Marine intelligent cable releasing hook control system and method

Abstract

The invention discloses a marine intelligent cable release hook control system and method, and relates to the technical field of ship control, and the system comprises a data collection module, a spectrum analysis module, a strategy selection module, a blocking judgment module and a stability regulation and control module. According to the invention, through intelligent data acquisition, analysis and strategy decision, a complex and changeable sea condition environment is effectively dealt with, efficient, stable and accurate control of a cable releasing hook action is realized, specifically through combination of wavelet analysis and particle swarm optimization, a sea condition change trend can be accurately analyzed and predicted, a cable releasing action strategy is highly matched with an actual environment, and a cable releasing effect is improved. Blind or lagged operation decisions are avoided, and the adaptability of the system to severe sea conditions is remarkably improved; real-time state feedback and dynamic regulation and control technologies are fused, fine control in the cable releasing action process is achieved, and the phenomenon of abnormal action or instability of the cable releasing hook is effectively avoided by means of dynamic electromagnetic structure release force adjustment and accurate blocked state detection.

Description

Technical Field

[0001] This invention relates to the field of ship control technology, specifically to a marine intelligent unmooring hook control system and method. Background Technology

[0002] During ship navigation and berthing, the release hook is one of the key safety devices. Its reliability, stability and response accuracy are directly related to the efficiency and safety of ship operations. Especially in harsh sea conditions, it is impossible to ensure that the release hook action is precisely matched with the actual environmental conditions, so as to avoid abnormal action or failure due to environmental disturbances.

[0003] Currently, most marine undock control technologies rely on manual experience-based decision-making or simple sensor signal feedback mechanisms. These technologies respond to general sea state changes through pre-set action strategies. This approach is relatively mature in terms of technology and has advantages such as simple structure and convenient implementation, and can basically meet the ship undock requirements under normal circumstances.

[0004] However, as the complexity and frequency of change in the marine operating environment continue to increase, existing methods often suffer from problems such as delayed strategic decision-making and large deviations in action response when facing high-intensity and rapidly changing complex sea conditions. This can easily lead to unstable or even obstructed unhooking actions, increasing equipment wear and operational risks, and reducing the efficiency and safety of ship operations. Summary of the Invention

[0005] To solve the above-mentioned technical problems, the present invention provides a marine intelligent release hook control system, comprising:

[0006] Data acquisition module S11: Collects sea state fluctuation data and cable load data of the environment where the detachment hook is located; establishes an environmental disturbance state sequence based on the sea state fluctuation data; and generates a dynamic load sequence based on the cable load data.

[0007] Spectrum analysis module S12: Performs Fast Fourier Transform on the dynamic load sequence to extract the spectral features of the sequence, including the dominant frequency spectral amplitude and spectral energy distribution; constructs a preparatory strategy library for the cable release hook response action based on the spectral features;

[0008] Strategy selection module S13: Inputs the environmental disturbance state sequence into a pre-established deep recurrent neural network model to generate an environmental state prediction sequence; selects the initial unhooking action strategy from the unhooking hook response action preparation strategy library based on the environmental state prediction sequence;

[0009] Obstruction determination module S14: Triggers the release action of the permanent magnet electromagnetic structure in the unhooking mechanism according to the initial action strategy of unhooking, collects the displacement and action torque at the moment of release, and determines whether the unhooking action is obstructed based on the displacement and action torque.

[0010] Stability control module S15: If the unhooking action is not obstructed, it acquires real-time cable load data and unhooking mechanism displacement data, and dynamically adjusts the release current of the permanent magnet electromagnetic structure to keep the unhooking mechanism moving smoothly.

[0011] Furthermore, the sea state fluctuation data includes changes in wave height, wave period, and ship attitude angle; the establishment of an environmental disturbance state sequence based on the sea state fluctuation data specifically includes:

[0012] Collect sea state fluctuation data within a preset time period;

[0013] Wave height and wave period are input into the wavelet analysis model for time-frequency decomposition;

[0014] The steps for constructing the wavelet analysis model are as follows:

[0015] Historical wave heights and wave periods are selected as training samples to construct an initial set of wavelet decomposition functions, which are defined by two key parameters: center frequency and bandwidth.

[0016] An adaptive particle swarm optimization algorithm is used to optimize the parameters of the wavelet decomposition function set;

[0017] The optimized set of wavelet decomposition functions is used to generate a wavelet analysis model.

[0018] Based on the optimized wavelet analysis model, the time-frequency decomposition results of the wave signal are obtained, and the time-frequency energy distribution characteristics are extracted, including the spectral energy distribution range of the wave signal at different frequencies and time periods, the main frequency region, and the frequency variation value.

[0019] Based on the time-frequency energy distribution characteristics and the changes in the ship's attitude angle, an environmental disturbance state sequence is constructed. The real-time attitude change data of the ship includes the roll angle, pitch angle, and bow angle.

[0020] Furthermore, the steps for optimizing the wavelet decomposition function set parameters using the adaptive particle swarm optimization algorithm include:

[0021] A particle swarm is randomly initialized, where each particle in the swarm represents a set of wavelet decomposition function parameters, including the center frequency and bandwidth of the wavelet, to form the initial particle swarm.

[0022] The time-frequency energy error of the training samples is used as the fitness function to evaluate the fitness value of each particle;

[0023] The particle velocity and position are adjusted based on the fitness value, and the iteration continues until the fitness reaches a predetermined convergence threshold, at which point the optimal parameters are output.

[0024] Furthermore, the steps for generating a dynamic load sequence based on cable load data are as follows:

[0025] Based on the cable tension sensor installed on the release hook bearing structure, the magnitude of the cable tension on the release hook is continuously measured in real time.

[0026] Set the sampling frequency to 5–20 Hz, continuously measure and store raw load data for 30 seconds to 2 minutes to obtain an initial cable load data sequence.

[0027] The initial load data sequence is denoised and outlier data points are corrected to obtain high-quality dynamic load data.

[0028] High-quality dynamic load data is normalized to generate high-quality normalized load values;

[0029] A dynamic load sequence is constructed based on multiple sets of high-quality normalized load values.

[0030] Furthermore, the specific content of constructing the unhooking hook response action preparation strategy library based on spectral characteristics includes:

[0031] The spectral characteristics of the dynamic load sequence are used as input variables and input into the nonlinear dynamic simulation model.

[0032] The mechanical response characteristics of the unhooking hook action under the corresponding spectral characteristics were obtained based on the nonlinear dynamic simulation model.

[0033] Generate a sequence of unhooking response action strategies based on the mechanical response characteristics;

[0034] A preliminary strategy library is constructed by mapping spectral features to action strategy sequences.

[0035] Furthermore, the steps for constructing the nonlinear dynamic simulation model are as follows:

[0036] The friction coefficient, elastic modulus, and motion constraints of the release hook mechanism were collected, and the nonlinear dynamic control equations for the release hook action were constructed.

[0037] Based on the actual motion characteristics of the release hook mechanism, a generalized coordinate system for the release hook mechanism is defined.

[0038] Based on the defined generalized coordinates, the kinetic energy of the unmooring hook mechanism and the potential energy of the elastic component of the unmooring hook mechanism are calculated; based on the obtained friction coefficient, the friction torque is calculated, and the dynamic equation of the unmooring hook mechanism is constructed according to the kinetic energy of the unmooring hook mechanism, the potential energy of the elastic component of the unmooring hook mechanism, and the friction torque.

[0039] The control equations are solved using the finite element numerical method to construct a nonlinear dynamic simulation model.

[0040] The steps for constructing a nonlinear dynamic simulation model are as follows:

[0041] A finite element geometric model of the release hook mechanism is established; the obtained friction coefficient, elastic modulus and motion constraints are set as parameters and boundary conditions for finite element simulation; the established nonlinear control equations are numerically solved using finite element simulation software to obtain detailed dynamic response data of the release hook mechanism under spectral characteristic input conditions, so as to construct a complete nonlinear dynamic simulation model.

[0042] Furthermore, the steps for generating a deep recurrent neural network model include:

[0043] Historical environmental prediction data is acquired and divided into a first test set and a first training set. The historical environmental prediction data includes environmental disturbance state sequences and corresponding environmental state prediction sequences.

[0044] Construct a regression network by taking the environmental perturbation state sequence in the first training set as the input of the regression network and the corresponding environmental state prediction sequence as the output of the regression network, and train the regression network to obtain the initial deep recurrent neural network.

[0045] The initial deep recurrent neural network model is validated using the first test set. The initial deep recurrent neural network whose output is less than or equal to the preset test error threshold is used as the deep recurrent neural network model.

[0046] The logic for selecting the initial unhooking action strategy from the unhooking response action preparation strategy library based on the environmental state prediction sequence is as follows:

[0047] The predicted wave time-frequency energy distribution characteristics and predicted ship attitude change data in the environmental state prediction sequence are subjected to joint time-domain and frequency-domain processing to determine their corresponding environmental state prediction feature vectors.

[0048] Using the obtained environmental state prediction feature vector as the input feature, the stored spectral feature vector that is closest to the current input feature vector is searched in the unhooking hook response action preparation strategy library.

[0049] Determine the action policy sequence corresponding to the spectral feature vector with the highest similarity, and select the first decoupling action policy from this action policy sequence as the initial decoupling action policy.

[0050] Furthermore, the release action of the permanent magnet electromagnetic structure is achieved by eliminating the attraction force of the permanent magnet structure through a transient current pulse; the process of determining whether the disengagement action is obstructed based on the displacement and action torque specifically includes:

[0051] A feature vector for determining motion sluggishness is constructed based on the displacement and action torque at the moment of release of the release hook mechanism;

[0052] The motion blockage determination feature vector is input into a preset blocked classification model for classification to obtain a determination label, which includes a motion blockage marker and a motion unblocked marker.

[0053] Based on the action obstruction flag, a policy correction scheme is triggered, which is as follows:

[0054] Based on the current environmental state prediction sequence and the action obstruction marker, the strategy selection module S13 is repeatedly executed to match a suitable decoupling initial action strategy.

[0055] The steps for generating the hindered classification model include:

[0056] Obtain historical obstruction judgment data, and divide the historical obstruction judgment data into a second training set and a second test set. The historical obstruction judgment data includes action obstruction judgment feature vectors and their corresponding judgment labels.

[0057] Configure the initial classifier by using the action retardation judgment feature vector in the second training set as the input data of the initial classifier and the corresponding judgment label in the second training set as the output data of the initial classifier. Train the initial classifier to obtain the initial classification network.

[0058] The initial classification network is validated using a second test set. The output of the initial classification network with an accuracy greater than or equal to the preset test accuracy is used as a pre-built hindered classification model.

[0059] Furthermore, the logic for dynamically adjusting the release force of the permanent magnet electromagnetic structure to maintain the smooth operation of the cable release mechanism is as follows:

[0060] Calculate the load deviation between the real-time cable load data and the preset target load data, and calculate the displacement deviation between the displacement data of the cable release mechanism and the preset target displacement data.

[0061] Determine whether the absolute values ​​of the load deviation and displacement deviation exceed the preset deviation threshold. If the load deviation is greater than the preset positive load deviation threshold or the displacement deviation is greater than the preset positive displacement deviation threshold, then reduce the release current.

[0062] If the load deviation value is less than the preset negative load deviation threshold, or the displacement deviation value is less than the preset negative displacement deviation threshold, the release current will be increased.

[0063] A method for controlling a marine intelligent release hook, applied to any of the marine intelligent release hook control systems described in the preceding claims, characterized in that the method comprises:

[0064] S21: Collect sea state fluctuation data and cable load data of the environment where the detachment hook is located; establish an environmental disturbance state sequence based on the sea state fluctuation data, and generate a dynamic load sequence based on the cable load data;

[0065] S22: Perform a Fast Fourier Transform on the dynamic load sequence to extract the spectral features of the sequence, including the dominant frequency spectral amplitude and spectral energy distribution; construct a preparatory strategy library for the uncoupling hook response action based on the spectral features;

[0066] S23: Input the environmental disturbance state sequence into a pre-established deep recurrent neural network model to generate an environmental state prediction sequence; select the initial unhooking action strategy from the unhooking response action preparation strategy library based on the environmental state prediction sequence;

[0067] S24: Trigger the release action of the permanent magnet electromagnetic structure in the unhooking mechanism according to the initial action strategy of unhooking, collect the displacement and action torque at the moment of release, and determine whether the unhooking action is blocked based on the displacement and action torque. If the unhooking action is blocked, then feed back to correct the initial action strategy of unhooking and return to step S23.

[0068] S25: If the unhooking action is not obstructed, real-time cable load data and unhooking mechanism displacement data are obtained, and the release current of the permanent magnet electromagnetic structure is dynamically adjusted to keep the unhooking mechanism moving smoothly.

[0069] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0070] This invention effectively addresses complex and ever-changing sea conditions through intelligent data acquisition, analysis, and strategy decision-making. It achieves efficient, stable, and precise control of the release hook action. By combining wavelet analysis with particle swarm optimization, it can accurately analyze and predict sea condition trends, ensuring that the release hook action strategy is highly matched with the actual environment. This avoids blind or delayed operational decisions and significantly improves the system's adaptability to severe sea conditions.

[0071] Furthermore, by integrating real-time status feedback and dynamic control technology, precise control is achieved during the uncoupling process. With the help of dynamic electromagnetic structure release force adjustment and accurate obstruction detection, abnormal or unstable uncoupling hook movements are effectively avoided, significantly improving the reliability and service life of the uncoupling mechanism, thereby enhancing the safety and efficiency of ship operations. Attached Figure Description

[0072] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0073] Figure 1 This is a block diagram of a marine intelligent unmooring hook control system provided in Embodiment 1 of the present invention;

[0074] Figure 2 The flowchart shows a marine intelligent unmooring hook control method provided in Embodiment 2 of the present invention. Detailed Implementation

[0075] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0076] Example 1

[0077] Please see Figure 1 As shown in the figure, this embodiment discloses a marine intelligent unmooring hook control system, the system comprising:

[0078] Data acquisition module S11: Collects sea state fluctuation data and cable load data of the environment where the detachment hook is located; establishes an environmental disturbance state sequence based on the sea state fluctuation data; and generates a dynamic load sequence based on the cable load data.

[0079] Specifically, the sea state fluctuation data includes changes in wave height, wave period, and ship attitude angle; the establishment of an environmental disturbance state sequence based on the sea state fluctuation data specifically includes:

[0080] S111: Collect sea state fluctuation data within a preset time period;

[0081] It should be noted that: a shipborne wave altimeter is used to measure wave height in real time; a period sensor is used to measure wave period in real time; and an inertial navigation system (INS) is used to measure changes in hull attitude angles in real time, including real-time angle data for the three axes of roll, pitch, and bow.

[0082] The wave height, wave period, and ship attitude angle changes are all collected and stored in real time at a specified sampling frequency (e.g., 5Hz to 10Hz);

[0083] S112: Input wave height and wave period into the wavelet analysis model for time-frequency decomposition;

[0084] Specifically, the steps for constructing the wavelet analysis model are as follows:

[0085] S112.1: Select historical wave heights and wave periods as training samples to construct an initial set of wavelet decomposition functions. The wavelet decomposition functions are defined by two key parameters: center frequency and bandwidth.

[0086] It should be noted that the wavelet decomposition function is a special mathematical waveform, shaped like an oscillating wave packet. It can effectively detect changes in wave signals within different time and frequency ranges. Simply put, this function can decompose complex wave signals according to different frequencies and time periods, so as to clearly know the intensity and frequency distribution of waves within different time periods.

[0087] S112.2: The parameters of the wavelet decomposition function set are optimized using an adaptive particle swarm optimization algorithm;

[0088] Specifically, the optimization of wavelet decomposition function set parameters using the adaptive particle swarm optimization algorithm includes:

[0089] A particle swarm is randomly initialized, where each particle in the swarm represents a set of wavelet decomposition function parameters, including the center frequency and bandwidth of the wavelet, to form the initial particle swarm.

[0090] The time-frequency energy error of the training samples is used as the fitness function to evaluate the fitness value of each particle. The fitness function is expressed as follows:

[0091]

[0092] Where β is the fitness function, E real,n E represents the actual time-frequency energy value of the nth historical sample point. pred,n The wavelet prediction time-frequency energy value for the nth sample point;

[0093] The particle velocity and position are adjusted based on the fitness value. The particle velocity update formula is expressed as follows:

[0094] v i (t+1)=w·v i (t)+c1·r1·[pbest i (t)-x i (t)]+c2·r2·[gbest(t)-x i (t)];

[0095] In the formula, v i (t) represents the velocity of the i-th particle in the t-th iteration, w is the preset inertia weight factor, c1 and c2 are acceleration factors for the particle swarm optimization algorithm, and r1 and r2 are random numbers in the interval [0,1] used to improve the algorithm's random search capability. pbest igbest(t) represents the historical best position found by the i-th particle up to the t-th iteration, and gbest(t) represents the global best position found by the entire particle swarm up to the t-th iteration.

[0096] It should be noted that w is set by the experimenters based on historical data to balance the global search capability and local search capability of the algorithm. The acceleration factor is the learning factor, which represents the weight of the individual particle's own experience and the weight of the group's overall experience, respectively.

[0097] The particle position update formula is expressed as:

[0098] x i (t+1)=x i (t)+v i (t+1);

[0099] x i (t) represents the position of the i-th particle in the t-th iteration;

[0100] Iterate until the fitness reaches a predetermined convergence threshold, then output the optimal parameters;

[0101] S112.3: Use the optimized set of wavelet decomposition functions to generate a wavelet analysis model;

[0102] S113: Based on the optimized wavelet analysis model, the time-frequency decomposition results of the wave signal are obtained, and the time-frequency energy distribution characteristics are extracted, including the spectral energy distribution range of the wave signal at different frequencies and time periods, the main frequency region, and the frequency variation value.

[0103] S114: Construct an environmental disturbance state sequence based on the time-frequency energy distribution characteristics and the changes in the ship's attitude angle. The real-time attitude change data of the ship includes the roll angle, pitch angle, and bow angle. The environmental disturbance state sequence is represented as follows:

[0104] S env (t)={E tf (t),θ roll (t),θ pitch (t),θ yaw (t)};

[0105] E tf (t) represents the wave time-frequency energy distribution characteristics, θ roll (t),θ pitch (t),θ yaw (t) represent the changes in roll angle, pitch angle, and bow angle, respectively;

[0106] Specifically, the steps for generating a dynamic load sequence based on cable load data are as follows:

[0107] Based on the cable tension sensor installed on the release hook bearing structure, the magnitude of the cable tension on the release hook is continuously measured in real time.

[0108] It should be noted that cable tension sensors include, but are not limited to, strain gauge and pressure sensors.

[0109] Set the sampling frequency to 5–20 Hz, continuously measure and store raw load data for 30 seconds to 2 minutes to obtain an initial cable load data sequence.

[0110] The initial load data sequence is denoised and outlier data points are corrected to obtain high-quality dynamic load data.

[0111] It should be noted that: data denoising utilizes existing signal filtering methods (such as Butterworth low-pass filters) to remove high-frequency noise interference in order to obtain a smoother load signal; outlier correction employs median filtering or sliding window averaging filtering methods to correct outlier data points in the load data sequence, ensuring the accuracy and stability of the load data;

[0112] High-quality dynamic load data is normalized and represented as follows:

[0113]

[0114] Where L(t) represents the high-quality dynamic load data at time t, L max For maximum high-quality dynamic load data, L min For minimum high-quality dynamic load data, L norm (t) represents the high-quality normalized load value at time t;

[0115] A dynamic load sequence is constructed based on multiple sets of high-quality normalized load values;

[0116] The above steps automatically optimize the parameters of the wavelet decomposition function using an adaptive particle swarm optimization algorithm to adapt to the analysis of complex sea state wave characteristics. Compared with traditional fixed-parameter wavelet analysis or wavelet analysis with manually adjusted parameters, it has obvious advantages and can more accurately reflect the true characteristics of ocean waves, providing a high-quality data foundation for subsequent decoupling strategies.

[0117] Spectrum analysis module S12: Performs Fast Fourier Transform on the dynamic load sequence to extract the spectral features of the sequence, including the dominant frequency spectral amplitude and spectral energy distribution; constructs a preparatory strategy library for the cable release hook response action based on the spectral features;

[0118] It should be noted that: the dominant frequency is the main frequency component in the load change; the spectral amplitude is the magnitude of the amplitude corresponding to each frequency component; and the spectral energy distribution is the proportion of load change energy in each frequency range.

[0119] Specifically, the method for constructing a preparatory strategy library for unhooking response actions based on spectral characteristics includes:

[0120] S121: The spectral characteristics of the dynamic load sequence are used as input variables and input into the nonlinear dynamic simulation model;

[0121] Specifically, the steps for constructing the nonlinear dynamic simulation model are as follows:

[0122] S121.1: Collect the friction coefficient, elastic modulus and motion constraints of the release hook mechanism, and construct the nonlinear dynamic control equations for the release hook action;

[0123] It should be noted that: the friction coefficient is obtained through standard friction tests; the elastic modulus is obtained through standard material mechanics tests (such as tensile or compression tests); the motion constraint conditions are determined according to the design and installation of the release hook structure, including the constraint type and motion range of the rotary joint and sliding joint.

[0124] S121.2: Based on the actual motion characteristics of the release hook mechanism, define the generalized coordinate q of the release hook mechanism. i ;

[0125] It should be noted that: the so-called unmooring hook mechanism refers to the unmooring hook mechanical component, and the rotation angle of the rotating component of the unmooring hook mechanism is used as the generalized coordinate, or the displacement of the sliding component can be selected as the generalized coordinate;

[0126] Based on the defined generalized coordinates, the kinetic energy of the release hook mechanism is calculated and expressed as:

[0127]

[0128] In the formula, T is the kinetic energy of the release hook mechanism, and J is the given moment of inertia. The angular velocity of the turning angle;

[0129] It should be noted that the moment of inertia J is determined by the mechanical design parameters of the release hook mechanism, such as material density and geometric dimensions (e.g., cross-sectional area, length, mass distribution).

[0130] The potential energy of the elastic component of the release hook mechanism is calculated and expressed as:

[0131]

[0132] In the formula, V is the potential energy of the release hook mechanism, k is the given elastic stiffness coefficient, and x is the elastic variable of the elastic component.

[0133] It should be noted that: elastic components include springs or beams, and k is determined based on the elastic modulus;

[0134] Based on the obtained friction coefficient, the friction torque is calculated and expressed as:

[0135] Q i =-μNr

[0136] Among them, Q i Let be the frictional torque corresponding to the i-th generalized coordinate, μ be the obtained friction coefficient, N be the normal load value, and r be the effective frictional force radius.

[0137] It should be noted that: when measuring the friction coefficient or simulating actual working conditions, a normal force is applied to the contact surface of the release hook using a loading device; a force sensor (such as a pressure sensor or strain gauge force gauge) is installed to record the magnitude of the applied normal force; the data output by the sensor is recorded by the data acquisition system and converted into a specific normal load value;

[0138] r refers to the actual effective distance of the frictional force acting on the contact surface, which is usually determined based on the geometry of the unhooking mechanism and the actual distribution of the contact surface;

[0139] Construct the dynamic equations of the uncoupling hook mechanism:

[0140]

[0141] Where L is the Lagrange function, L = TV; Let the Lagrangian function L be used with respect to the generalized coordinate q. i For partial derivatives, For the Lagrangian function L with respect to the generalized velocity The rate of change of the partial derivative with time, Let be the derivative of the i-th generalized coordinate with respect to time;

[0142] It should be noted that: generalized coordinate q i Used to describe the position, angle, and other degrees of freedom of motion of the uncoupling hook mechanism. This indicates the velocity of the generalized coordinate system.

[0143] S121.3: Solve the control equations using the finite element numerical method to construct a nonlinear dynamic simulation model;

[0144] Specifically, the steps for constructing a nonlinear dynamic simulation model are as follows:

[0145] A finite element geometric model of the release hook mechanism is established; the obtained friction coefficient, elastic modulus and motion constraints are set as parameters and boundary conditions for finite element simulation; the established nonlinear control equations are numerically solved using finite element simulation software to obtain detailed dynamic response data of the release hook mechanism under spectral characteristic input conditions, so as to construct a complete nonlinear dynamic simulation model.

[0146] It should be noted that finite element simulation software includes, but is not limited to, ABAQUS and ANSYS;

[0147] S122: The mechanical response characteristics of the unhooking hook action under the corresponding frequency spectrum characteristics are obtained by simulation based on the nonlinear dynamic simulation model;

[0148] The dominant frequency, spectral amplitude, and spectral energy distribution in the spectral characteristics are used as the excitation inputs to the nonlinear dynamic simulation model; the simulation load input signal is constructed using the dominant frequency, and is expressed as:

[0149] F(t) = A f sin(2πf dom t)

[0150] Where F(t) is the simulated load input signal, t is time, and f is the input signal. dom As the dominant frequency, A f To be related to the dominant frequency f dom The corresponding spectral amplitude;

[0151] Based on the spectral energy distribution, an auxiliary excitation input signal is constructed for the secondary frequencies adjacent to the dominant frequency, which serves as an external excitation.

[0152] The simulation duration was set to 30s to 120s, and the time step was set to 0.01s to 0.1s. Nonlinear dynamic finite element simulation was run to obtain the motion trajectory data of the release hook mechanism response.

[0153] Extract the mechanical response characteristics of the unhooking mechanism from the simulation results, including the initial moment of the unhooking threshold, the duration of the unhooking action, and the peak torque of the unhooking mechanism;

[0154] It should be noted that: the initial moment when the unhooking mechanism reaches the preset unhooking threshold is determined based on the displacement or angle data; the duration of the unhooking action is determined based on the time required for the unhooking mechanism to complete the unhooking process from the start of the action in the simulation results; and the peak torque of the unhooking mechanism is determined based on the peak torque value experienced by the unhooking mechanism during the unhooking process.

[0155] The beneficial effects of steps S121 and S122 are as follows:

[0156] By collecting the friction coefficient, elastic modulus, and motion constraints of the uncoupling mechanism, a nonlinear dynamic simulation model is constructed. Using the spectral characteristics of the dynamic load sequence as input, the specific action response process of the uncoupling mechanism under complex dynamic loads is accurately simulated. The initial triggering time of the uncoupling threshold, the duration of the uncoupling action, and the peak torque borne by the mechanism during the uncoupling process are clearly obtained. This avoids the problem that traditional empirical methods cannot accurately describe nonlinear friction, elasticity, and constraints, and provides data support for the formulation of subsequent uncoupling action strategies, thereby improving the safety and general applicability of the uncoupling control.

[0157] S123: Generate a sequence of disengagement response action strategies based on mechanical response characteristics;

[0158] Based on the initial moment of the unhooking threshold, the duration of the unhooking action, and the peak torque of the unhooking mechanism, a unhooking response action strategy is constructed, expressed as:

[0159] P k ={t start,k ,t dur,k M act,k}

[0160] P k For the k-th decoupling response action strategy, t start,k Starting time point t dur,k M represents the duration of the action. act,k This represents the peak torque of the release hook mechanism;

[0161] For each set of spectral characteristics, repeat the above steps to generate a complete sequence of decoupling response action strategies.

[0162] S124: A preliminary strategy library is constructed by mapping spectral features to action strategy sequences;

[0163] Construct a policy library data structure, using spectral features as the mapping input and the corresponding decoupling response action policy as the mapping output, to establish a one-to-one mapping relationship between spectral features and decoupling response action policies. The spectral features as the mapping input are represented as follows:

[0164] f = {f dom A f E dist}

[0165] Where f represents the spectral characteristics of the mapped input, f dom As the dominant frequency, A f E represents the spectral amplitude corresponding to the dominant frequency. dist Characteristics of spectral energy distribution;

[0166] The decoupling response strategy for the mapped output is represented as follows:

[0167] P k ={t start,k ,t dur,k M act,k}

[0168] Based on the multidimensional interpolation algorithm, the spectral features formed by all simulation data and the corresponding action strategies are used as training data to fit and establish a nonlinear mapping relationship from spectral features to decoupling response action strategies.

[0169] It should be noted that multidimensional interpolation methods include, but are not limited to, radial basis function interpolation (RBF interpolation) and cubic spline interpolation;

[0170] The acquired nonlinear mapping relationships are stored in a computer database to build a complete library of preparatory strategies for unhooking response actions;

[0171] The beneficial effects of steps S123 and S124 are as follows:

[0172] By establishing a mapping relationship between load spectrum characteristics and decoupling action strategies, a pre-optimized strategy library is formed, enabling rapid matching of dynamic load conditions and timely provision of corresponding decoupling strategies in actual work, avoiding response delays and errors caused by real-time calculations. At the same time, by using a multi-dimensional interpolation algorithm to fit the nonlinear relationship between spectrum characteristics and action strategies, the adaptability of the strategy library to unknown load conditions is improved, effectively expanding the applicability and control accuracy of the system under complex dynamic conditions.

[0173] Strategy selection module S13: Inputs the environmental disturbance state sequence into a pre-established deep recurrent neural network model to generate an environmental state prediction sequence; selects the initial unhooking action strategy from the unhooking hook response action preparation strategy library based on the environmental state prediction sequence;

[0174] Specifically, the steps for generating a deep recurrent neural network model include:

[0175] Historical environmental prediction data is acquired and divided into a first test set and a first training set. The historical environmental prediction data includes environmental disturbance state sequences and corresponding environmental state prediction sequences.

[0176] It should be noted that the environmental state prediction sequence is the wave time-frequency energy distribution characteristics and ship attitude change data for a future time period;

[0177] Construct a regression network by taking the environmental perturbation state sequence in the first training set as the input of the regression network and the corresponding environmental state prediction sequence as the output of the regression network, and train the regression network to obtain the initial deep recurrent neural network.

[0178] The initial deep recurrent neural network model is validated using the first test set. The initial deep recurrent neural network whose output is less than or equal to the preset test error threshold is used as the deep recurrent neural network model.

[0179] Specifically, the logic for selecting the initial unhooking action strategy from the unhooking response action preparation strategy library based on the environmental state prediction sequence is as follows:

[0180] The predicted wave time-frequency energy distribution characteristics and predicted ship attitude change data in the environmental state prediction sequence are subjected to joint time-domain and frequency-domain processing to determine their corresponding environmental state prediction feature vectors.

[0181] Using the obtained environmental state prediction feature vector as the input feature, the stored spectral feature vector that is closest to the current input feature vector is searched in the unhooking hook response action preparation strategy library.

[0182] It should be noted that the similarity between the feature vectors is achieved using a distance metric algorithm, specifically including but not limited to Euclidean distance, cosine similarity, or Mahalanobis distance.

[0183] Determine the action policy sequence corresponding to the spectral feature vector with the highest similarity, and select the first decoupling action policy from this action policy sequence as the initial decoupling action policy.

[0184] Obstruction determination module S14: Triggers the release action of the permanent magnet electromagnetic structure in the unhooking mechanism according to the initial action strategy of unhooking, collects the displacement and action torque at the moment of release, and determines whether the unhooking action is obstructed based on the displacement and action torque.

[0185] It should be noted that the displacement data is acquired by a displacement sensor installed on the moving parts of the uncoupling hook mechanism. The specific sensor types include, but are not limited to, linear differential transformers (LVDTs) or grating displacement sensors.

[0186] The action torque data is obtained in real time by a torque sensor installed on the rotating shaft or bearing component of the uncoupling hook mechanism. The torque sensor type includes, but is not limited to, strain gauge torque sensor and piezoelectric torque sensor.

[0187] Specifically, the release action of the permanent magnet electromagnetic structure is achieved by eliminating the attraction force of the permanent magnet structure through a transient current pulse; the process of determining whether the disengagement action is obstructed based on the displacement and action torque specifically includes:

[0188] S141: Construct a feature vector for determining motion sluggishness based on the displacement and action torque at the instant of release of the release hook mechanism;

[0189] Represented as:

[0190] X b =[Δd,M act ] T

[0191] In the formula, X b Let M be the feature vector for determining motion lag, Δd be the displacement change value of the release hook mechanism at the instant of release, and M be the feature vector for determining motion lag. act This represents the instantaneous torque applied when the release hook mechanism is engaged; the superscript T indicates the transpose of the vector.

[0192] It should be noted that: Δd represents the displacement increment from the start of the release hook mechanism's action to the current measurement time, and M... act This represents the real-time measured axial torque value of the uncoupling mechanism at the instant of release during the uncoupling action. The superscript T indicates the transpose of the vector, that is, the conversion of the row vector into a column vector.

[0193] S142: Input the motion blockage determination feature vector into the preset blockage classification model for classification and determination to obtain determination labels, the determination labels including motion blockage markers and motion unblocked markers;

[0194] Specifically, the steps for generating the hindered classification model include:

[0195] Obtain historical obstruction judgment data, and divide the historical obstruction judgment data into a second training set and a second test set. The historical obstruction judgment data includes action obstruction judgment feature vectors and their corresponding judgment labels.

[0196] Configure the initial classifier by using the action retardation judgment feature vector in the second training set as the input data of the initial classifier and the corresponding judgment label in the second training set as the output data of the initial classifier. Train the initial classifier to obtain the initial classification network.

[0197] The initial classification network is validated using a second test set. The output of the initial classification network with an accuracy greater than or equal to the preset test accuracy is used as a pre-built hindered classification model.

[0198] It should be noted that the hindered classification model can be a support vector machine;

[0199] S143: Based on the action obstruction flag, trigger a policy correction scheme, wherein the policy correction scheme is as follows:

[0200] Based on the current environment state prediction sequence and the action obstruction marker, the strategy selection module S13 is repeatedly executed to match a suitable decoupling initial action strategy.

[0201] Stability control module S15: If the unhooking action is not obstructed, it acquires real-time cable load data and unhooking mechanism displacement data, and dynamically adjusts the release current of the permanent magnet electromagnetic structure to keep the unhooking mechanism moving smoothly.

[0202] Specifically, the logic for dynamically adjusting the release force of the permanent magnet electromagnetic structure to maintain the smooth operation of the cable release mechanism is as follows:

[0203] Calculate the load deviation between the real-time cable load data and the preset target load data, and calculate the displacement deviation between the displacement data of the cable release mechanism and the preset target displacement data.

[0204] Represented as:

[0205] ΔL=L real (τ)-L ref

[0206] ΔS=S real (τ)-S ref

[0207] In the formula, ΔL is the load deviation value, L real (τ) represents the real-time cable load data at time τ, L ref The target load data is preset, ΔS is the displacement deviation value, and S is the displacement deviation value. real (τ) represents the displacement data of the release hook mechanism measured in real time, S ref Preset target displacement data;

[0208] Determine whether the absolute values ​​of the load deviation and displacement deviation exceed the preset deviation threshold. If the load deviation is greater than the preset positive load deviation threshold or the displacement deviation is greater than the preset positive displacement deviation threshold, then reduce the release current.

[0209] If the load deviation value is less than the preset negative load deviation threshold, or the displacement deviation value is less than the preset negative displacement deviation threshold, the release current is increased.

[0210] In the execution steps of the strategy selection module S13 and the obstruction determination module S14, a deep recurrent neural network is used to predict the real-time collected environmental disturbance state sequence, achieving an accurate estimate of future environmental change trends. The prediction results are then quickly matched with a pre-built action strategy library, effectively improving the foresight and accuracy of the release hook action decision. Furthermore, based on the obstruction determination model, the system monitors in real-time whether the release action of the release hook mechanism is obstructed, enabling the system to quickly identify anomalies and promptly trigger the action strategy correction process, significantly reducing response delays under abnormal conditions. Simultaneously, in the absence of obstruction, the release current is dynamically adjusted to correct load and displacement deviations in real time, achieving refined control of the release hook mechanism's actions and ensuring the stability, reliability, and smoothness of the mechanism's operation.

[0211] Example 2

[0212] Please see Figure 2As shown, based on a unified inventive concept, this embodiment discloses a marine intelligent release hook control method, including:

[0213] S21: Collect sea state fluctuation data and cable load data of the environment where the detachment hook is located; establish an environmental disturbance state sequence based on the sea state fluctuation data, and generate a dynamic load sequence based on the cable load data;

[0214] S22: Perform a Fast Fourier Transform on the dynamic load sequence to extract the spectral features of the sequence, including the dominant frequency spectral amplitude and spectral energy distribution; construct a preparatory strategy library for the uncoupling hook response action based on the spectral features;

[0215] S23: Input the environmental disturbance state sequence into a pre-established deep recurrent neural network model to generate an environmental state prediction sequence; select the initial unhooking action strategy from the unhooking response action preparation strategy library based on the environmental state prediction sequence;

[0216] S24: Trigger the release action of the permanent magnet electromagnetic structure in the unhooking mechanism according to the initial action strategy of unhooking, collect the displacement and action torque at the moment of release, and determine whether the unhooking action is blocked based on the displacement and action torque. If the unhooking action is blocked, then feed back to correct the initial action strategy of unhooking and return to step S23.

[0217] S25: If the unhooking action is not obstructed, real-time cable load data and unhooking mechanism displacement data are obtained, and the release current of the permanent magnet electromagnetic structure is dynamically adjusted to keep the unhooking mechanism moving smoothly.

[0218] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A marine intelligent release hook control system, characterized in that, The system includes: Data acquisition module S11: Collects sea state fluctuation data and cable load data of the environment where the detachment hook is located; establishes an environmental disturbance state sequence based on the sea state fluctuation data; and generates a dynamic load sequence based on the cable load data. Spectrum analysis module S12: Performs Fast Fourier Transform on the dynamic load sequence to extract the spectral features of the sequence, including the dominant frequency spectral amplitude and spectral energy distribution; constructs a preparatory strategy library for the cable release hook response action based on the spectral features; Strategy selection module S13: Inputs the environmental disturbance state sequence into a pre-established deep recurrent neural network model to generate an environmental state prediction sequence; selects the initial unhooking action strategy from the unhooking hook response action preparation strategy library based on the environmental state prediction sequence; Obstruction determination module S14: Triggers the release action of the permanent magnet electromagnetic structure in the unhooking mechanism according to the initial action strategy of unhooking, collects the displacement and action torque at the moment of release, and determines whether the unhooking action is obstructed based on the displacement and action torque. Stability control module S15: If the unhooking action is not obstructed, it acquires real-time cable load data and unhooking mechanism displacement data, and dynamically adjusts the release current of the permanent magnet electromagnetic structure to keep the unhooking mechanism moving smoothly.

2. The marine intelligent release hook control system according to claim 1, characterized in that, The sea state fluctuation data includes changes in wave height, wave period, and ship attitude angle; the establishment of an environmental disturbance state sequence based on the sea state fluctuation data specifically includes: Collect sea state fluctuation data within a preset time period; Wave height and wave period are input into the wavelet analysis model for time-frequency decomposition; The steps for constructing the wavelet analysis model are as follows: Historical wave heights and wave periods are selected as training samples to construct an initial set of wavelet decomposition functions, which are defined by two key parameters: center frequency and bandwidth. An adaptive particle swarm optimization algorithm is used to optimize the parameters of the wavelet decomposition function set; The optimized set of wavelet decomposition functions is used to generate a wavelet analysis model. Based on the optimized wavelet analysis model, the time-frequency decomposition results of the wave signal are obtained, and the time-frequency energy distribution characteristics are extracted, including the spectral energy distribution range of the wave signal at different frequencies and time periods, the main frequency region, and the frequency variation value. Based on the time-frequency energy distribution characteristics and the changes in the ship's attitude angle, an environmental disturbance state sequence is constructed. The real-time attitude change data of the ship includes the roll angle, pitch angle, and bow angle.

3. The marine intelligent release hook control system according to claim 2, characterized in that, The steps for optimizing the wavelet decomposition function set parameters using the adaptive particle swarm optimization algorithm include: A particle swarm is randomly initialized, where each particle in the swarm represents a set of wavelet decomposition function parameters, including the center frequency and bandwidth of the wavelet, to form the initial particle swarm. The time-frequency energy error of the training samples is used as the fitness function to evaluate the fitness value of each particle; The particle velocity and position are adjusted based on the fitness value, and the iteration continues until the fitness reaches a predetermined convergence threshold, at which point the optimal parameters are output.

4. A marine intelligent release hook control system according to claim 3, characterized in that, The steps for generating a dynamic load sequence based on cable load data are as follows: Based on the cable tension sensor installed on the release hook bearing structure, the magnitude of the cable tension on the release hook is continuously measured in real time. Set the sampling frequency to 5–20 Hz, continuously measure and store raw load data for 30 seconds to 2 minutes to obtain an initial cable load data sequence. The initial load data sequence is denoised and outlier data points are corrected to obtain high-quality dynamic load data. High-quality dynamic load data is normalized to generate high-quality normalized load values; A dynamic load sequence is constructed based on multiple sets of high-quality normalized load values.

5. A marine intelligent release hook control system according to claim 4, characterized in that, The aforementioned strategy library for constructing a release hook response action based on spectral characteristics specifically includes: The spectral characteristics of the dynamic load sequence are used as input variables and input into the nonlinear dynamic simulation model. The mechanical response characteristics of the unhooking hook action under the corresponding spectral characteristics were obtained based on the nonlinear dynamic simulation model. Generate a sequence of unhooking response action strategies based on the mechanical response characteristics; A preliminary strategy library is constructed by mapping spectral features to action strategy sequences.

6. A marine intelligent release hook control system according to claim 5, characterized in that, The steps for constructing the nonlinear dynamic simulation model are as follows: The friction coefficient, elastic modulus, and motion constraints of the release hook mechanism were collected, and the nonlinear dynamic control equations for the release hook action were constructed. Based on the actual motion characteristics of the release hook mechanism, a generalized coordinate system for the release hook mechanism is defined. Based on the defined generalized coordinates, the kinetic energy of the unmooring hook mechanism and the potential energy of the elastic component of the unmooring hook mechanism are calculated; based on the obtained friction coefficient, the friction torque is calculated, and the dynamic equation of the unmooring hook mechanism is constructed according to the kinetic energy of the unmooring hook mechanism, the potential energy of the elastic component of the unmooring hook mechanism, and the friction torque. The control equations are solved using the finite element numerical method to construct a nonlinear dynamic simulation model; The steps for constructing a nonlinear dynamic simulation model are as follows: A finite element geometric model of the release hook mechanism is established; the obtained friction coefficient, elastic modulus and motion constraints are set as parameters and boundary conditions for finite element simulation; the established nonlinear control equations are numerically solved using finite element simulation software to obtain detailed dynamic response data of the release hook mechanism under spectral characteristic input conditions, so as to construct a complete nonlinear dynamic simulation model.

7. A marine intelligent release hook control system according to claim 6, characterized in that, The steps for generating a deep recurrent neural network model include: Historical environmental prediction data is acquired and divided into a first test set and a first training set. The historical environmental prediction data includes environmental disturbance state sequences and corresponding environmental state prediction sequences. Construct a regression network by taking the environmental perturbation state sequence in the first training set as the input of the regression network and the corresponding environmental state prediction sequence as the output of the regression network, and train the regression network to obtain the initial deep recurrent neural network. The initial deep recurrent neural network model is validated using the first test set. The initial deep recurrent neural network whose output is less than or equal to the preset test error threshold is used as the deep recurrent neural network model. The logic for selecting the initial unhooking action strategy from the unhooking response action preparation strategy library based on the environmental state prediction sequence is as follows: The predicted wave time-frequency energy distribution characteristics and predicted ship attitude change data in the environmental state prediction sequence are subjected to joint time-domain and frequency-domain processing to determine their corresponding environmental state prediction feature vectors. Using the obtained environmental state prediction feature vector as the input feature, the stored spectral feature vector that is closest to the current input feature vector is searched in the unhooking hook response action preparation strategy library. Determine the action policy sequence corresponding to the spectral feature vector with the highest similarity, and select the first decoupling action policy from this action policy sequence as the initial decoupling action policy.

8. A marine intelligent release hook control system according to claim 7, characterized in that, The release action of the permanent magnet electromagnetic structure is achieved by eliminating the attraction force of the permanent magnet structure through a transient current pulse; the process of determining whether the unhooking action is obstructed based on the displacement and the action torque specifically includes: A feature vector for determining motion sluggishness is constructed based on the displacement and action torque at the instant of release of the release hook mechanism; The motion blockage determination feature vector is input into a preset blocked classification model for classification to obtain a determination label, which includes a motion blockage marker and a motion unblocked marker. Based on the action obstruction flag, a policy correction scheme is triggered, which is as follows: Based on the current environmental state prediction sequence and the action obstruction marker, the strategy selection module S13 is repeatedly executed to match a suitable decoupling initial action strategy. The steps for generating the hindered classification model include: Obtain historical obstruction judgment data, and divide the historical obstruction judgment data into a second training set and a second test set. The historical obstruction judgment data includes action obstruction judgment feature vectors and their corresponding judgment labels. Configure the initial classifier by using the action retardation judgment feature vector in the second training set as the input data of the initial classifier and the corresponding judgment label in the second training set as the output data of the initial classifier. Train the initial classifier to obtain the initial classification network. The initial classification network is validated using a second test set. The output of the initial classification network with an accuracy greater than or equal to the preset test accuracy is used as a pre-built hindered classification model.

9. A marine intelligent release hook control system according to claim 8, characterized in that, The logic for dynamically adjusting the release force of the permanent magnet electromagnetic structure to maintain the smooth operation of the release hook mechanism is as follows: Calculate the load deviation between the real-time cable load data and the preset target load data, and calculate the displacement deviation between the release hook mechanism displacement data and the preset target displacement data; Determine whether the absolute values ​​of the load deviation and displacement deviation exceed the preset deviation threshold. If the load deviation is greater than the preset positive load deviation threshold or the displacement deviation is greater than the preset positive displacement deviation threshold, then reduce the release current. If the load deviation value is less than the preset negative load deviation threshold, or the displacement deviation value is less than the preset negative displacement deviation threshold, the release current will be increased.

10. A marine intelligent release hook control method, applied to the marine intelligent release hook control system according to any one of claims 1-9, characterized in that, The method includes: S21: Collect sea state fluctuation data and cable load data of the environment where the detachment hook is located; establish an environmental disturbance state sequence based on the sea state fluctuation data, and generate a dynamic load sequence based on the cable load data; S22: Perform a Fast Fourier Transform on the dynamic load sequence to extract the spectral features of the sequence, including the dominant frequency spectral amplitude and spectral energy distribution; construct a preparatory strategy library for the uncoupling hook response action based on the spectral features; S23: Input the environmental disturbance state sequence into a pre-established deep recurrent neural network model to generate an environmental state prediction sequence; select the initial unhooking action strategy from the unhooking response action preparation strategy library based on the environmental state prediction sequence; S24: Trigger the release action of the permanent magnet electromagnetic structure in the unhooking mechanism according to the initial action strategy of unhooking, collect the displacement and action torque at the moment of release, and determine whether the unhooking action is blocked based on the displacement and action torque. If the unhooking action is blocked, then feed back to correct the initial action strategy of unhooking and return to step S23. S25: If the unhooking action is not obstructed, real-time cable load data and unhooking mechanism displacement data are obtained, and the release current of the permanent magnet electromagnetic structure is dynamically adjusted to keep the unhooking mechanism moving smoothly.

Images