A ship crane anti-swing adaptive monitoring system and a control method thereof

Abstract

The application provides a ship crane anti-swing adaptive monitoring system and a control method thereof. The system comprises a perception module, a core processing module and an execution and feedback module connected with each other. The execution and feedback module is used for converting anti-swing control instructions into driving signals, controlling crane movement to offset swing, and forming a closed-loop control loop. A man-machine interaction module is connected with the core processing module, and is used for system state display, early warning information pushing and parameter configuration. The core processing module comprises a data fusion and state estimation unit. An extended Kalman filtering algorithm is adopted to fuse multi-source sensor data and estimate a two-dimensional swing angle and angular velocity of a load in real time. An adaptive anti-swing control unit adopts an adaptive controller based on deep reinforcement learning, and calculates optimal crane movement compensation instructions in real time according to current state estimation values to solve the problems of strong operator dependence, poor control robustness and inability to adaptively adjust in the prior art.

Description

Technical Field

[0001] This invention relates to an anti-sway adaptive monitoring system for ship cranes and its control method. Background Technology

[0002] Ship cranes are core equipment for maritime cargo handling and marine engineering operations, widely used in areas such as offshore replenishment, offshore platform construction, and maritime rescue. During maritime operations, ships are affected by environmental factors such as wind, waves, and currents, resulting in six degrees of freedom of motion: roll, pitch, heave, sway, yaw, and bow roll. These motions are transmitted to the hoisted cargo through the crane, causing severe and unpredictable swaying of the cargo. Hoisted swaying seriously threatens operational safety, easily leading to cargo collisions, personnel injuries, and significantly reducing loading and unloading efficiency.

[0003] Traditional methods for anti-swaying cranes primarily rely on the operator's experience to perform counter-swaying operations. This involves the operator manually controlling the crane to move in the opposite direction based on observed cargo swaying. This method demands a very high level of operator skill, and its effectiveness is inconsistent in complex sea conditions, making it difficult to guarantee operational safety.

[0004] To address these issues, existing technologies have proposed automatic control methods based on preset models. For example, PID controllers are used to adjust the crane's movement based on swing angle feedback, or input shaping techniques are employed to filter operator commands and reduce swing excitation. While these methods are effective under ideal conditions, in actual offshore operations, parameters such as load weight, rope length, and wave conditions frequently change, leading to significant deviations between the preset model and the actual system. This results in a sharp decline in control effectiveness and poor robustness.

[0005] Chinese patent CN120589600A discloses a method for anti-swaying of ship cranes based on model predictive control. This method establishes a dynamic model of the crane-load system and performs rolling optimization to solve for control commands based on the model. However, this method relies on an accurate mathematical model, and its control performance deteriorates significantly when system parameters change or when unmodeled disturbances exist.

[0006] Chinese patent CN121454937A discloses a ship crane sway monitoring device that uses an inertial measurement unit mounted on the hook to measure the sway angle. Although this solution achieves real-time monitoring of the sway state, it relies on only a single sensor and is easily affected by factors such as ship movement and electromagnetic interference in the complex marine environment. The state estimation accuracy is insufficient, and it does not provide an effective adaptive control method.

[0007] Chinese patent CN121404963A discloses an anti-sway control system for ship cranes, which employs multi-sensor data fusion technology to improve state estimation accuracy and designs a controller based on a linear quadratic regulator. While this scheme improves upon multi-sensor fusion, its control algorithm remains based on a fixed linear model, making it unable to adapt to dynamically changing operating environments. Its control effectiveness is limited under load variations or strong disturbances.

[0008] Therefore, there is a need for a ship crane anti-sway monitoring system that can sense the swaying state of the crane, make intelligent decisions to generate optimal control commands, and has strong adaptive capabilities, in order to solve the technical problems of poor robustness and weak adaptability of existing technologies under complex sea conditions and changing operating conditions. Summary of the Invention

[0009] The purpose of this invention is to overcome the shortcomings of the prior art by providing a ship crane anti-sway adaptive monitoring system and its control method, thereby solving the problems of strong operator dependence, poor control robustness, and inability to adaptively adjust in the prior art. This objective of the invention is achieved as follows:

[0010] This invention proposes an adaptive anti-sway monitoring system for ship cranes, comprising: a sensing module for real-time acquisition of data on the crane's sway state, crane body motion state, and environmental disturbances; a core processing module connected to the sensing module for fusing the sensing data, estimating the crane's sway angle and angular velocity in real time, and generating anti-sway control commands; an execution and feedback module connected to the core processing module and the crane drive system for converting the anti-sway control commands into drive signals to control the crane's movement to counteract the sway and form a closed-loop control circuit; and a human-machine interaction module connected to the core processing module for system status display, early warning information push, and parameter configuration. The core processing module includes a data fusion and state estimation unit that uses an extended Kalman filter algorithm to fuse multi-source sensor data and estimate the crane's two-dimensional sway angle and angular velocity in real time; and an adaptive anti-sway control unit that uses a deep reinforcement learning-based adaptive controller to calculate the optimal crane motion compensation command in real time based on the current state estimate.

[0011] Furthermore, the sensing module includes a hook attitude sensor for sensing the hook attitude; a crane motion sensor for sensing the crane motion state; and an environmental sensor for sensing environmental disturbances.

[0012] Furthermore, the sensing module includes an inertial measurement unit and a visual marker camera mounted on the hook.

[0013] Furthermore, the sensing module includes encoders installed on the crane's slewing mechanism and luffing mechanism.

[0014] Furthermore, the adaptive anti-sway control unit adopts a deep deterministic strategy gradient algorithm, and its training environment is a ship-crane-load coupled dynamics simulation model. Its reward function comprehensively considers the sway suppression effect, energy consumption and operation smoothness.

[0015] This invention also proposes a method for anti-sway control of a ship crane, comprising the following steps: real-time acquisition of crane sway state, crane body motion state, and environmental disturbance data through multi-source sensors; fusing the multi-source sensor data using an extended Kalman filter algorithm to estimate the two-dimensional sway angle and angular velocity of the crane; inputting the state estimate into a pre-trained deep reinforcement learning controller to calculate the optimal crane motion compensation command; converting the control command into a drive signal to drive the crane to perform compensation motion; continuously monitoring the crane state after compensation, and feeding back the monitored crane state data to the data fusion step to form closed-loop control.

[0016] Furthermore, the multi-source sensors include a hook attitude sensor, a crane motion sensor, and an environmental sensor.

[0017] Furthermore, the deep reinforcement learning controller employs a deep deterministic policy gradient algorithm, whose reward function comprehensively considers swing suppression effect, energy consumption, and operational smoothness.

[0018] This invention also proposes a computer-readable storage medium storing a computer program. When executed by a processor, the computer program performs the following steps: real-time acquisition of data on the sway state of the hoist, the motion state of the crane body, and environmental disturbances from multiple sources of sensors; fusing the multi-source sensor data using an extended Kalman filter algorithm to estimate the two-dimensional sway angle and angular velocity of the hoist; inputting the state estimate into a pre-trained deep reinforcement learning controller to calculate the optimal crane motion compensation command; converting the control command into a drive signal to drive the crane to perform the compensation motion; continuously monitoring the hoist state after compensation and feeding back the monitored hoist state data to the data fusion step to form a closed-loop control.

[0019] Furthermore, the deep reinforcement learning controller employs a deep deterministic policy gradient algorithm, whose reward function comprehensively considers swing suppression effect, energy consumption, and operational smoothness.

[0020] Compared with the prior art, the beneficial effects of the present invention are: by fusing multiple heterogeneous sensors, including the coordinated work of the hook inertial measurement unit, visual marker camera, crane encoder, ship attitude reference system and wind speed and direction sensor, the accuracy and reliability of the load sway state estimation are significantly improved, overcoming the shortcomings of single sensors being susceptible to the influence of ship motion, electromagnetic interference and other factors, and providing accurate and reliable state information for control decision-making.

[0021] The adaptive anti-sway control unit based on the deep deterministic strategy gradient algorithm does not rely on a precise mathematical model. It can learn the optimal anti-sway strategy autonomously through continuous training in the ship-crane-load coupled dynamics simulation environment. It has strong robustness and adaptability to changes in load weight, rope length and unknown wind and wave disturbances, and effectively solves the problem that the control effect of traditional PID control or input shaping technology drops sharply when the working conditions change.

[0022] The data fusion and state estimation unit employs the extended Kalman filter algorithm. Based on the crane-load kinematic equations, it effectively fuses multi-source data with noise and delay, and accurately estimates the load's two-dimensional swing angle and angular velocity in real time, providing high-quality input to the controller and ensuring control accuracy.

[0023] Experimental data show that the hoisting sway stabilization time under the control system of this invention is 68% longer than that of uncontrolled natural decay, and the overall operation efficiency is increased by more than 30%. The automatic and rapid anti-sway control reduces the sway amplitude and stabilization time of the cargo, reduces the risk of cargo collision and personnel injury, frees the operator from the tense manual anti-sway operation, and significantly improves the safety and efficiency of offshore operations.

[0024] The system has full-cycle monitoring capabilities. The human-machine interface module provides a real-time visual monitoring platform that displays the hoist swing trajectory, system status, and issues abnormal warnings. It also has data recording and analysis functions, which can be used to optimize work processes, conduct accident reviews, and train operators. Attached Figure Description

[0025] Figure 1 This is a comparative schematic diagram of the oscillation elimination process under a typical working scenario in Embodiment 3 of the present invention. Detailed Implementation

[0026] To enhance understanding of the present invention, the present invention will be further described in detail below with reference to embodiments and accompanying drawings. These embodiments are only used to explain the present invention and do not constitute a limitation on the scope of protection of the present invention.

[0027] Example 1

[0028] This embodiment discloses a ship crane anti-sway adaptive monitoring system. The system adopts a hierarchical distributed architecture and includes four core parts: a sensing module, a core processing module, an execution and feedback module, and a human-computer interaction module.

[0029] The sensing module is used to collect real-time data on the sway status of the hoist, the motion status of the crane itself, and environmental disturbances. The sensing module establishes a data connection with the core processing module via CAN bus and Ethernet, enabling synchronous acquisition and transmission of data from multiple sensor sources.

[0030] The core processing module uses an industrial-grade embedded computer running a real-time Linux kernel. It connects to the sensing module to fuse and process the sensed data, estimate the load's sway angle and angular velocity in real time, and generate anti-sway control commands. The core processing module internally includes a data fusion and state estimation unit and an adaptive anti-sway control unit.

[0031] The data fusion and state estimation unit employs an extended Kalman filter algorithm to fuse multi-source sensor data and estimate the two-dimensional swing angle and angular velocity of the hoist in real time. Specifically, this unit first performs spatiotemporal alignment and outlier removal on the multi-channel sensor data. Then, it establishes a state transition model based on the crane-hoist kinematic equations, using the theoretical hook position calculated from encoder data as the prediction input and inertial measurement unit and visual data as the observation input. Through a prediction-update iterative process, it outputs the fused actual swing angles α and β of the hoist relative to the vertical line and its angular velocity.

[0032] The adaptive anti-sway control unit employs an adaptive controller based on deep reinforcement learning, which calculates the optimal crane motion compensation command in real time based on the current state estimate. This control unit incorporates a pre-trained deep deterministic policy gradient algorithm network. It receives the state estimate, target position, and environmental data output by the data fusion and state estimation unit, and outputs the current optimal slewing angular velocity compensation and amplitude angular velocity compensation through neural network forward computation.

[0033] The execution and feedback module connects to the core processing module and the crane drive system. It converts anti-sway control commands into drive signals, controlling the crane's movement to counteract sway and forming a closed-loop control circuit. This module includes a motion controller and a servo drive system. The motion controller superimposes control commands onto the operator's original operating commands to generate a master control command, driving the servo valve group or frequency converter to perform compensating motion. After the crane moves, the sensing module continuously monitors the new state after compensation and feeds the monitoring data back to the core processing module, forming closed-loop control.

[0034] The human-machine interface module is connected to the core processing module and is used for system status display, early warning information push, and parameter configuration. This module includes an industrial-grade touch screen display unit and a data storage and analysis unit. The touch screen display unit displays the hoist's three-dimensional trajectory, sway parameters, system status, and early warning information in real time. The data storage and analysis unit records data throughout the entire operation process and supports playback and efficiency analysis.

[0035] The workflow of this system embodiment is as follows: After system startup, the sensing module continuously collects data, the core processing module performs data fusion and state estimation, the adaptive anti-sway control unit calculates control commands, and the execution and feedback module drives the crane to perform compensating motion, forming a closed-loop control. Tested under typical sea conditions, the system can accurately suppress load sway in real time.

[0036] Example 2

[0037] This embodiment further discloses a specific sensor configuration scheme for the sensing module.

[0038] The sensing module includes a hook attitude sensor, a crane motion sensor, and an environmental sensor.

[0039] The hook attitude sensor includes an inertial measurement unit (IMU) mounted on the hook and a visual marker camera. The IMU is a miniature, waterproof, and shockproof IMU, fixed to the hook, used to measure the hook's three-axis acceleration and angular velocity. The visual marker camera is an industrial-grade camera, mounted at the end of the crane boom, aligned with visual markers in the hook area, used to assist in positioning the hook.

[0040] The crane motion sensors include encoders mounted on the crane's slewing and luffing mechanisms. These encoders are absolute encoders, mounted on the crane's slewing platform and boom hinge points respectively, used to measure the slewing angle and luffing angle. Additionally, the crane's hoisting mechanism is also equipped with encoders for measuring rope length.

[0041] Environmental sensors include a hull attitude reference system and wind speed and direction sensors. The hull attitude reference system, mounted on top of the bridge, measures the hull's six degrees of freedom motion attitude. The wind speed and direction sensors are ultrasonic anemometers, mounted high on the hull, and measure ambient wind speed and direction.

[0042] Data from each sensor is acquired synchronously via CAN bus and Ethernet at a frequency of 100Hz. After spatiotemporal alignment and outlier removal, the data is input to the data fusion and state estimation unit of the core processing module. After multi-source data fusion, the accuracy of sway state estimation is significantly improved, overcoming the shortcomings of single sensors being susceptible to factors such as ship motion and electromagnetic interference.

[0043] Example 3

[0044] This embodiment further discloses the training and deployment scheme of the deep deterministic policy gradient algorithm controller in the adaptive anti-sway control unit.

[0045] The adaptive anti-sway control unit adopts a deep deterministic policy gradient algorithm. Its training environment is a ship-crane-load coupled dynamics simulation model. Its reward function comprehensively considers the sway suppression effect, energy consumption and operation smoothness.

[0046] The ship-crane-load coupled dynamics simulation model is established based on the Lagrange equations, considering the coupling effects of the ship's six-degree-of-freedom motion, the crane's rotation and luffing motion, the load's two-dimensional swaying motion, and wind and wave disturbances. The simulation environment is configured with various operating parameters, including different sea state levels, different load weights, and different rope lengths, to train the robustness and adaptability of the controller.

[0047] The reward function is designed as a weighted sum of the square of the negative yaw angle, the square of the yaw angular velocity, and the square of the control action. Specifically, the reward function R equals the square of the negative α multiplied by the square of the yaw angle θ, plus the square of the yaw angular velocity ω, plus the square of the control action α, where α, β, and γ are weighting coefficients. In this embodiment, the weighting coefficients α, β, and γ are set to 1.0, β, and γ to 0.1. This reward function design enables the policy network to minimize energy consumption and operational amplitude while suppressing yaw, avoiding secondary perturbations.

[0048] The deep deterministic policy gradient algorithm uses a two-layer fully connected neural network with 256 hidden neurons. The training process employs an experience replay mechanism and a target network soft update mechanism, converging after 10,000 training iterations. After training, the policy network parameters are fixed and deployed to the embedded computer of the core processing module.

[0049] The controller's workflow is as follows: First, it reads the current state estimate, including the swing angle θ, swing angular velocity ω, crane position, and environmental information. Then, it inputs the state estimate into a pre-trained deep deterministic policy gradient algorithm network. The network performs forward computation and outputs the optimal control action, including the slewing angular velocity compensation and the amplitude angular velocity compensation. The original control action output by the network is smoothed and limited to prevent system jitter and ensure safety. Finally, the processed control action is sent to the execution module to drive the crane servo system to perform the compensation motion. The system enters a state of waiting for the next control cycle, with a control cycle of 10 milliseconds, and then returns to step one, forming a continuously operating closed-loop control system.

[0050] The sway suppression effect of the present invention was verified in actual ship tests. Figure 1 The diagram illustrates a comparison of the sway reduction process under typical operating scenarios, with the horizontal axis representing time and the vertical axis representing the sway angle θ. The dashed lines in the diagram represent the sway angle change under uncontrolled natural decay, while the solid lines represent the sway angle change under the control of the system of this invention.

[0051] Test data shows that when swaying occurs after the cargo is lifted, the sway angle reaches a peak of approximately 12 degrees at about 13 seconds without control, then slowly decays, finally reaching approximately 0 degrees after about 22 seconds. However, with the control system of this invention, the system activates quickly, rapidly estimates the sway state, and generates a reverse compensation motion command, driving the crane boom to perform intelligently planned compensation motion. The sway angle reaches a peak of approximately 10 degrees at about 11 seconds, a decrease of about 17% compared to the uncontrolled situation, then rapidly decays, reaching approximately 0 degrees after about 15 seconds. The sway stabilization time under the control system of this invention is improved by 68% compared to the uncontrolled natural decay, resulting in an overall work efficiency improvement of over 30%.

[0052] Automatic and rapid anti-sway control reduces cargo sway amplitude and stabilization time, lowers the risk of cargo collisions and personnel injuries, frees operators from tedious manual anti-sway operations, and significantly improves the safety and efficiency of offshore operations.

[0053] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A ship crane anti-sway adaptive monitoring system, characterized in that, The system includes a sensing module for real-time acquisition of data on the hoist's sway status, the crane's motion status, and environmental disturbances; a core processing module connected to the sensing module for fusing the sensing data, estimating the hoist's sway angle and angular velocity in real time, and generating anti-sway control commands; an execution and feedback module connected to the core processing module and the crane drive system for converting the anti-sway control commands into drive signals to control the crane's movement to counteract the sway and form a closed-loop control circuit; and a human-machine interface module connected to the core processing module for system status display, early warning information push, and parameter configuration. The core processing module includes a data fusion and state estimation unit that uses an extended Kalman filter algorithm to fuse multi-source sensor data and estimate the hoist's two-dimensional sway angle and angular velocity in real time. The adaptive anti-sway control unit adopts an adaptive controller based on deep reinforcement learning, which calculates the optimal crane motion compensation command in real time based on the current state estimate.

2. The ship crane anti-sway adaptive monitoring system according to claim 1, characterized in that, The sensing module includes a hook attitude sensor for sensing the hook attitude; a crane motion sensor for sensing the crane motion status; and an environmental sensor for sensing environmental disturbances.

3. The ship crane anti-sway adaptive monitoring system according to claim 1, characterized in that, The sensing module includes an inertial measurement unit and a visual marker camera mounted on the hook.

4. The ship crane anti-sway adaptive monitoring system according to claim 1, characterized in that, The sensing module includes encoders installed on the crane's slewing mechanism and luffing mechanism.

5. The ship crane anti-sway adaptive monitoring system according to claim 1, characterized in that, The adaptive anti-sway control unit adopts a deep deterministic strategy gradient algorithm. Its training environment is a ship-crane-load coupled dynamics simulation model. Its reward function comprehensively considers the sway suppression effect, energy consumption and operation smoothness.

6. A method for anti-sway control of a ship crane, characterized in that, Includes the following steps: The hoisting sway state, crane body motion state and environmental disturbance data are collected in real time by multiple source sensors; the extended Kalman filter algorithm is used to fuse the multi-source sensor data to estimate the two-dimensional sway angle and angular velocity of the hoisting. The state estimate is input into a pre-trained deep reinforcement learning controller to calculate the optimal crane motion compensation command; the control command is converted into a drive signal to drive the crane to perform the compensation motion; the load state after compensation is continuously monitored, and the monitored load state data is fed back to the data fusion step to form a closed-loop control.

7. The anti-sway control method for ship cranes according to claim 6, characterized in that, The multi-source sensors include a hook attitude sensor, a crane motion sensor, and an environmental sensor.

8. The anti-sway control method for ship cranes according to claim 6, characterized in that, The deep reinforcement learning controller employs a deep deterministic policy gradient algorithm, whose reward function comprehensively considers swing suppression effect, energy consumption, and operational smoothness.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it performs the following steps: real-time acquisition of data on the swaying state of the hoist, the motion state of the crane body, and environmental disturbances through multi-source sensors; and estimation of the two-dimensional swaying angle and angular velocity of the hoist by fusing the multi-source sensor data using an extended Kalman filter algorithm. The state estimate is input into a pre-trained deep reinforcement learning controller to calculate the optimal crane motion compensation command; the control command is converted into a drive signal to drive the crane to perform the compensation motion; the load state after compensation is continuously monitored, and the monitored load state data is fed back to the data fusion step to form a closed-loop control.

10. The computer-readable storage medium according to claim 9, characterized in that, The deep reinforcement learning controller employs a deep deterministic policy gradient algorithm, whose reward function comprehensively considers swing suppression effect, energy consumption, and operational smoothness.

Images