An intelligent construction method and system for an offshore super-large anti-collision box cofferdam

Abstract

The present application relates to the field of foundation construction technology, and discloses a method and system for intelligent construction of super-large anti-collision box cofferdams at sea, the method comprising: real-time collection and fusion of multi-source heterogeneous sensor data, optimizing and generating control instructions based on the current state and target trajectory, and continuously executing the control instructions until the position and attitude deviations meet the preset thresholds and are stably maintained, for monitoring the health status of the components and dynamically reconstructing the control strategy in the event of failure. The system comprises: an embedded autonomous hardcover system, which is integrated into the anti-collision box body and consists of the following modules: a multi-source perception module, an onboard collaborative AI decision module, and an active attitude control module, whose actuators include omnidirectional thrusters and a dynamic ballast water system. The present invention achieves high-precision, anti-disturbance intelligent lowering and positioning of super-large components at sea through a high-speed and robust perception-decision-execution closed loop, thereby improving the efficiency and safety of offshore construction.

Description

Technical Field

[0001] The present invention relates to the technical field of foundation construction, and in particular to an intelligent construction method and system for an offshore super-large anti-collision box cofferdam. Background Art

[0002] Ultra-large anti-collision box cofferdams, due to their complex structure, enormous dimensions, and heavy weight, present significant challenges in manufacturing, transporting, and precisely lowering them in offshore engineering. Traditional cofferdam construction methods suffer from inherent drawbacks such as long construction periods, high overall costs, and difficulty ensuring accurate positioning. Especially in complex and volatile sea conditions, traditional manual operations and experience-based control methods often lag behind actual requirements, making it difficult to achieve efficient, safe, and precise operation of ultra-large components.

[0003] While existing technical solutions have introduced methods such as using steel casings as guide piles and lifting and connecting sections with floating cranes during the construction and lowering of anti-collision box cofferdams, their level of intelligence is relatively low. This results in significant room for improvement in construction speed, construction safety, and final installation quality. In particular, current technology is insufficiently resistant to external disturbances such as complex ocean currents and waves, making it difficult to achieve accurate and rapid closed-loop control. This undoubtedly increases construction risks and limits operational efficiency, resulting in a failure to fully meet the demands for digital, intelligent, and high-end construction of ultra-large anti-collision box cofferdams.

[0004] Therefore, the present invention proposes an intelligent construction method and system for an offshore super-large anti-collision box cofferdam to address the deficiencies of the prior art. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides an intelligent construction method for super-large offshore anti-collision box cofferdams, which solves the problem that the super-large offshore anti-collision box cofferdams mainly rely on external large-scale equipment and manual monitoring and decision-making during the lowering and placement process, resulting in low control accuracy, limited operation window and high safety risks.

[0006] To achieve the above objectives, the present invention provides an intelligent construction method for an ultra-large offshore anti-collision box cofferdam. By integrating an embedded autonomous fine-fitting system inside the cofferdam body, autonomous, continuous, and high-precision closed-loop control of the cofferdam's posture and position is achieved.

[0007] A first aspect of the present invention provides a method for intelligently constructing an offshore ultra-large anti-collision box cofferdam, the method comprising the following steps:

[0008] S1. Integrate an embedded autonomous assembly system in the anti-collision box cofferdam body, the embedded autonomous assembly system including an active attitude control module, a multi-source perception module, and an onboard collaborative AI decision-making module;

[0009] S2. After the anti-collision box cofferdam is hoisted into the water, the embedded autonomous hardcover system is powered on and initialized, and the multi-source sensing module is driven to collect real-time status data of the anti-collision box cofferdam, wherein the real-time status data includes structural stress data;

[0010] S3. The onboard collaborative AI decision module performs collaborative optimization calculations based on a model predictive control algorithm to generate control instructions;

[0011] S4, the onboard collaborative AI decision module performs online adaptive correction on the parameters of the hydrodynamic model by continuously comparing the predicted output of the hydrodynamic model built into the embedded autonomous hardcover system with the real-time status data;

[0012] S5. The active attitude control module receives and executes the control instruction, and performs continuous closed-loop adjustment on the attitude and plane position of the anti-collision box cofferdam until the anti-collision box cofferdam is stabilized within a preset accuracy range of the target installation position.

[0013] In one embodiment, the specific configuration of the embedded autonomous hardcover system in step S1 is defined as follows:

[0014] The active attitude control module consists of a dynamic ballast system and multiple azimuthing thrusters. The dynamic ballast system generates attitude restoring torques by transferring water mass at high speed between tanks within the crash box cofferdam, which are used to adjust pitch and roll attitude. The azimuthing thrusters are arranged along the perimeter of the crash box cofferdam, providing horizontal translational control forces and rotational control torques to adjust its planar position and heading angle.

[0015] The multi-source perception module includes: a satellite positioning system for obtaining the global coordinates of the anti-collision box cofferdam; an inertial measurement unit for measuring its attitude angular velocity and attitude angle; and multiple stress sensors arranged at key nodes of the structure for obtaining real-time structural stress data.

[0016] The onboard collaborative AI decision-making module is an embedded computing unit that internally stores and runs a hydrodynamic model, a model predictive control algorithm module, and an online adaptive correction algorithm module.

[0017] In one embodiment, the data collection and transmission process in step S2 is defined as follows:

[0018] The power-on initialization triggers the satellite positioning system, inertial measurement unit, and all stress sensors within the multi-source perception module to enter a continuous operating state. The multi-source perception module integrates the collected position data, attitude data, and structural stress data into a unified data frame format and transmits it to the onboard collaborative AI decision-making module in real time at a preset frequency via the internal data bus of the embedded autonomous hardcover system.

[0019] In one embodiment, the collaborative optimization calculation in step S3 is described in detail:

[0020] The core of the model predictive control algorithm is to generate the optimal control instructions for each control cycle by solving a finite-horizon optimal control problem. This optimal control problem is structured as solving an objective function consisting of a weighted trajectory tracking term and a control increment penalty term.

[0021] Its mathematical expression can be described as:

[0022] ;

[0023] Where, is the discrete time point of the current control cycle; Based on The impact of momentary state on future The predicted value of the system output (i.e. the position and posture of the anti-collision box cofferdam) at the first step; For the future The expected output on the reference trajectory of the step; For the future The control increment of the step; To predict the time domain length, To control the time domain length; and are the weight matrices used to adjust the trajectory tracking error and control the increment size, respectively.

[0024] The first term is the trajectory tracking term, which is used to minimize the cumulative deviation between the predicted output sequence and the reference trajectory sequence.

[0025] The second term is the control increment penalty term, which is used to suppress drastic changes in control instructions, thereby ensuring smooth operation of the active attitude control module.

[0026] While solving the above objective function, the optimization calculation must also meet a series of constraints:

[0027] Structural stress constraints: ,in For the predicted The stress value of each stress measuring point, The upper limit of safety stress preset for this measuring point.

[0028] Energy consumption constraints: ,in is the predicted instantaneous total power consumption of the active attitude control module, A safe power limit is set for the autonomous energy module in the embedded autonomous system. This constraint ensures that the generated control instructions are physically executable.

[0029] In one embodiment, the online adaptive correction process in step S4 is defined as follows:

[0030] The onboard collaborative AI decision-making module uses a recursive least squares algorithm, takes the error between the predicted output of the hydrodynamic model and the real-time status data sent by the multi-source perception module as the algorithm input, and updates the key parameters in the hydrodynamic model online and iteratively, specifically including the fluid damping coefficient and the additional mass coefficient, so that the model can more accurately reflect the actual hydrodynamic characteristics of the anti-collision box cofferdam under the current sea conditions, thereby improving the prediction accuracy of the model predictive control algorithm.

[0031] In one embodiment, the basis for determining whether the precise positioning is completed in step S5 is defined as follows:

[0032] The onboard collaborative AI decision-making module compares the real-time position and attitude data received from the multi-source perception module with the target installation position and target attitude to determine the planar position deviation and attitude inclination deviation. If the values ​​of the planar position deviation and attitude inclination deviation remain below their respective preset tolerance thresholds for a continuous period of time, the system is considered to have achieved precise positioning.

[0033] In one embodiment, the method further includes step S6: performing online diagnosis and fault-tolerant control on the embedded autonomous hardcover system.

[0034] The onboard collaborative AI decision-making module continuously diagnoses the operating status of each thruster and pump valve in the active attitude control module, as well as each sensor in the multi-source perception module. If any component failure is detected, the onboard collaborative AI decision-making module dynamically adjusts the system model it relies on for prediction to reflect only the status of currently available, fully functional components. It then continues collaborative optimization calculations based on the adjusted system model to generate fault-tolerant control instructions.

[0035] In one embodiment, the method is performed while a large external floating crane provides vertical suspension force to the anti-collision box cofferdam. The control instructions generated by the onboard collaborative AI decision-making module are specifically used to autonomously compensate for horizontal and attitude disturbances caused by environmental factors such as waves and currents acting on the anti-collision box cofferdam during the lowering of the anti-collision box cofferdam.

[0036] A second aspect of the present invention provides an intelligent construction system for an offshore super-large anti-collision box cofferdam. The system is an embedded autonomous fine-fitting system integrated into the anti-collision box cofferdam body. The embedded autonomous fine-fitting system includes:

[0037] A multi-source perception module is used to collect real-time status data of the anti-collision box cofferdam at a preset frequency, wherein the real-time status data includes global position, posture, and structural stress data of multiple key nodes;

[0038] An active attitude control module, fixedly connected to the anti-collision box cofferdam, for receiving control instructions and applying control force and control torque to the anti-collision box cofferdam;

[0039] The onboard collaborative AI decision module is an embedded computing unit, whose data input is connected to the output of the multi-source perception module and whose data output is connected to the input of the active attitude control module. The onboard collaborative AI decision module is configured to perform the following operations:

[0040] Based on a model predictive control algorithm, with the goal of minimizing the deviation between the predicted motion state of the anti-collision box cofferdam and the target installation position, and simultaneously using the structural stress data received from the multi-source perception module as a safety constraint, a collaborative optimization calculation is performed to generate control instructions;

[0041] Performing online adaptive correction on the parameters of the hydrodynamic model by continuously comparing the predicted output of the hydrodynamic model built into the system with the real-time status data received from the multi-source sensing module;

[0042] The control instruction is sent to the active attitude control module to drive the active attitude control module to perform continuous closed-loop adjustment on the attitude and plane position of the anti-collision box cofferdam until the anti-collision box cofferdam is stabilized within a preset accuracy range of the target installation position.

[0043] Beneficial effects:

[0044] 1. This invention integrates an embedded autonomous assembly system within the anti-collision box cofferdam, forming a complete internal closed loop of perception, decision-making, and execution, achieving autonomous control of the cofferdam's posture and position. This method utilizes a model predictive control algorithm to calculate precise control instructions in real time, actively compensating for environmental disturbances and transforming passive manual adjustments into active autonomous corrections. This improves the final placement accuracy of ultra-large offshore anti-collision box cofferdams, enabling highly automated intelligent construction and reducing reliance on operator experience and external auxiliary equipment.

[0045] 2. This invention incorporates the structural safety of the cofferdam as a core constraint in the control algorithm, significantly improving the safety of the construction process. When collaboratively optimizing and calculating control instructions, the system uses real-time monitored structural stress data as a hard safety boundary, ensuring that any generated control actions will not cause stress on key parts of the cofferdam to exceed preset safety thresholds. This predictive safety protection mechanism can fundamentally avoid hazardous conditions and safeguard the structural integrity of the anti-collision box cofferdam throughout the entire intelligent construction process.

[0046] 3. This invention enhances the system's adaptability to complex marine environments by introducing an online adaptive correction mechanism. By continuously comparing the hydrodynamic model's predictions with the actual motion state, the system updates the model parameters online, accurately reflecting the actual hydrodynamic characteristics of the current sea conditions. This ensures that the control model maintains high accuracy, ensuring stable and effective control of the intelligent construction system under variable sea conditions and broadening the meteorological window for offshore operations.

[0047] 4. This invention significantly improves the reliability and robustness of the entire system by incorporating online diagnostic and fault-tolerant control capabilities. The system continuously self-diagnoses the operating status of key components, such as sensors and thrusters. If a component failure is detected, it dynamically adjusts its internal control model to continue the mission relying solely on the remaining functioning components. This design prevents unintended interruptions to the entire intelligent construction process of the offshore ultra-large anti-collision box cofferdam due to localized failures, ensuring operational continuity and safety even under extreme operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is the overall process flow chart of the present invention;

[0049] Figure 2 This is a schematic diagram of the large floating crane of the present invention being anchored in place;

[0050] Figure 3 This is an elevation view of the wind cable arrangement for auxiliary positioning of the present invention;

[0051] Figure 4 This is a plan view of the wind cable arrangement for auxiliary positioning of the present invention;

[0052] Figure 5 This is a block diagram of the intelligent construction system for the offshore super-large anti-collision box cofferdam of the present invention;

[0053] Figure 6 This is a flow chart of the intelligent construction method of the super-large anti-collision box cofferdam at sea according to the present invention;

[0054] Figure 7 This is a schematic diagram of the model predictive control principle of the present invention.

[0055] Among them, 1. Floating crane anchor; 2. Giant anti-collision box; 3. Pile foundation construction platform; 4. Construction plank road; 5. Large floating crane; 6. Embedded autonomous precision assembly system; 7. Multi-source perception module; 8. Active attitude control module; 9. Onboard collaborative AI decision-making module. DETAILED DESCRIPTION

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0057] Embodiments of the present invention provide a method and system for intelligently constructing an ultra-large offshore crash box cofferdam, addressing the existing challenges of low precision, high safety risks, and reliance on manual experience in cofferdam lowering operations. An embedded autonomous assembly system 6 is pre-integrated within the main structure of the giant crash box 2. This system autonomously and continuously controls the planar position and posture of the giant crash box 2 in a closed-loop manner, while a large external floating crane 5 provides the primary vertical lifting force. This system actively compensates for environmental disturbances such as waves and currents, enabling highly accurate, unmanned placement.

[0058] Reference Figure 5 , Figure 5 This is a block diagram of the components of an intelligent construction system for an ultra-large offshore anti-collision cofferdam, according to one embodiment of the present invention. The system is an embedded autonomous assembly system 6 integrated within the giant anti-collision cofferdam 2, comprising a multi-source perception module 7, an active attitude control module 8, and an onboard collaborative AI decision-making module 9.

[0059] The multi-source sensing module 7 is used to collect the state data of the giant anti-collision box 2 and the surrounding environment data in real time. In a specific embodiment, the multi-source sensing module 7 includes:

[0060] A satellite positioning system is used to obtain the global three-dimensional coordinates of the giant crash box 2; an inertial measurement unit is used to measure its three-axis angular velocity and attitude angle; multiple stress sensors are deployed at key nodes of the cofferdam structure; and multiple tilt sensors are used to precisely measure its tilt attitude. These sensors are physically installed at predetermined locations on the giant crash box 2 to comprehensively obtain its kinematic and dynamic information.

[0061] Active attitude control module 8 is used to receive and execute control commands from the onboard collaborative AI decision-making module 9, applying active control forces and torques to the giant anti-collision system 2. In one embodiment, active attitude control module 8 comprises a dynamic ballast water system and multiple azimuth thrusters. The dynamic ballast water system generates restoring torques to adjust pitch and roll attitude by transferring water mass between tanks within the cofferdam at high speed. The azimuth thrusters are evenly distributed along the perimeter of the cofferdam to provide horizontal translational control forces and bow rotation control torques.

[0062] The onboard collaborative AI decision-making module 9 is the data processing and decision-making core of the entire embedded autonomous assembly system 6. Its hardware is an industrial-grade embedded computing unit. Its data input is connected to the multi-source perception module 7 via an internal data bus to receive real-time status data. Its data output is connected to the active attitude control module 8 via the same bus to issue control commands. The onboard collaborative AI decision-making module 9 internally incorporates a hydrodynamic model, a model predictive control algorithm module, and an online adaptive correction algorithm module.

[0063] Reference Figure 2 , which shows the external collaboration scene when the system of the present invention is in operation. The large floating crane 5 completes the anchoring in place near the pile foundation construction platform 3 through its floating crane anchor 1 system, providing a stable benchmark for subsequent lifting operations. The embedded autonomous finishing system 6 provided by the present invention plays a role in the process of the large floating crane 5 lowering the giant anti-collision box 2. The two work together to complete the intelligent construction task. At the same time, you can refer to Figure 3 and Figure 4 The wind cable arrangement shown is used to assist in mooring the giant anti-collision casing 2 to increase operational safety in severe sea conditions.

[0064] The specific steps of the intelligent construction method of the offshore super-large anti-collision box cofferdam provided by the present invention will be described in detail below.

[0065] Reference Figure 6 , Figure 6 This is a flow chart of a method for intelligently constructing an offshore ultra-large anti-collision box cofferdam according to one embodiment of the present invention. The method comprises the following steps:

[0066] Step S1: System Integration:

[0067] This step is completed during the onshore manufacturing phase before the giant crash-proof box 2 leaves the factory. Its purpose is to physically and permanently integrate the various hardware components of the embedded autonomous hardcover system 6 into the main structure of the giant crash-proof box 2. Figure 1 The overall process shown is carried out in coordination, and the specific implementation method includes the following:

[0068] During the cofferdam block fabrication and in-plant segment assembly phase, component installation and wiring are performed according to pre-designed layout drawings. First, prefabricated mounting bases and brackets are installed for each subunit of the multi-source perception module 7, active attitude control module 8, and onboard collaborative AI decision-making module 9. These are then secured to the cofferdam's steel structure via welding or high-strength bolts.

[0069] Specifically, the installation of the multi-source sensing module 7 includes placing and fixing multiple stress sensors in the stress concentration area of ​​the cofferdam structure determined by finite element analysis; installing the inertial measurement unit at a location close to the geometric center of the cofferdam with less vibration interference; and installing the antenna of the satellite positioning system in an unobstructed area on the top of the cofferdam to ensure the quality of signal reception.

[0070] The installation of the active attitude control module 8 includes symmetrically installing multiple omnidirectional thrusters below the waterline of the outer wall of the cofferdam to provide effective translation and rotation control force; and arranging the water pumps, valves and connecting pipes of the dynamic ballast water system between the preset water tanks inside the cofferdam.

[0071] The installation of the onboard collaborative AI decision-making module 9 includes encapsulating the embedded computing unit as the hardware core, together with its power supply module and communication interface, in a chassis with waterproof, salt spray and impact resistance functions, and fixing the chassis in a cabin inside the cofferdam that is easy to maintain and has a relatively stable environment.

[0072] After all components are installed, the system's internal electrical connections are made. Through pre-installed, shielded, and waterproof dedicated cables and data buses, all sensors in the multi-source perception module 7 and all actuators in the active attitude control module 8 are connected to the data interfaces of the onboard collaborative AI decision-making module 9, forming a complete, closed hardware system. After integration is complete, a power-on test is performed to verify the functional integrity of each component and the reliability of the communication links.

[0073] Step S2: System activation and data collection:

[0074] This step is performed after the giant anti-collision casing 2 is hoisted into the water by the large floating crane 5. The operator sends a power-on initialization instruction to the embedded autonomous finishing system 6 via a wireless communication link, or the system automatically executes the initialization program after detecting that the main power is turned on.

[0075] System power-up initialization triggers all sensor subunits within the multi-source sensing module 7 to enter a continuous operating state. Specifically, the satellite positioning system begins receiving satellite signals and calculating the cofferdam's real-time three-dimensional geographic coordinates; the inertial measurement unit begins continuously outputting three-axis angular velocity and attitude angle data; and all stress sensors and inclination sensors deployed on the structure begin measuring at a preset sampling frequency.

[0076] The data fusion unit within the multi-source perception module 7 is responsible for processing and integrating heterogeneous raw data from different sensors. This process includes adding precise timestamps to each piece of data and integrating the collected position data, attitude data, and structural stress data into a data frame with a unified data structure. This formatted data frame ensures the consistency and interpretability of information transmitted within the system.

[0077] Finally, the multi-source perception module 7 continuously sends the packaged data frames to the onboard collaborative AI decision module 9 through the internal data bus of the embedded autonomous hardcover system 6 at a preset fixed frequency (for example, 10 Hz), providing real-time and accurate input for its subsequent optimization calculations.

[0078] Step S3: Collaborative decision-making and control instruction generation:

[0079] This step is cyclically executed by the onboard collaborative AI decision module 9 within each control cycle. It receives the real-time status data frame from step S2 and performs collaborative optimization calculations based on it to generate the optimal control instructions at the current moment. The specific technical principles of this step are explained below.

[0080] The onboard collaborative AI decision-making module 9 pre-stores a six-degree-of-freedom hydrodynamic model describing the giant crash box 2's motion in water. This model, a set of mathematical equations, calculates the giant crash box 2's motion state sequence over a finite future time period (the prediction horizon) based on the current state input (including position, attitude, velocity, and angular velocity) and a set of candidate control inputs (i.e., the action sequence of each actuator in the active attitude control module 8). This prediction process forms the basis for subsequent optimal control, enabling the system to foresee future dynamics and plan countermeasures in advance.

[0081] Reference Figure 7 , Figure 7 This is a schematic diagram of the model predictive control principle according to one embodiment of the present invention. The onboard collaborative AI decision module 9 uses a model predictive control algorithm to generate control instructions in each control cycle by solving a constrained finite-time optimal control problem.

[0082] The core of the optimal control problem is to solve an objective function The minimum value of , the objective function in one embodiment can be expressed by the following formula:

[0083] ;

[0084] Where, is the discrete time point of the current control cycle; Based on The impact of momentary state on future The predicted value of the system output (i.e. the position and posture of the anti-collision box cofferdam) at the first step; For the future The expected output on the reference trajectory of the step; For the future The control increment of the step; To predict the time domain length, To control the time domain length; and are positive definite weight matrices used to adjust the relative importance of trajectory tracking error and control increment in the objective function. These matrices are usually diagonal matrices, and the elements on the diagonal reflect the importance of the corresponding state or control quantity; and :These two symbols represent the weighted two-norm square. For a vector and a positive definite weight matrix , the weighted two-norm square operation is defined as: ;in, is a vector Therefore, the formula represents the state prediction error vector Multiply by the weight matrix The weighted squared norm of quantifies the deviation between the predicted state and the reference state, and is calculated based on The matrix weights the deviations of the different components. Represents the control increment vector Multiply by the weight matrix The weighted two-norm square of quantizes the size of the control increment and is calculated based on The matrix weights the changes in the different control input components. This symbol indicates that at the current moment For the future The predicted value of . It emphasizes that the prediction is based on the information at the current moment.

[0085] The first term in the objective function , defined as the trajectory tracking term. Its physical significance lies in quantifying and minimizing the cumulative deviation between the predicted motion state sequence of the cofferdam and the reference trajectory sequence leading to the target installation position over the entire prediction time domain. This term drives the cofferdam to accurately track the preset path.

[0086] The second term in the objective function , defined as the control increment penalty term. Its physical significance is to penalize drastic changes in control instructions between adjacent control cycles. This term serves to smooth the output force of the active attitude control module 8, ensuring smooth movement and avoiding unnecessary impact on the actuator and cofferdam structure.

[0087] While solving the above objective function to obtain the optimal control increment sequence, the optimization process must also meet a series of hard constraints to ensure the safety and physical feasibility of the control instructions. These constraints include:

[0088] While solving the above objective function to obtain the optimal control increment sequence, the optimization process must also meet a series of hard constraints to ensure the safety and physical feasibility of the control instructions. These constraints include:

[0089] Structural stress constraints: ,in .also, The predicted value in the future based on the hydrodynamic model and structural mechanics model is The first step caused by the control action The stress value of each measuring point is The upper limit of the safety stress preset for this measuring point. This constraint ensures that any control instructions generated by the system will not endanger the structural safety of the cofferdam.

[0090] Energy consumption constraints: Here, For the prediction of the future Step 8, the instantaneous total power consumption of all actuators (thrusters, water pumps, etc.) of the active attitude control module, and A safe output power limit set for the autonomous energy module in the embedded autonomous system 6. This constraint ensures that the control instructions are executable on the energy supply.

[0091] Actuator physical constraints: and This constraint defines the physical limits of each actuator in the active attitude control module 8, such as the minimum / maximum thrust of each azimuthing thruster and the maximum water transfer rate of the dynamic ballast water system.

[0092] The onboard collaborative AI decision module 9 solves the constrained optimal control problem in each control cycle and obtains a coverage control time domain. The optimal control increment sequence According to the rolling optimization principle of model predictive control, only the first element in this sequence Used to calculate the actual control instructions at the current moment , and sends the instruction to step S5. In the next control cycle ( The system will repeat the above process.

[0093] Step S4: Online adaptive correction:

[0094] This step, performed in parallel with step S3, aims to continuously improve the prediction accuracy of the built-in hydrodynamic model to address the dynamic changes in the real ocean environment. Due to the complexity and time-varying nature of environmental factors such as waves and currents, the predicted output of a pre-set hydrodynamic model with fixed parameters will inevitably deviate from the actual motion response of the giant crash box 2. This step ensures that the model accurately reflects the current hydrodynamic characteristics by identifying and correcting key model parameters online and in real time.

[0095] In one embodiment of the present invention, the onboard collaborative AI decision module 9 implements online adaptive correction of the hydrodynamic model by executing a recursive least squares (RLS) algorithm.

[0096] First, the hydrodynamic model describing the cofferdam motion is rewritten into a parametric linear form. Taking the hydrodynamic force on a certain degree of freedom (such as sway) as an example, the model can be expressed as:

[0097] ;

[0098] Where, At discrete time points System outputs that can be measured by sensors, such as the total hydrodynamic force acting on the cofferdam; Is a parameter vector to be identified, whose elements include the key unknown parameters of the model, such as the added mass coefficient, linear damping coefficient, quadratic damping coefficient, etc. For example, . is a discrete point in time A measurable regression vector whose elements are kinematic variables corresponding to the parameters to be identified, such as acceleration, velocity, and velocity squared measured by an inertial measurement unit. For example, .

[0099] The core of the recursive least squares algorithm is to use each new set of measurement data , recursively update the parameter vector Estimated value of The iterative process is performed once in each control cycle and includes the following calculations:

[0100] Calculate the forecast error: use the parameter estimate of the previous moment and the regression vector at the current moment , calculate the error between the model predicted output and the actual measured output :

[0101] ;

[0102] Calculate gain vector: Calculate the Kalman gain vector The gain vector determines the prediction error How much to use to correct parameter estimates:

[0103] ;

[0104] Where, is the forgetting factor, which is a constant between (0, 1]. As the algorithm is trained, it gradually reduces the influence weight of old data, thereby being able to track time-varying parameters. is the covariance matrix at the previous moment, which represents the degree of uncertainty of the parameter estimates.

[0105] Update parameter estimates: Update old parameter estimates Plus the prediction error and gain vector The correction amount determined is used to obtain the optimal estimated value of the parameter at the current moment :

[0106] ;

[0107] Update covariance matrix: Update the covariance matrix , for the next round ( moment) is used for iterative calculation.

[0108] ;

[0109] Where, is the identity matrix.

[0110] By repeatedly executing the above recursive calculation in each control cycle, the onboard collaborative AI decision module 9 can obtain a set of real-time updated hydrodynamic model parameter vectors that best fit the current working conditions. This updated set of parameters is immediately applied to the hydrodynamic model in step S3 for motion state prediction and model predictive control calculations for the next control cycle. This online adaptive correction mechanism forms a closed-loop feedback loop for model updates, ensuring that the model underlying the control system maintains high fidelity, thereby improving the robustness and environmental adaptability of the overall control scheme.

[0111] Step S5: Closed-loop adjustment and precise positioning:

[0112] In this step, during each control cycle, the active posture control module 8 receives and executes the control instructions from the onboard collaborative AI decision-making module 9, and at the same time, the onboard collaborative AI decision-making module 9 continuously determines the in-position status.

[0113] The process of the active attitude control module 8 executing the control instruction includes:

[0114] Command reception and analysis: The active attitude control module 8 receives the control command vector from the onboard collaborative AI decision module 9 through the internal data bus This vector contains the exact set values ​​for all actuator subunits in the module. In one embodiment, the structure of this vector is ,in The sub-vector contains the thrust magnitude and thrust direction angle of all omnidirectional thrusters. The sub-vector contains the flow rate and direction of all dynamic ballast water system pumps. The instruction parsing unit within the module assigns the values ​​in the vector to the corresponding actuators based on the preset mapping relationship.

[0115] Instruction Execution: The underlying driver circuitry of the active attitude control module 8 converts the resolved numerical setpoints into the electrical signals required by various physical actuators, such as pulse-width modulation (PWM) signals for controlling the speed of the thruster motors or voltage signals for controlling the opening of valves in the ballast water system. These electrical signals drive the actuators to produce physical actions, thereby applying precise translational control forces and attitude adjustment torques to the giant crash box 2.

[0116] At the same time, the onboard collaborative AI decision module 9 continues to execute the in-place status determination process, which is the basis for terminating the entire closed-loop control process. The determination process includes:

[0117] Deviation calculation: In each control cycle, the module calculates the deviation based on the real-time status data obtained from the multi-source perception module 7. and the preset target location status , calculate the current position deviation and attitude deviation. Position deviation Defined as the Euclidean distance between the current coordinate and the target coordinate, The attitude deviation is defined as the absolute value of the difference between the current attitude angle and the target attitude angle, that is, .

[0118] In-position condition judgment: The module compares the calculated deviation with a set of preset thresholds, which include position deviation thresholds and attitude deviation threshold The module continuously checks whether the following conditions are met simultaneously: , , ,and .

[0119] Stability verification: If and only if all conditions in step 2 are met within a preset, continuous time window When all the above conditions are met, the system determines that the system has completed the precise positioning. This stability verification step is intended to eliminate false judgments caused by transient disturbances or measurement noise.

[0120] Once stability verification is passed, the onboard collaborative AI decision module 9 stops sending active control commands to the active attitude control module 8 (or sends a zero-increment command to maintain the current state) and sends a status signal to the external master control platform indicating task completion. This completes the closed-loop adjustment and precise positioning steps.

[0121] Step S6: Online diagnosis and fault-tolerant control:

[0122] This step is an optional enhanced implementation of the method of the present invention, which aims to improve the reliability and robustness of the entire embedded autonomous hardcover system 6. This step runs continuously throughout the entire operation process and includes two parts: online diagnosis and fault tolerance control.

[0123] Online diagnosis is performed by the onboard collaborative AI decision module 9, which is used to monitor the health status of each hardware component in the multi-source perception module 7 and the active posture control module 8 in real time.

[0124] For the diagnosis of sensor components, an analytical redundancy method is used. This method includes:

[0125] Readings from physically redundant sensors of the same type (e.g., multiple tilt sensors) are cross-compared. Functionally related sensors of different types (e.g., the position output of a satellite positioning system and the position output integrated by an inertial measurement unit) are checked for consistency. If the deviation between the comparison or verification results of any sensor group consistently exceeds a preset fault threshold, or if a sensor experiences signal loss, constant value, or abnormal noise levels, that sensor is diagnosed as faulty.

[0126] For the diagnosis of actuator components, a model-based approach is adopted. This approach includes:

[0127] The onboard collaborative AI decision module 9 takes the control command sent to a specific actuator as input and uses its internal hydrodynamic model to predict the system response (e.g., acceleration or angular velocity) that the command should produce. Simultaneously, it compares the actual system response measured by the multi-source perception module 7. If the residual between the model's predicted response and the actual measured response is consistently greater than a preset threshold, the actuator is not operating as expected and is diagnosed as faulty. The system also monitors the feedback electrical signals (e.g., motor current) from each actuator; any abnormal current values ​​can also serve as a basis for fault diagnosis.

[0128] When the online diagnosis part identifies and confirms that a component has failed, the onboard collaborative AI decision module 9 immediately starts the fault-tolerant control program and dynamically adjusts its internal algorithms and control strategies.

[0129] If a sensor failure is diagnosed, the onboard collaborative AI decision module 9 will remove the data source of the faulty sensor from its state estimation and data fusion algorithm, and only rely on the remaining, fully functional sensor data to continue to generate an estimate of the cofferdam state.

[0130] If an actuator failure is diagnosed (e.g., an omnidirectional thruster failure), the onboard collaborative AI decision module 9 dynamically modifies the constraints in the model predictive control (MPC) optimization problem in step S3. Specifically, it updates the physical constraints associated with the faulty actuator, for example, setting its maximum and minimum output capabilities (e.g., thrust) to zero in the constraint equations. During the next control cycle's optimization solution, the MPC algorithm automatically recalculates and assigns control tasks under these new constraints, compensating the control torque originally borne by the faulty actuator with the remaining functional actuators by solving the optimal solution.

[0131] Through this online diagnosis and fault-tolerant control mechanism, the present invention can avoid catastrophic failure of the system when some components fail. Instead, it can reconstruct the control strategy and use the system redundancy to continue to perform the task, thereby maximizing the continuity and safety of offshore operations.

[0132] In order to more specifically illustrate the technical solution of the present invention and its implementation effect, a complete application example will be given below with the background of a super-large anti-collision box cofferdam lowering project. The scenes and steps described in this example should be combined with the attached Figure 1 To the attached Figure 7 Understand.

[0133] In a typical operation scenario, refer to the attached Figure 2The large floating crane 5 has already completed the anchoring and positioning in the waters near the pile foundation construction platform 3 through its floating crane anchor system 1. The large floating crane 5 has lifted a giant anti-collision box 2, which was integrated with the embedded autonomous finishing system 6 of the present invention during the onshore manufacturing stage, from the transport barge. At this time, the giant anti-collision box 2 is suspended above the water surface, ready for the final lowering and precise positioning operation. In certain working conditions, please refer to the attached Figure 3 and attached Figure 4 In the manner shown, wind cables are added to improve operational stability.

[0134] The operator at the control console at the work site sends a system activation command via a wireless data link to the embedded autonomous hardcover system 6 located inside the giant crashproof housing 2. Upon receiving the command, the onboard collaborative AI decision module 9 powers on and executes an initialization procedure, which includes a self-test of all system hardware components to confirm the functional integrity of the multi-source perception module 7 and the active attitude control module 8. After passing the self-test, the system enters standby mode. All sensor subunits within the multi-source perception module 7, including the satellite positioning system, inertial measurement unit, stress sensor, and tilt sensor, begin collecting and transmitting real-time data on the initial state of the cofferdam (position and attitude) and environmental conditions (such as initial sway) to the onboard collaborative AI decision module 9.

[0135] The operator instructed the large floating crane 5 to slowly lower the giant anti-collision casing 2. During the lowering process, the cofferdam encountered a sudden cross-current, causing it to deviate from the planned vertical lowering path. The system executed the following disturbance compensation closed-loop control:

[0136] Perception: The satellite positioning system and inertial measurement unit within the multi-source perception module 7 instantly detect unexpected lateral position drift and roll attitude tilt caused by ocean currents within milliseconds. This deviation data is packaged and sent to the decision module in real time.

[0137] Decision-making: After receiving the state deviation data, the onboard collaborative AI decision-making module 9 immediately executes the following within one control cycle (for example, 0.1 seconds): Figure 7 The model predictive control (MPC) optimization calculation based on the principle shown is performed. This calculation primarily aims to minimize trajectory tracking error, treating ocean current disturbances as external forces to be overcome. The optimization process generates a set of optimal control instructions while satisfying all constraints. These constraints include ensuring that the stress values ​​at key structural points, measured by the stress sensors within the multi-source sensing module 7, remain below a safety upper limit under the control force, and ensuring that the total power required for the control instructions does not exceed the system's energy supply limit.

[0138] Execution: The active attitude control module 8 immediately interprets and executes the control instructions from the decision module. Its circumferentially arranged omnidirectional thrusters work in concert to generate a net thrust equal in magnitude to, and opposite in direction to, the current, actively counteracting the lateral drift caused by the current. Simultaneously, the dynamic ballast water system fine-tunes itself based on the instructions, generating a restoring torque by rapidly transferring water between tanks within the cofferdam to actively suppress the rolling attitude caused by the current.

[0139] Through the above-mentioned rapid closed loop of perception-decision-execution, the system effectively compensates for external environmental disturbances, enabling the giant anti-collision box 2 to resist the impact of ocean currents and quickly return to the predetermined vertical lowering trajectory. The entire process requires no human intervention.

[0140] When the bottom of the giant anti-collision box 2 approaches its final installation elevation on the pile foundation construction platform 3, the onboard collaborative AI decision module 9 automatically switches the focus of the control strategy to high-precision position and attitude maintenance. At this time, the objective function of the model predictive control will focus more on minimizing the deviation between the current state and the final target state. The system makes fine adjustments at the millimeter level, counteracting small underwater disturbances through continuous, small, and coordinated control outputs of the omnidirectional thrusters and dynamic ballast water system. This process continues until the real-time position deviation and attitude deviation of the cofferdam are stably less than the preset threshold value in step S5 within a preset duration.

[0141] Once the onboard collaborative AI decision module 9 confirms that the criteria for precise positioning have been met, it automatically sends a task completion confirmation signal to the master control platform on the large floating crane 5 and stops all control outputs from the active attitude control module 8, placing the system in a locked hold state. Upon receiving this signal, on-site operators can proceed with subsequent permanent fixing operations, such as welding the cofferdam to the foundation structure on the pile foundation construction platform 3 or achieving final locking through other mechanical connection methods, thus completing the intelligent lowering of the entire offshore ultra-large anti-collision box cofferdam.

[0142] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent construction method for an offshore super-large anti-collision box cofferdam, characterized in that: The method comprises the following steps: S1. Integrate an embedded autonomous assembly system in the anti-collision box cofferdam body, the embedded autonomous assembly system including an active attitude control module, a multi-source perception module, and an onboard collaborative AI decision-making module; S2. Hoist the anti-collision box cofferdam into the water and activate the embedded autonomous assembly system to drive the multi-source perception module to collect and send real-time status data of the anti-collision box cofferdam to the onboard collaborative AI decision module, wherein the real-time status data includes structural stress data; S3. The onboard collaborative AI decision-making module, based on a model predictive control algorithm, aims to minimize the deviation between the predicted motion state of the anti-collision box cofferdam and the target installation position, and simultaneously uses the structural stress data as a safety constraint to perform collaborative optimization calculations to generate control instructions; The model predictive control algorithm is specifically used to solve an optimization problem in each control cycle with trajectory tracking error and control increment as the objective function and structural stress data as the safety constraint, generate an optimal control instruction sequence, and execute the first instruction of the optimal control instruction sequence. The trajectory tracking error is intended to minimize the deviation between the predicted motion state of the anti-collision box cofferdam and the target installation position, and the control increment is intended to limit the variation range of the control instruction to ensure control smoothness. S4. The onboard collaborative AI decision-making module continuously compares the predicted output of the hydrodynamic model built into the embedded autonomous hardcover system with the real-time status data, and performs online adaptive correction on the parameters of the hydrodynamic model to improve the prediction accuracy of the model predictive control algorithm; S5. The active attitude control module receives and executes control instructions, and performs continuous closed-loop adjustment on the attitude and plane position of the anti-collision box cofferdam until the anti-collision box cofferdam is stabilized within a preset accuracy range of the target installation position and is accurately positioned.

2. The intelligent construction method of an offshore super-large anti-collision box cofferdam according to claim 1 is characterized in that: In step S1, the embedded autonomous hardcover system includes an active attitude control module, a multi-source perception module, and an onboard collaborative AI decision module. The steps include: The active attitude control module is composed of a dynamic ballast water system and multiple omnidirectional thrusters, wherein: The dynamic ballast water system generates attitude restoring torque by transferring water mass between internal water tanks; The omnidirectional thruster provides translation and rotation control forces in the horizontal plane; The multi-source perception module is used to obtain a satellite positioning system for the absolute position of the anti-collision box cofferdam, an inertial measurement unit for measuring the attitude, and a plurality of stress sensors for obtaining structural stress data; The onboard collaborative AI decision-making module is an embedded computing unit, which integrates a hydrodynamic model, a model predictive control algorithm module, and an online adaptive correction algorithm module.

3. The intelligent construction method of an offshore super-large anti-collision box cofferdam according to claim 1 is characterized in that: In step S2, the anti-collision box cofferdam is hoisted into the water and the embedded autonomous assembly system is activated to drive the multi-source perception module to collect and send real-time status data of the anti-collision box cofferdam to the onboard collaborative AI decision module. The steps of the real-time status data including structural stress data include: Performing power-on initialization on the embedded autonomous hardcover system, wherein the power-on initialization triggers all sensors in the multi-source perception module to enter a continuous working state; The multi-source perception module integrates the acquired position data, posture data and structural stress data of the anti-collision box cofferdam into a data frame in a unified format, and sends it to the onboard collaborative AI decision-making module in real time through the internal data bus of the embedded autonomous hardcover system.

4. The intelligent construction method of an offshore super-large anti-collision box cofferdam according to claim 1 is characterized in that: In step S3, the onboard collaborative AI decision module performs collaborative optimization calculations based on a model predictive control algorithm to generate control instructions with the goal of minimizing the deviation between the predicted motion state of the anti-collision box cofferdam and the target installation position, and simultaneously uses the structural stress data as a safety constraint. The steps include: The objective function solved by the collaborative optimization calculation is composed of a trajectory tracking term and a control increment penalty term: The trajectory tracking term is used to quantify and minimize the cumulative deviation between the predicted motion state sequence of the anti-collision box cofferdam and the reference trajectory sequence leading to the target installation position in the prediction time domain; The control increment penalty term is used to penalize drastic changes in the control instructions between adjacent control cycles, so as to smooth the output force of the active attitude control module and ensure smooth movement.

5. The intelligent construction method of an offshore super-large anti-collision box cofferdam according to claim 4 is characterized in that: The collaborative optimization calculation not only solves the objective function but also satisfies the energy consumption constraint; The energy consumption constraint is used to ensure that the generated control instructions are executable within the power supply range of the autonomous energy module of the embedded autonomous hardcover system; The onboard collaborative AI decision-making module uses the instantaneous total power consumption of the active attitude control module as a constraint variable and ensures that the calculated value of the instantaneous total power consumption does not exceed the safe power upper limit set by the autonomous energy module.

6. The intelligent construction method of an offshore super-large anti-collision box cofferdam according to claim 1 is characterized in that: In step S4, the onboard collaborative AI decision module continuously compares the predicted output of the hydrodynamic model built into the embedded autonomous hardcover system with the real-time status data, and performs online adaptive correction on the parameters of the hydrodynamic model to improve the prediction accuracy of the model predictive control algorithm. The steps include: Online adaptive correction is performed by the onboard collaborative AI decision-making module, which uses a recursive least squares algorithm to take the error between the predicted output of the hydrodynamic model and the real-time state data as input, and iteratively updates the fluid damping coefficient and additional mass coefficient in the hydrodynamic model.

7. The intelligent construction method of an offshore super-large anti-collision box cofferdam according to claim 1 is characterized in that: In step S5, the active attitude control module receives and executes the control instruction, and continuously adjusts the attitude and plane position of the anti-collision box cofferdam in a closed loop until the anti-collision box cofferdam is stabilized within a preset accuracy range of the target installation position. The steps of completing the precise positioning include: The basis for determining whether precise positioning is completed is the plane position deviation and attitude inclination deviation monitored by the multi-source perception module; When the plane position deviation and the attitude inclination angle deviation are both continuously smaller than the respective preset allowable error thresholds within a continuous preset period of time, it is confirmed that the anti-collision box cofferdam is stable within the preset accuracy range of the target installation position.

8. The intelligent construction method of an offshore super-large anti-collision box cofferdam according to claim 7 is characterized in that: The continuous closed-loop adjustment is carried out while an external large floating crane provides vertical hanging force for the anti-collision box cofferdam; the control instructions generated by the onboard collaborative AI decision-making module are used to autonomously compensate for the disturbance caused by waves and currents acting on the anti-collision box cofferdam during the lowering process of the anti-collision box cofferdam.

9. The intelligent construction method of an offshore super-large anti-collision box cofferdam according to claim 1 is characterized in that: The method further comprises: S6. Perform online diagnosis and fault-tolerant control on the embedded autonomous hardcover system: The onboard collaborative AI decision-making module continuously diagnoses the operating status of the active posture control module and the multi-source perception module. When a component failure is diagnosed, the system model on which the model predictive control algorithm relies is dynamically adjusted to reflect the currently available component status, and collaborative optimization calculations are continued based on the adjusted system model to generate control instructions.

10. An intelligent construction system for an offshore super-large anti-collision box cofferdam, applied to the method according to any one of claims 1 to 9, characterized in that: The system is an embedded autonomous fine-fitting system integrated into the anti-collision box cofferdam body, and the embedded autonomous fine-fitting system includes: A multi-source sensing module is used to collect real-time status data of the anti-collision box cofferdam, wherein the real-time status data includes structural stress data; An active attitude control module, configured to apply a control force and a control torque to the anti-collision box cofferdam; An onboard collaborative AI decision-making module, connected to the multi-source perception module and the active attitude control module, configured to perform the following operations: Based on a model predictive control algorithm, with the goal of minimizing the deviation between the predicted motion state of the anti-collision box cofferdam and the target installation position, and simultaneously using the structural stress data received from the multi-source perception module as a safety constraint, a collaborative optimization calculation is performed to generate control instructions; By continuously comparing the predicted output of the hydrodynamic model built into the embedded autonomous hardcover system with the real-time status data, the parameters of the hydrodynamic model are adaptively corrected online; The control instruction is sent to the active attitude control module to drive the active attitude control module to perform continuous closed-loop adjustment on the attitude and plane position of the anti-collision box cofferdam until the anti-collision box cofferdam is stabilized within a preset accuracy range of the target installation position.

Images