A large-scale sinking well sinking posture active correction control system under complex geological conditions

Abstract

The present application belongs to the technical field of civil engineering, and provides a large-scale sinking well posture active correction control system under complex geological conditions, which comprises: a posture adjustment demand acquisition module: the deviation between the current posture of the sinking well and the target posture is acquired, and a direction adjustment demand is generated. By establishing a quantitative comparison mechanism of geological mutation time and system online correction lag time, dynamic risk assessment and hierarchical response of the correction window are realized. When the development speed of geological mutation is faster than the system response capability, the risk of missing the best correction window can be identified in advance, and hierarchical intervention measures from command fine tuning, accelerated execution to emergency stop and auxiliary anti-inclination are automatically triggered according to the low, medium and high risk levels, so that uncontrolled accidents such as sudden sinking and rolling are effectively avoided, the safety of the sinking well structure and construction is ensured, and the initiative and reliability of the large-scale sinking well posture control under complex geological conditions are improved. The present application breaks through the problem that the existing system can predict but cannot respond in time.

Description

Technical Field

[0001] This invention belongs to the field of civil engineering technology, specifically a large caisson sinking attitude active correction control system under complex geological conditions. Background Technology

[0002] As a key structure in major projects such as bridge anchorage foundations, shield tunneling shafts, and deep-water wharves, the accuracy and stability of attitude control during the sinking process of large caissons directly affect the quality and safety of the project. Existing caisson sinking control methods mainly rely on manual experience and conventional measurement methods, and passive or semi-active correction is carried out through methods such as zoned soil removal, external loading, and high-pressure water jetting. In recent years, some projects have introduced automated control systems based on sensor monitoring, and attempted to use proportional-integral-derivative (PID) control and logical judgment rules to achieve auxiliary adjustment of soil removal operations. However, these existing systems generally lack the ability to accurately model the coupled dynamics of caissons and soil under complex geological conditions, making it difficult to cope with control problems such as strong coupling of multiple variables, large time delays, and strong nonlinearity, and even more so, unable to predict and actively respond to geological changes in advance. However, when traversing complex geological conditions such as quicksand layers and weak interlayers, sudden geological changes often occur, such as sudden soil liquefaction and instantaneous loss of bearing capacity. This can cause the caisson to sink or roll over within seconds. At this time, due to the inherent online correction lag time (usually tens of seconds to several minutes) from the perception of geological parameters and control decisions to the action of the soil extraction mechanism, and the speed of geological changes often far exceeds this lag time, the control system cannot complete an effective response within the optimal correction window even if it has issued instructions in advance. It is predicted, but it is too late. The existence of this control blind zone directly leads to serious loss of control over the caisson's attitude, causing cracks in the well wall, damage to the cutting edge, permanent structural deformation, and even causing the surrounding surface to collapse and construction safety accidents, resulting in significant delays in the construction period and economic losses. Therefore, the present invention provides an active deviation correction control system for the sinking attitude of large caissons under complex geological conditions. Summary of the Invention

[0003] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0004] The technical solution adopted by this invention to solve its technical problem is: an active deviation correction control system for the sinking attitude of a large caisson under complex geological conditions, comprising: Attitude Adjustment Requirement Acquisition Module: Acquires the deviation between the current attitude of the caisson and the target attitude, and generates orientation adjustment requirements; Feedforward predictive control module: Based on the direction adjustment requirements and combined with the pre-built caisson-soil coupled dynamic model, it outputs the current optimal soil extraction command through predictive analysis. Geological mutation early warning module: During the execution of the current optimal soil extraction command, the geological parameters at the bottom and around the caisson are monitored in real time, and the geological mutation is predicted based on mutation characteristics. Lag Dynamic Assessment Module: If a geological mutation occurs, the module outputs the geological mutation time to the occurrence of the mutation, dynamically calculates the total lag time of online correction from perception to execution of the current control system, compares it with the geological mutation time, and outputs the risk level of missing the optimal correction window. Adaptive Decision Execution Module: Based on the output risk level, it triggers early warning signals and makes adaptive decision adjustments.

[0005] The beneficial effects of this invention are as follows: This invention establishes a quantitative comparison mechanism between the geological mutation time and the system's online correction lag time, enabling dynamic risk assessment and graded response to the correction window. When the geological mutation develops faster than the system's response capability, it can identify the risk of missing the optimal correction window in advance and automatically trigger graded intervention measures based on low, medium, and high risk levels, ranging from fine-tuning of instructions and accelerated execution to emergency soil cessation and auxiliary anti-tilting. This effectively avoids runaway accidents such as sudden sinking and rollover, ensuring the safety of the caisson structure and construction, and improving the initiative and reliability of attitude control for large caissons under complex geological conditions. It also overcomes the problem that existing systems can predict but cannot react in time. Attached Figure Description

[0006] The invention will now be further described with reference to the accompanying drawings.

[0007] Figure 1 This is a system block diagram of an embodiment of the present invention; Figure 2 This is a flowchart of the steps in an embodiment of the present invention. Detailed Implementation

[0008] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0009] Example Please see Figure 1 and Figure 2 As shown in the embodiment of the present invention, an active deviation correction control system for the sinking attitude of a large caisson under complex geological conditions includes: Attitude Adjustment Requirement Acquisition Module: Acquires the deviation between the current attitude of the caisson and the target attitude, and generates orientation adjustment requirements; In this module, the specific process is as follows: the current attitude data of the caisson is collected in real time by a high-precision sensor array deployed on the top and side walls of the caisson; The sensor array includes: At least three biaxial inclinometers are evenly distributed along the circumference of the caisson to measure the tilt angle of the caisson about the X-axis and Y-axis; At least three differential global positioning system or Beidou positioning terminals are evenly distributed around the top of the caisson to measure the three-dimensional spatial coordinates of each measuring point on the top of the caisson. And at least one gyroscope or fiber optic inertial measurement unit installed at the center of the caisson to measure the torsional angle of the caisson around the Z-axis; The collected multi-source sensor data is time-stamped and spatially registered to eliminate sampling delay differences and local installation errors between different sensors, thus forming the current attitude vector of the caisson. : ; in, , The current horizontal coordinates of the center point of the top surface of the caisson. The current sinking elevation of the center point of the top surface of the caisson. The current tilt angles around the X and Y axes. The current torsion angle around the Z-axis; Read the target attitude vector of the caisson at the current construction stage from the construction design documents. : ; in, , The target horizontal coordinates are the center point of the top surface of the caisson. The target sinking elevation is the center point of the top surface of the caisson. The target tilt angles around the X and Y axes. The target torsion angle around the Z-axis; The attitude deviation vector is obtained by subtracting the current attitude vector of the caisson from the target attitude vector component by component. An orientation adjustment demand vector is generated based on the attitude deviation vector. Each component of the orientation adjustment demand vector is the corresponding correction demand amount. Its magnitude is positively correlated with the component corresponding to the attitude deviation vector, and its sign indicates the desired direction of movement for correction. For example, if the tilt angle deviation around the X-axis in the attitude deviation vector is greater than zero, it indicates that the caisson is tilted in the positive X direction. In this case, the component corresponding to the direction adjustment demand vector is negative, which means that soil needs to be taken on the negative X-direction side to correct the tilt.

[0010] Feedforward predictive control module: Based on the direction adjustment requirements and combined with the pre-built caisson-soil coupled dynamic model, it outputs the current optimal soil extraction command through predictive analysis. In this module, the specific process is as follows: obtain the orientation adjustment demand vector, read the current actual attitude (position, tilt, torsion) of the caisson, the cumulative soil removal volume of each compartment, and real-time earth pressure monitoring data; The coupled dynamic model of the caisson and the soil is pre-established in the form of a control effect matrix. The rows of the matrix correspond to the six degrees of freedom of the caisson, and the columns correspond to each independent soil extraction chamber. Each element in the matrix represents the magnitude of attitude change of the caisson in a certain degree of freedom when a unit volume of soil is removed from the chamber. For example, one element represents "how many degrees the caisson will tilt around the X-axis if 1 cubic meter of soil is removed from compartment A"; another element represents "how many millimeters the caisson will sink as a whole if 1 cubic meter of soil is removed from compartment B". The control effect matrix is ​​obtained in the following way: In the early stage of caisson sinking, in shallow areas with relatively clear geological conditions, independent and quantitative soil sampling tests were carried out on each compartment in turn. Soil was taken from only one compartment at a time, and the volume of soil taken was recorded. At the same time, sensors were used to measure the attitude change of the caisson in six degrees of freedom caused by the result. The attitude change was divided by the volume of soil taken to obtain the influence coefficient of that compartment on that degree of freedom. The test results from multiple compartments are summarized, outliers caused by accidental disturbances are removed, and the initial control effect matrix is ​​obtained by least squares regression. For certain compartments that should theoretically have similarity coefficients due to structural symmetry (such as symmetrically distributed outer compartments), the average value is taken to enhance the robustness of the matrix. During the subsequent sinking process, after each complete zonal soil sampling cycle, the predicted attitude change of the model is compared with the actual monitored attitude change. The prediction error is calculated, and the recursive least squares algorithm is used to make small step corrections to the elements in the control effect matrix that are greatly affected by the current soil sampling action, so that the model gradually approaches the true response characteristics of the current stratum. The correction process is continuous to cope with the uneven geological changes along the depth. The model constructed in the above manner can reflect the response relationship between the caisson and the soil in the current construction stage, and whether it is a quantitative basis for subsequent optimization of soil extraction instructions. With the goal of minimizing the attitude deviation, a constrained numerical optimization algorithm is used to solve for the optimal soil extraction scheme. The process is as follows: the current attitude deviation is taken as the initial state. The amount of soil removed from each compartment is taken as a variable to be solved. Each variable has a corresponding upper and lower limit to prevent sudden subsidence caused by excessive soil removal in a single operation. Using the control effect matrix, the total attitude change caused by any set of soil-taking variables is calculated, which is the linear superposition of the effects of each compartment. The goal is to make the current attitude, after adding the predicted changes, as close as possible to the target attitude. The penalty for tilt angle deviation is the heaviest, followed by torsional deviation, then horizontal position deviation, while deviation in sinking is allowed to be within a certain range. The predicted attitude must not exceed the safety threshold (e.g., the maximum allowable tilt angle) in any degree of freedom, and the rate of change of soil removal volume in each compartment should not be too large in order to avoid drastic fluctuations in the actuators. Numerical optimization algorithms, such as constrained quadratic programming or gradient projection, are used to solve the problem. After iterative convergence, a set of optimal soil extraction allocation schemes for each compartment are obtained. The soil removal volume of each compartment obtained from the above solution is output as the current optimal soil removal command.

[0011] As one implementation method of this embodiment, the gradient projection method is used for solving, including: The actual amount of soil taken from each compartment in the previous control cycle is used as the starting point for the current iteration calculation. If it is the first calculation, the amount of soil taken from each compartment is evenly distributed to a base value, and two stopping conditions are set: one is to reach the preset maximum number of cycles (e.g., 100 times), and the other is that the target cost value calculated in two consecutive cycles changes very little (which can be considered as convergence). Under the current soil extraction plan, calculate whether slightly increasing the soil extraction volume of a single compartment will increase or decrease the target cost. The faster the increase, the more soil extraction should be reduced for that compartment; the faster the decrease, the more soil extraction should be increased for that compartment. Combining these trends of increase and decrease for all compartments forms a multi-dimensional optimization direction. The target cost consists of two parts: the first part is the comprehensive evaluation value of the caisson's attitude deviation, which is to combine the deviations in all directions according to the degree of engineering hazard into a total score, with tilt deviation having the heaviest deduction, followed by torsional deviation, then horizontal position deviation, and sinking depth deviation not being scored as long as it is within a certain range; the second part is the penalty for excessive changes in soil extraction volume compared to the previous cycle, that is, it is not desirable for the soil extraction volume of each compartment to fluctuate drastically. Considering that the amount of soil removed from each compartment cannot exceed the safety limit or be negative, and that the total amount of soil removed from all compartments must meet the minimum requirements for maintaining the sinking progress, it is necessary to project the optimization direction obtained in the previous step: For compartments that have reached their upper limit and whose optimization direction still needs to increase, their direction is forcibly changed to zero (no longer increasing); for compartments that have reached their lower limit and whose optimization direction still needs to decrease, their direction is forcibly changed to zero; if the current total soil extraction volume is close to or below the required minimum value, and the optimization direction attempts to further reduce the total volume, the direction component is corrected so that the total volume no longer decreases; after the above corrections, a feasible adjustment direction is obtained that both points to cost reduction and does not violate any hard constraints; In feasible adjustment directions, start from zero and gradually increase the adjustment range while observing the changes in target cost. Target cost usually decreases first and then increases. Through a simple trial and error method from large to small, for example, first try a large step, and if the cost increases instead, retreat and reduce the step size to find the adjustment range that reduces the target cost the most. Based on the found optimal adjustment range, move the current soil extraction volume of each compartment one step in the feasible direction to obtain a new soil extraction plan. Perform boundary checks on the new plan: clamp the compartments that exceed the upper limit to the upper limit, set the compartments that are below zero to zero, and if the total soil extraction volume is lower than the required minimum value, supplement it proportionally according to the remaining available space in each compartment. Compare the changes in target cost before and after the update. If the change is very small (indicating that it is close to the optimal solution) or the preset maximum number of iterations has been reached, stop the calculation and output the current soil extraction plan as the optimal soil extraction command; otherwise, return to the second step and continue iterative optimization based on the new plan. The final soil extraction allocation scheme for each compartment will be taken as the current optimal soil extraction command.

[0012] Geological mutation early warning module: During the execution of the current optimal soil extraction command, the geological parameters at the bottom and around the caisson are monitored in real time, and the geological mutation is predicted based on mutation characteristics. It should be noted that the value of obtaining the geological mutation time lies in transforming the geological risk from a qualitative judgment of whether a mutation will occur to a quantitative parameter of when the mutation will occur. This provides a comparable time benchmark for the online correction lag time of the control system itself. Without this time parameter, the lag time is merely an isolated performance indicator and cannot be quantitatively correlated with the urgency of external threats. With the geological mutation time, the control system can assess in real time whether its response capability can keep up with the development speed of the geological risk by comparing the two, and then output the risk level and guide the graded decision-making. In this module, the specific process is as follows: during the execution of the current optimal soil sampling command, the geological parameters of the soil at the bottom of the caisson and the surrounding soil are continuously collected in real time, including but not limited to: pore water pressure, soil pressure, sinking speed and acceleration, side wall friction, and groundwater level changes. The pore water pressure is obtained in real time by pore water pressure gauges installed at the bottom of the cutting foot and at the center of each compartment. Earth pressure is obtained through earth pressure sensors at the bottom of the inner partition wall and at the foot of the cutting edge. The sinking velocity and acceleration were obtained from the high-frequency settling rate provided by the inertial measurement unit at the top of the caisson; Sidewall friction is estimated using steel bars or strain gauges embedded in the well wall. Groundwater level changes are calculated from the water level data of observation wells around the caisson; Based on the system's built-in preset mutation feature library, it identifies whether geological mutations will occur; Among them, the mutation feature library contains various typical geological mutations (such as liquefaction of quicksand layers, shear failure of soft soil, brittle fracture of hard layers, etc.) that can be observed before they occur, and pre-stores samples of several typical precursor time series curves. For example, the precursor modes are: pore water pressure rises sharply to more than 80% of the effective overburden pressure within tens of seconds; the settlement acceleration fluctuates abnormally for several seconds and is unrelated to the soil removal operation; the soil pressure reading in a certain area suddenly drops by more than 30%, while the soil pressure in adjacent areas increases; the settlement speed suddenly accelerates and loses stability with the amount of soil removed remaining unchanged. The precursor time series curves are: the pore water pressure rise curve 30 seconds before a certain sand layer liquefaction occurs, and the abnormal waveform of the sinking acceleration 20 seconds before a certain sudden sinking. The real-time monitored geological parameters are compared with the mutation feature database. The process is as follows: Sliding window processing is performed on the real-time data stream of each sensor (the window duration is usually 10 to 30 seconds), and key feature values ​​within the sliding window are extracted, including but not limited to: rate of change, maximum change amplitude, temporal correlation with soil sampling command, and duration of abnormal fluctuations. The extracted feature values ​​are compared item by item with the preset thresholds and criteria of each precursor pattern in the mutation feature library, and the dynamic time warping (DTW) similarity matching method is used for identification. The DTW similarity is calculated between the current real-time acquired geological parameter sequence and the precursor time series curve of the same type of parameter in the feature library. Specifically, the DTW similarity calculation process is as follows: each data point of the geological parameter sequence curve and the precursor time series curve are paired, and the absolute difference between each pair of values ​​is calculated. All paired differences are arranged into a table, with the rows of the table corresponding to each point of the real-time sequence and the columns corresponding to each point of the template sequence. Starting from the top left corner of the table, the goal is to reach the bottom right corner. Each time, the movement can only be one grid to the right, one grid down, or one grid to the lower right. For each grid passed, the difference corresponding to that grid is accumulated. The goal is to find a path that minimizes the total accumulated difference from the starting point to the end point, which is the optimal path. The differences in all grids on the optimal path are added together to obtain the total accumulated difference (dynamic time warping distance). The total accumulated difference is divided by the number of steps traversed by the path to obtain the average difference per step, which is the normalized DTW similarity. Among them, the DTW algorithm allows two sequences to be non-linearly scaled and aligned on the time axis, thus it can match sequences with different occurrence rates but similar waveforms. For example, if the actual pore water pressure rises at twice the speed or half the speed of the sample, as long as the curve shape (such as an S-shape that is slow at first and then rapid) is similar, it can still be effectively identified. If the DTW similarity is less than the preset similarity threshold, the current monitored sequence is determined to be a successful match with the mutation precursor pattern; otherwise, it is determined to be a mismatch, and the monitoring continues in the next sliding window. The similarity threshold is statistically derived from historical samples: for example, the average DTW distance under normal fluctuations is 30, and the distance when similar to the precursor of mutation is 10, then the similarity threshold can be set to 15. If a successful match is determined, the probability of abrupt change is confirmed by combining other characteristics, such as the ratio of the absolute value of pore water pressure to the effective overburden pressure. Specifically: Obtain the historical database, which includes the specific values ​​of various monitoring characteristics within 30 seconds before each geological change (change) in multiple completed caisson projects with similar geological conditions to this project, as well as the same characteristic values ​​corresponding to a large number of normal construction periods (without change) during the same period. For cases where a precursor pattern of a mutation has been successfully matched within the current sliding window, the following features are extracted from the window: DTW similarity, ratio of pore water pressure to effective overburden pressure, percentage decrease in soil pressure, and duration of abnormal fluctuations in settlement acceleration. For each feature, the following statistics are performed in the historical database: the value distribution of the feature in historical mutation samples and the value distribution of the feature in historical normal samples are calculated respectively; Let A be the currently measured feature value. Count the number of historical mutation samples whose feature value is less than or equal to A, and divide it by the total number of mutation samples to obtain the cumulative mutation probability. At the same time, count the number of samples with feature values ​​less than or equal to A in the historical normal samples, divide the number of samples by the total number of normal samples, and obtain the normal cumulative probability. Divide the cumulative probability of mutation by the cumulative probability of normality to obtain the likelihood ratio corresponding to the feature. If the cumulative probability of normality is zero, the likelihood ratio is set to a preset maximum value, such as 1000. Repeat the above process for all selected independent features to obtain multiple likelihood ratios, and multiply the multiple likelihood ratios to obtain the comprehensive likelihood ratio; The overall mutation probability is calculated using the overall likelihood ratio. The formula for calculating the overall mutation probability is: Overall mutation probability = Overall likelihood ratio / (Overall likelihood ratio + 1). If the overall mutation probability is greater than the mutation probability threshold, then a geological mutation is determined to occur. If the overall mutation probability is less than or equal to the mutation probability threshold, it is determined that no geological mutation will occur. The mutation probability threshold is based on measured data from a large number of completed well projects in the historical database. It statistically analyzes the distribution range of the comprehensive probability before a geological mutation occurs and the distribution range of the comprehensive probability when no mutation occurs. The quantile value that minimizes the sum of the two types of errors (missed mutation and false alarm mutation) and prioritizes reducing the risk of missed detection is selected. At the same time, it is combined with the acceptable level of safety risk for this project (i.e., the maximum allowable missed detection rate). The comprehensive probability value corresponding to this quantile is used as the mutation probability threshold, which is usually a fixed value between 0.5 and 0.8 (such as 0.7). It can be fine-tuned during construction based on the accuracy feedback of the previous round of early warnings.

[0013] Lag Dynamic Assessment Module: If a geological mutation occurs, the module outputs the geological mutation time to the occurrence of the mutation, dynamically calculates the total lag time of online correction from perception to execution of the current control system, compares it with the geological mutation time, and outputs the risk level of missing the optimal correction window. In this module, the process of outputting the time until the geological mutation occurs is as follows: Based on the successfully matched precursor patterns of mutations, the type of mutation currently under warning is determined. The mutation types include: liquefaction of quicksand layer, shear failure of soft soil, and brittle fracture of hard layer. Each mutation type corresponds to a key monitoring parameter used for time calculation. For example, the mapping relationship is as follows: Liquefaction of quicksand layer: The key monitoring parameter is pore water pressure, measured in kilopascals, and the sensor is installed at the bottom of the cutting edge; Soft soil shear failure: The key monitoring parameter is the settlement acceleration, measured in meters per square second, and the sensor is an inertial measurement unit installed on the top of the caisson; Brittle fracture of hard layer: The key monitoring parameter is earth pressure, measured in kilopascals, and the sensor is installed at the bottom of the inner partition wall; Read all the measurement values ​​of key monitoring parameters in the past 30 seconds and arrange them from morning to night to form a real-time change sequence; Obtain the start time of the current sliding window, which is the moment when the precursor feature is first detected; Extract a subsequence of changes from the start time to the current time from the real-time change sequence, wherein the subsequence of changes is at least 10 seconds long; Take the last 5 seconds of data from the changed subsequence and use linear regression to calculate the rate of change of that data segment, i.e., the slope of the linear regression. If the rate of change is less than or equal to zero, it is determined that the current parameter has no upward or downward trend and cannot be extrapolated. The remaining time of the historical reference is directly used as the geological change time. If the rate of change is greater than zero, the danger threshold corresponding to the mutation type is read from the mutation feature library, the monitoring parameter value at the current moment (the last value of the real-time change sequence) is obtained, and the difference between the corresponding danger threshold and the monitoring parameter value is calculated to obtain the parameter difference. If the parameter difference is less than or equal to zero, it means that the danger threshold has been exceeded and the geological mutation time is 0 seconds. Otherwise, the ratio of the parameter difference to the rate of change is calculated to obtain the extrapolated remaining time. For example, the specific values ​​are as follows: Liquefaction of quicksand layer: The danger threshold is 100% of the effective overburden pressure, which is calculated based on the current burial depth of the caisson bottom and the soil unit weight, in kPa; Shear failure of soft soil: The danger threshold is when the sinking acceleration reaches 0.5 m / s² (empirical value, which can be pre-input); Brittle fracture of hard layer: The danger threshold is when the soil pressure drops to 50% of the initial value. The minimum value between the extrapolated remaining time and the historical reference remaining time is taken as the final geological change time. The historical reference remaining time is retrieved from the historical database. It is a historical event record that is the same as the current mutation type and has a successful dynamic time warping similarity match. For each successfully matched historical event, the time interval between the successful matching time of the sliding window and the actual occurrence time of the mutation is recorded. The median of all time intervals is calculated as the historical reference remaining time. If there is no matching record in the historical database, the historical reference remaining time is a preset default value. This default value is set based on similar engineering experience. For example, it is 20 seconds for quicksand liquefaction, 10 seconds for soft soil shear failure, and 15 seconds for hard layer brittle fracture. It is entered into the system before construction. The process of dynamically calculating the total online correction lag time of the current control system from sensing to execution is as follows: The total online correction lag time is obtained by summing three measurable components: perception lag time, decision lag time, and execution lag time. Among them, the high-precision time synchronization module records the acquisition time of each sensor data packet during acquisition, and the control system receiving program records the reception time when the data packet is received; The sensing lag time is the difference between the receiving time and the acquisition time. The maximum value among the sensing lag times of all sensors is recorded as the maximum sensing lag time. Record the start time when the gradient projection method iterative solution is executed, and record the end time when the current optimal soil sampling command is output and sent to the execution interface. The decision lag time is the difference between the end time and the start time, and is updated to the actual time of the most recent decision in each control cycle. Install motion sensors on each soil-collecting device (such as grab bucket and sludge suction valve), such as grab bucket closing limit switch and hydraulic system pressure switch. Record the time when the control system issues a soil-collecting command, and record the time when the motion sensor detects that the device has started to move. The execution lag time is the difference between the start time of the action and the time when the command is issued. The maximum execution lag time among all currently running soil extraction equipment is recorded as the maximum execution lag time. The maximum perception lag time, decision lag time, and maximum execution lag time are summed to obtain the total online correction lag time. The process of outputting the risk level of missing the optimal correction window is as follows: The time margin is obtained by calculating the difference between the geological abrupt change time and the total lag time of online correction. If the time margin is greater than the safety upper limit threshold, then low risk is output, indicating that the control system has enough time to complete the correction action and will not miss the best correction window. If the time margin is greater than zero but less than or equal to the safety upper limit threshold, then the output is medium risk, indicating that the correction window is about to close and that it is necessary to speed up the execution or take auxiliary measures. If the time margin is less than or equal to zero, a high risk is output, indicating that the best correction window has been missed or is about to be missed, and there is a risk of attitude loss of control. The safety upper limit threshold is based on the time distribution from early warning to sudden change under similar geological conditions in historical engineering data. Combined with the measured statistical distribution of the total lag time of online correction of this control system, the critical time value that enables the control system to complete effective correction before the sudden change occurs is taken. Specifically, by analyzing the time margin of successful correction cases and the time margin of failed cases in completed caisson projects, the dividing point that can distinguish between the two situations of "having enough time" and "not having enough time" is selected. At the same time, the shortest action time of the actuator from the issuance of the instruction to the actual soil extraction and the minimum evolution time of geological parameters from abnormality to sudden change are considered to comprehensively determine the threshold. It is preset before construction based on the experience of similar projects and can be adjusted according to the actual early warning accuracy during construction.

[0014] Adaptive decision execution module: Based on the output risk level, it triggers early warning signals and makes adaptive decision adjustments; In this module, the specific process is as follows: based on the received risk level, different levels of early warning signals are triggered, as detailed below: Low risk level: A blue warning signal is triggered, which is displayed on the control room screen in a low-frequency flashing manner. At the same time, a text reminder is sent to the mobile terminal of the construction management personnel, stating that the risk of geological change is low, the current corrective capability is sufficient, adjust the current optimal soil sampling instruction, and continue to execute according to the original instruction. Medium risk level: A yellow warning signal is triggered, which will be displayed on the control room screen in the form of medium-frequency flashing, while an intermittent buzzing sound will be emitted, and a text reminder will be sent to the construction management personnel, stating that there is a risk of geological change and the correction window will be closed soon. Please pay attention and prepare emergency measures. The upper limit of soil removal volume for each compartment in the current optimal soil removal command is temporarily reduced to 70% of the original upper limit. At the same time, the weight of the attitude deviation comprehensive evaluation value in the target cost in the gradient projection method iterative solution is increased by 20%, prioritizing attitude stability and allowing a reduction in sinking rate. The adjusted command is immediately issued for execution. High risk level: Triggering a red warning signal, the warning will be displayed on the control room screen with high-frequency flashing, accompanied by a continuous buzzing sound, and an emergency voice call will be sent to construction management personnel with the text message: "High risk of geological change, the best window for correction has been missed, and the emergency plan must be implemented immediately." Immediately cease all current soil extraction operations and trigger the emergency mechanism: issue a hard interruption command to all chambers inside the well to stop soil extraction, and issue an emergency auxiliary command to the external auxiliary systems (such as high-pressure water jetting and external loading equipment), requiring the initiation of anti-tilting measures according to the preset emergency plan (such as loading on the opposite side of the tilt). The above emergency commands have the highest priority and cannot be interrupted by other tasks.

[0015] This embodiment has at least the following effects: The geological mutation early warning module transforms geological risks from a qualitative judgment of whether they will occur to a quantitative time parameter of when they will occur, enabling the control system to proactively predict mutations before they occur, rather than simply correcting deviations that have already occurred. The lag dynamic assessment module calculates the total lag time of online correction from perception to execution in real time, and compares it quantitatively with the geological change time to form a safety indicator of time margin, thereby clearly answering the question of whether there is enough time. Based on the time margin, three levels of risk (low, medium, and high) are output, and corresponding early warning signals and soil sampling instructions are triggered to achieve risk classification and control, avoiding false alarms or omissions. In the event of sudden geological changes such as liquefaction of quicksand layers and shear failure of soft soil, the prediction often fails because it is too late. By estimating the time of geological changes in advance and dynamically comparing it with the system lag time, dangerous scenarios that cannot be responded to in time even if the prediction is correct can be identified. Emergency mechanisms can be triggered in high-risk situations, reducing serious consequences such as well wall cracking, structural damage, and safety accidents caused by sudden sinking and rolling. The core function of this embodiment is to compare the development speed of external threats (geological change time) with the response speed of internal systems (total online correction lag time) and make hierarchical decisions accordingly, thus optimizing the blind spot problem of not being able to assess whether it can keep up with the change.

[0016] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An active deviation correction control system for the sinking attitude of a large caisson under complex geological conditions, characterized in that: include: Attitude Adjustment Requirement Acquisition Module: Acquires the deviation between the current attitude of the caisson and the target attitude, and generates orientation adjustment requirements; Feedforward predictive control module: Based on the direction adjustment requirements and combined with the pre-built caisson-soil coupled dynamic model, it outputs the current optimal soil extraction command through predictive analysis. Geological mutation early warning module: During the execution of the current optimal soil extraction command, the geological parameters at the bottom and around the caisson are monitored in real time, and the geological mutation is predicted based on mutation characteristics. Lag Dynamic Assessment Module: If a geological mutation occurs, the module outputs the geological mutation time to the occurrence of the mutation, dynamically calculates the total lag time of online correction from perception to execution of the current control system, compares it with the geological mutation time, and outputs the risk level of missing the optimal correction window. The process for obtaining the total online correction lag time is as follows: The high-precision time synchronization module records the acquisition time of each sensor data packet during acquisition, and the control system receiving program records the reception time when the data packet is received; the sensing lag time is the difference between the reception time and the acquisition time, and the maximum value among the sensing lag times of all sensors is recorded as the maximum sensing lag time. The start time is recorded when the gradient projection method iterative solution begins, and the end time is recorded when the current optimal soil sampling command is output and sent to the execution interface; the decision lag time is the difference between the end time and the start time, and is updated to the actual time of the most recent decision in each control cycle; Calculate the maximum execution lag time, sum the maximum perception lag time, decision lag time, and maximum execution lag time to obtain the total online corrected lag time; Adaptive Decision Execution Module: Based on the output risk level, it triggers early warning signals and makes adaptive decision adjustments.

2. The active deviation correction control system for the sinking attitude of a large caisson under complex geological conditions according to claim 1, characterized in that: The process of obtaining the generation direction adjustment requirement is as follows: Real-time acquisition of the current attitude data of the caisson forms the current attitude vector of the caisson. The target attitude vector of the caisson at the current construction stage is read. The current attitude vector of the caisson and the target attitude vector are subtracted component by component to obtain the attitude deviation vector. The direction adjustment requirement vector is generated based on the attitude deviation vector.

3. The active deviation correction control system for the sinking attitude of a large caisson under complex geological conditions according to claim 1, characterized in that: The process of outputting the current optimal soil extraction command is as follows: The pre-built caisson-soil coupled dynamic model is invoked. The model adopts the form of a control effect matrix. The matrix elements represent the influence coefficients of the unit soil volume of each compartment on the attitude changes of the six degrees of freedom of the caisson. The matrix is ​​obtained by taking soil samples from a single compartment during the initial stage of caisson sinking and then correcting it through least squares regression and online recursion. With the goal of minimizing attitude deviation, the amount of soil taken from each compartment is taken as the variable to be solved. The upper and lower limits of the amount of soil taken from each compartment and the lower limit of the total amount of soil taken are set as constraints. The total attitude change caused by soil taking is predicted by the control effect matrix. The constrained numerical optimization algorithm is used to solve the problem. After iterative convergence, the soil taking allocation scheme of each compartment is obtained as the current optimal soil taking command output.

4. The active deviation correction control system for the sinking attitude of a large caisson under complex geological conditions according to claim 1, characterized in that: The process of predicting whether a geological mutation will occur based on mutation feature identification is as follows: Geological parameters of the bottom and surrounding soil of the caisson are collected in real time to obtain the built-in mutation feature library, which stores a variety of typical geological mutation precursor patterns and corresponding precursor time series curve samples. Sliding window processing is performed on the real-time data stream of each sensor to extract key feature values ​​within the window. The dynamic time warping algorithm is used to match the shape similarity between the currently acquired geological parameter sequence and the precursor time series curves in the feature library. If the DTW similarity is less than the preset similarity threshold, it is determined that the mutation precursor pattern has been successfully matched. By combining multiple independent features within a successfully matched window, the likelihood ratio of each feature is calculated and multiplied to obtain the overall likelihood ratio. The overall mutation probability is then calculated using the overall likelihood ratio. If the overall mutation probability is greater than the preset mutation probability threshold, it is determined that a geological mutation will occur.

5. The active deviation correction control system for the sinking attitude of a large caisson under complex geological conditions according to claim 4, characterized in that: The calculation process for DTW similarity is as follows: Pair each data point of the geological parameter sequence curve with the precursor time series curve, calculate the absolute difference between each pair of values, and arrange all paired differences into a table. The rows of the table correspond to the points of the real-time sequence, and the columns correspond to the points of the template sequence. Starting from the top left corner of the table, the goal is to reach the bottom right corner. Each time, the movement can only be one grid to the right, one grid down, or one grid to the lower right. For each grid passed, the difference corresponding to that grid is accumulated. The goal is to find a path that minimizes the total accumulated difference from the starting point to the ending point, which is the optimal path. The differences in all grids along the optimal path are added together to obtain the total accumulated difference. The total accumulated difference is divided by the number of steps traversed by the path to obtain the average difference per step, which is the normalized DTW similarity.

6. The active deviation correction control system for the sinking attitude of a large caisson under complex geological conditions according to claim 4, characterized in that: The likelihood ratio calculation process for each feature is as follows: Calculate the value distribution of each feature in historical mutation samples and the value distribution in historical normal samples respectively; Let A be the currently measured feature value. Count the number of historical mutation samples whose feature value is less than or equal to A, and divide it by the total number of mutation samples to obtain the cumulative mutation probability. At the same time, count the number of samples with feature values ​​less than or equal to A in the historical normal samples, divide the number of samples by the total number of normal samples, and obtain the normal cumulative probability. The likelihood ratio is obtained by dividing the cumulative probability of mutation by the cumulative probability of normality.

7. The active deviation correction control system for the sinking attitude of a large caisson under complex geological conditions according to claim 1, characterized in that: The process of obtaining the time of geological abrupt change is as follows: The mutation type currently under warning is determined based on the successfully matched mutation precursor pattern, and each mutation type corresponds to a key monitoring parameter used for time calculation. Read all the measurement values ​​of key monitoring parameters in the past 30 seconds, arrange them from morning to night to form a real-time change sequence, obtain the start time of the current sliding window, extract the change subsequence from the start time to the current time from the real-time change sequence, take the last 5 seconds of data segment in the change subsequence, and use the linear regression method to calculate the change rate of the data segment. If the rate of change is less than or equal to zero, then the remaining historical reference time is directly used as the geological abrupt change time. If the rate of change is greater than zero, the danger threshold corresponding to the mutation type is read from the mutation feature library, the monitoring parameter value at the current moment is obtained, the difference between the corresponding danger threshold and the monitoring parameter value is calculated, and the parameter difference is obtained. If the parameter difference is less than or equal to zero, the geological mutation time is 0 seconds. Otherwise, the ratio of the parameter difference to the rate of change is calculated to obtain the extrapolated remaining time. The minimum value between the extrapolated remaining time and the historical reference remaining time is taken as the final geological mutation time.

8. The active deviation correction control system for the sinking attitude of a large caisson under complex geological conditions according to claim 1, characterized in that: The calculation process for the maximum execution lag time is as follows: A motion sensor is installed on each soil sampling device. When the control system issues a soil sampling command, the time of the command issuance is recorded. When the motion sensor detects that the device has started to move, the time of the start of the action is recorded. The execution lag time is the difference between the time of the start of the action and the time of the command issuance. The maximum execution lag time among all currently running soil sampling devices is recorded as the maximum execution lag time.

9. The active deviation correction control system for the sinking attitude of a large caisson under complex geological conditions according to claim 1, characterized in that: The process of outputting the risk level of missing the optimal correction window is as follows: The time margin is obtained by calculating the difference between the geological abrupt change time and the total lag time of online correction. If the time margin is greater than the safety upper limit threshold, then output "low risk". If the time margin is greater than zero but less than or equal to the safety upper limit threshold, then output medium risk; If the time margin is less than or equal to zero, output "high risk".

Images