supply to the main oil circuit hydraulic pump is cut off.

9. The control method for intelligent grouting in a water-rich sandy layer tunnel according to claim 1, characterized in that, In step S500, receiving the parameter combination and expert rule base optimized based on the full-process data mining from the cloud platform specifically includes: Before officially applying the optimized parameter combination and expert rule base, first enable shadow mode verification at the edge. The shadow mode verification is configured to: run an algorithm model containing optimized weights in parallel in the background without outputting control commands, compare the difference between the output of the algorithm model and the current actual command, and automatically switch the algorithm model to the active state after the verification is passed.

Images