Ice load on ship hull in wave environment

Abstract

The present application relates to the field of polar ship engineering, and particularly relates to a method for testing the hull ice load characteristics of a ship in a wave environment, comprising: setting up a test physical framework and a test basic scene; constructing a collaborative decision-making framework for a model ice group using a swarm cooperation mechanism, and embedding a dielectric elastomer deformation feedback mechanism as a control input for decision-making, with the output being the movement state and contact pressure time series data of the model ice group; and extracting ice load component data to obtain a parameter set reflecting the ice load characteristics. The present application realizes dynamic interaction and collaborative decision-making between ice floes by constructing a floe group-specific neighborhood group and an adaptive function. Each ice floe model targets the ice-facing surface of the ship hull, accurately reproducing the collaborative drift, accumulated sliding and force chain formation-breaking process of the ice floe group in a real ice area. The magnetic field is used to adjust the movement posture of the ice floe in real time, avoiding test interruption caused by collision and deformation of the ice floe model, while ensuring that the ice floe group quickly returns to the collaborative movement trend.

Description

Technical Field

[0001] This invention relates to the field of polar marine engineering, and specifically to a test method for the ice load characteristics of a ship hull under wave conditions. Background Technology

[0002] With global warming and accelerated melting of the Arctic ice cap, polar resource development and the construction of the Ice Silk Road have become a global strategic focus. Dynamic positioning vessels, as core equipment for polar resource exploration and exploitation, directly determine the efficiency and safety of polar development through their safe operation capabilities. These vessels must withstand complex coupled loads from wind, waves, currents, and floating ice. The dynamic coupling effect between floating ice and waves is a key factor leading to hull load fluctuations and dynamic positioning system instability. Therefore, accurately acquiring the hull's floating ice load characteristics under wave conditions has become a core requirement for ship structural strength design and dynamic positioning system optimization.

[0003] However, existing methods for testing the characteristics of floating ice loads on ship hulls have many limitations: for example, the motion control precision of floating ice models is insufficient; traditional tests often simulate floating ice drift by towing, making it difficult to reproduce the real state of floating ice swarms in wave environments, such as coordinated movement, accumulation, and slippage; and the floating ice models lack targeted biomimetic design, with their textures and physical properties differing significantly from real sea ice, leading to distorted test scenarios. Furthermore, the separation of load components is poor; the data collected is affected by multiple factors such as current loads and wave loads, and the lack of effective separation methods makes it difficult to accurately extract pure ice load components, resulting in significant deviations between test results and actual operating conditions. These shortcomings severely restrict the reliability of test data and cannot provide effective support for the design of ships in ice-covered areas and the optimization of dynamic positioning systems. Therefore, there is an urgent need for a test method that can accurately simulate the polar ice-wave coupling environment and efficiently extract ice load characteristics. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing a test method for the ice load characteristics of ship hulls under wave conditions.

[0005] The technical solution of this invention to solve the above-mentioned technical problems is as follows: a method for testing the floating ice load characteristics of a ship hull under wave conditions, the method comprising: S1. Based on the experimental design parameters and the scale ratio preset by the ship model, the size of the floating ice model is determined, the floating ice model is made, and a model ice group is formed in the experimental pool. Based on the size of the experimental pool, the layout of the magnetorheological drive unit array is planned through a three-dimensional spatial layout algorithm. The measured spectrum data of polar waves is introduced into the wave generator, the wave generation parameters are adjusted, and the physical framework of the experiment is completed. S2. Based on the experimental physical framework, and combined with the ship model and the preset ice-facing angle, the ship model is fixedly deployed and its levelness is calibrated. After the waves are generated and stabilized, the initial position coordinates of each floating ice model in the model ice group are calibrated. Dielectric elastomer film sensors are attached to the surface of each floating ice model to complete the construction of the basic experimental scenario. S3. A collaborative decision-making framework for the model ice swarm is constructed using a bee colony collaboration mechanism, and a dielectric elastic body deformation feedback mechanism is embedded as the control input for decision-making. Through magnetic field regulation, the magnetorheological matrix is ​​driven to regulate the motion state of the model ice swarm. A topology optimization path planning algorithm is integrated to generate the global optimal trajectory of the hull model, complete the deployment of decision-making, enable the model ice swarm to reach the target ice-facing area, and finally output the motion state and contact pressure time series data of the model ice swarm. S4. Real-time acquisition of internal stress diagrams and flow velocity field data of floating ice models in the model ice group in the test pool, combined with contact pressure time series data, and obtaining multi-field coupled datasets through data segmentation and storage algorithm, performing blind source separation on the dataset, extracting ice load component data, and extracting time-frequency features from the ice load component data to obtain a parameter set reflecting the characteristics of floating ice loads.

[0006] In a preferred embodiment, S1 calculates the corresponding ice model thickness based on the actual dimensions of the target ice floe by using a scaling ratio conversion algorithm according to the experimental design parameters and the preset scaling ratio of the hull model, combined with the similarity principle. Then, by combining the roundness parameter and using a geometric parameter iteration algorithm, the ice model size of different geometric shapes is determined. A biomimetic texture parametric modeling algorithm is adopted to extract texture depth, texture spacing, and texture distribution density based on the freeze-thaw texture and collision texture of the target ice surface, and to construct a three-dimensional biomimetic texture model. The three-dimensional biomimetic texture model is used as the physical basis of the ice model for processing and acquisition. S1 further plans the layout of the magnetorheological drive unit array based on the size of the test water tank using a three-dimensional spatial layout algorithm, specifically as follows: Using a three-dimensional spatial layout algorithm and considering the size of the test pool, we determined that magnetorheological drive unit arrays of appropriate specifications should be arranged at the bottom and sides of the pool. The array spacing of the magnetorheological drive unit arrays is set according to the requirement of full magnetic field coverage, so as to ensure that the magnetic field covers the entire test area without any magnetic field blind spots. The magnetorheological drive unit is fixed on a preset mounting bracket, and the wiring is completed using a waterproof wiring process. Each magnetorheological drive unit is connected to an independent control interface. The magnetic field strength calibrator is started, and a closed-loop magnetic field calibration algorithm is used to calibrate the magnetic field output of each magnetorheological drive unit. The actual output value of each magnetorheological drive unit under different magnetic field strengths is detected point by point, and the deviation is corrected by PID adjustment algorithm.

[0007] In a preferred embodiment, S1 further includes: Based on measured polar wave spectrum data, a spectrum fitting algorithm was used to extract the frequency distribution and energy distribution characteristics of the waves. Wave height data at different locations in the test pool were collected in real time, and the energy loss rate during wave propagation was calculated. Based on the energy loss rate and the extracted frequency distribution and energy distribution characteristics, the motion amplitude and frequency of the wave generator push plate were dynamically adjusted to complete the adjustment of the wave generation parameters.

[0008] In a preferred embodiment, S2 specifically includes: After fixing the hull model in the preset position, adjust the hull model so that the ice-facing surface is consistent with the preset ice-facing angle; After the wave generator is started and the wave field stabilizes, the initial position coordinates of each ice floe model in the stable wave field are calibrated, and a dielectric elastomer thin film sensor is attached to the surface of each ice floe model. The capacitance signal is acquired from the dielectric elastomer film sensor. High-frequency noise in the capacitance signal is filtered out by a first-order low-pass filter algorithm, and temperature drift interference in the capacitance signal is eliminated by a baseline calibration algorithm. The capacitance signal is converted into the corresponding contact pressure using a capacitance-pressure conversion formula. A unique ID is assigned to each dielectric elastomer film sensor and bound to the corresponding ice floe model.

[0009] In a preferred embodiment, S3 constructs a collaborative decision-making framework for the model ice swarm using a bee colony cooperation mechanism, including: The ice-facing surface of the ship model is set as the target point of the floating ice model. The three-dimensional coordinate data of the edge and center of the ice-facing surface are collected by the three-dimensional coordinate acquisition algorithm to construct the target point coordinate matrix, which serves as the motion pointing reference for each individual floating ice model. Based on the size of the test pool and the control boundary of the magnetorheological drive unit array, the motion boundary of each floating ice model is divided by the boundary division algorithm. The maximum geometric size of the floating ice model in the model ice group is estimated using a radius estimation algorithm. A neighborhood search radius is set, and the number of floating ice per unit area in the test pool is multiplied by the neighborhood search area to obtain the minimum number of floating ice models that the floating ice model with the maximum geometric size will contact within the movement boundary. The neighborhood search area is calculated by the formula for the area of ​​a circle and the neighborhood search radius. The obtained minimum number is used as the interaction threshold for each floating ice model to interact with other floating ice models in the neighborhood. Based on the preset interaction threshold, the corresponding neighborhood retrieval radius is obtained by using a radius estimation algorithm for each floating ice model in the model ice group, and combined with the initial position coordinates of each floating ice, the interactive neighborhood floating ice models within the movement boundary of each floating ice model are retrieved and each floating ice model is constructed into a unique neighborhood group. Based on the initial position coordinates and target point coordinate matrix of each ice floe model, the straight-line distance from each ice floe model to the center region of the target point is calculated. According to the motion boundary corresponding to each ice floe model, the straight-line distance value is converted into a relative distance value. The real-time distance between each ice floe model and all neighboring ice floes is calculated. Based on the difference between the average real-time distance of each ice floe model and the average distance of all ice floe models, the distance deviation value of the neighborhood pairing of each ice floe model is obtained. The fitness function is constructed using the distance deviation value and the relative distance value as input.

[0010] In a preferred embodiment, after obtaining the relative distance value and the spacing deviation value, the relative distance value and the neighborhood pairing spacing deviation value of each ice floe model are multiplied by the corresponding fitting weights according to the fitness function preset, and then summed to obtain the real-time fitness value of each ice floe model. The fitness function is embedded into the bee colony algorithm. After the algorithm is deployed, the velocity of each ice floe model is initialized using a velocity initialization algorithm. Based on the size of each ice floe model, the magnetorheological drive unit array is activated, and an initial velocity is allocated. The allocation of the initial velocity is based on the size and density of each ice floe model, and its weight is obtained. Then, based on the linear relationship between the weight and the movement velocity of the ice floe model under the same magnetic field strength, the initial velocity is allocated to ensure that the initial velocity of each ice floe model is the same. The bee colony algorithm is then activated for iteration. The position coordinates of each ice floe model are collected in real time and substituted into the fitness function to calculate the corresponding fitness. The neighboring ice floe models in the dedicated neighborhood group are called to select the optimal fitness value in the neighborhood. The step size optimization algorithm is used to adjust the optimization step size based on the difference between the fitness of the ice floe model and the optimal fitness value in the neighborhood. After that, the velocity of the corresponding ice floe model is optimized based on the positional deviation between the ice floe model and the center area of ​​the target point and the optimal velocity of the neighboring ice floe models. The decision iteration is completed and continues to iterate until the model ice swarm forms a stable cooperative movement trend.

[0011] In a preferred embodiment, during decision iteration, S3 embeds a dielectric elastomer deformation feedback mechanism as the control input for collaborative decision-making, including: setting the pressure value that causes the ice floe model to deform as a pressure threshold; converting the capacitance signal collected in real time by the dielectric elastomer thin film sensor into pressure data; comparing the pressure data with the pressure threshold that causes the ice floe model to deform; and further configuring the following settings if the pressure data exceeds the pressure threshold: Obtain the unique ID, location coordinates, and pressure data of the corresponding ice floe model. Use the unique ID of the ice floe model to call the corresponding exclusive neighborhood group to obtain the range of the neighborhood ice floes. Based on the coordinates of the contact point between the corresponding ice floe model and the ship hull model and the neighborhood ice floe models, obtain the orientation of the contact point relative to the corresponding ice floe model. By changing the magnetic field strength and direction of the magnetorheological drive unit closest to the corresponding ice floe model, an opposite magnetic force is generated at the contact point relative to the orientation of the corresponding ice floe model, thereby adjusting the movement of the corresponding ice floe model, and adjusting the pressure data according to the real-time pressure data until the pressure data does not exceed the pressure threshold. After adjusting the corresponding ice floe model, restart the decision iteration. If the contact pressure is still higher than the threshold, continue to optimize and adjust the instructions until the contact is released, so as to ensure that the ice floe group can avoid collisions and model damage, and can quickly return to the cooperative motion trend.

[0012] In a preferred embodiment, S3 uses the wave phase distribution collected in real time in the test pool, i.e., the wave phase data collected in real time, the coordinate boundary of the ship model, the initial position coordinates of the model ice group, and the target ice-facing area as constraints, and uses complex plane analytical functions to complete environmental topology modeling, converting the ship model and the wave field influence area into topological node constraints. Multiple candidate paths with distinct topologies are generated using a graph search algorithm. Bézier curves are then used to smoothly model these candidate paths, specifically: Each candidate path is decomposed into several short paths according to the topological nodes. Each path is treated as a uniform motion segment to simplify energy consumption calculation and ensure that the magnetic field control requirements of each path can be accurately matched. Based on the correlation between the magnetic field strength and acceleration applied by the magnetorheological drive unit array, the magnetic field strength required for the ice floe model to move along each short path and the magnetorheological drive energy consumption of each short path are calculated. The total energy consumption of each candidate path is obtained by subtracting the corresponding energy consumption offset from the magnetorheological drive energy consumption of all short paths. Map the coordinates of all topological nodes, the coordinate boundaries of the ship model, and the coordinates of the neighboring ice floes along the candidate path to the same three-dimensional coordinate system. Calculate the shortest distance between each topological node and the ship model wheel, and the shortest distance between each topological node and the neighboring ice floes as the ice floes move along the candidate path. Compare these distances with the preset safe distance for ice floes to obtain two distance deviation values. Sum the distance deviation values ​​of all topological nodes to obtain the distance deviation value representing the collision risk for each candidate path. An optimization function is constructed with the objectives of minimizing the total energy consumption and the minimum spacing deviation for each candidate path. After cost optimization and topology path classification of candidate paths for the optimization function, the globally optimal path is selected. The optimal path is decomposed into multiple topology nodes, and a decision command is issued to start the wave generator. This allows the model ice swarm to reach the target ice-facing area after passing through multiple topology nodes. A timestamp synchronization algorithm is used to synchronize and correlate the position, velocity, attitude time series data of each floating ice model, the contact pressure time series data of the dielectric elastomer sensor, the wave phase data, and the magnetic field strength control data. Finally, the motion state and contact pressure time series data of the model ice swarm are output.

[0013] In a preferred embodiment, S4 builds a blind source separation model based on the principle of maximizing negative entropy. After randomly initializing the separation matrix using a parameter initialization algorithm, it starts an independent component analysis algorithm with the surface load data of the ship model, the flow field velocity field data, and the wave phase data as inputs to iteratively update the separation matrix. Specifically, it calculates the negative entropy values ​​of each component in the output layer of the separation model in real time using a negative entropy calculation algorithm, adjusts the element values ​​of the separation matrix with the goal of maximizing negative entropy using a gradient descent algorithm, calculates the change in negative entropy after each iteration, and obtains the ice load, flow load, and wave load components after the iteration ends.

[0014] In a preferred embodiment, after obtaining the ice load component data, the data is decomposed by wavelet packet transform to obtain several frequency bands. The wavelet packet coefficients of each frequency band are reconstructed to obtain the ice load signal of the corresponding frequency band, and the energy value of the ice load signal of each frequency band is calculated. For the ice load signal and energy data of each reconstructed frequency band, time-frequency features including energy ratio, peak frequency, kurtosis and skewness are extracted to construct a set of ice load time-frequency feature parameters to reflect the characteristics of floating ice load.

[0015] The beneficial effects of this invention are: by constructing a neighborhood group and fitness function specific to ice floes, dynamic interaction and collaborative decision-making among ice floes are realized. Each ice floe model takes the ice-facing surface of the ship as the target, and optimizes the motion state by combining the real-time distance deviation with neighboring ice floes and the relative value of the target distance. This accurately reproduces the collaborative drift, accumulation and slip, and force chain formation-fracture process of ice floes in real ice areas. Furthermore, the motion attitude of the ice floes is adjusted in real time by magnetic field control, avoiding experimental interruption caused by collision deformation of the ice floe model, while ensuring that the ice floe group quickly returns to the collaborative motion trend. Based on measured polar wave spectrum data, a topology optimization path planning algorithm was used, with wave phase and ship coordinate boundaries as constraints. Environmental topology modeling was completed using complex plane analytical functions. The generated globally optimal path minimized energy consumption driven by magnetorheological forces and avoided collision risks through spacing deviation calculations. Simultaneously, a phase synchronization algorithm was used to calibrate the wave and ice floe motion phases, ensuring precise matching between wave propagation and ice floe drift, thus reproducing the dynamic characteristics of polar ice-wave coupling. Furthermore, after decomposing the path into topology nodes, a timestamp synchronization algorithm was used to correlate ice floe motion, pressure data, and wave phases, providing a high-quality dataset with temporal consistency for subsequent load component separation, indirectly improving the accuracy of pure ice load extraction. Attached Figure Description

[0016] Figure 1 This is a flowchart of the present invention; Figure 2 An ice floe model prepared from the polyethylene material used in this invention; Figure 3 This is a diagram showing the interannual variation of the effective wave height in various polar sea areas, which is referenced by the wave-generating parameters of the wave generator in this invention. Figure 4 This diagram shows the peak surface load distribution of the ship model of the present invention under different ice-facing angles. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] As attached Figure 1 As shown in the figure, this embodiment provides a method for testing the floating ice load characteristics of a ship hull under wave conditions, including the following steps: S1. Based on the experimental design parameters and the scale ratio preset by the ship model, the size of the floating ice model is determined, the floating ice model is made, and a model ice group is formed in the experimental pool. Based on the size of the experimental pool, the layout of the magnetorheological drive unit array is planned through a three-dimensional spatial layout algorithm. The measured spectrum data of polar waves is introduced into the wave generator, the wave generation parameters are adjusted, and the physical framework of the experiment is completed. S1, based on the experimental design parameters, uses a scaling ratio conversion algorithm. Using the pre-set scaling ratio of the hull model as a basis and combining the similarity principle, it substitutes the actual size of the target ice floe to calculate the corresponding ice floe model thickness. Combined with the roundness parameter, it uses a geometric parameter iterative algorithm to determine the size of ice floe models with different geometric shapes. A biomimetic texture parametric modeling algorithm is adopted to extract texture depth, texture spacing, and texture distribution density based on the freeze-thaw texture and collision texture of the target ice surface, and to construct a three-dimensional biomimetic texture model. The three-dimensional biomimetic texture model is used as the physical basis of the ice model for processing and acquisition. In some other specific implementations, please provide the specific materials selected for the ice floe model and the reasons for that selection, as follows: The density of polypropylene material ranges from 890 to 920 kg / m³. According to actual measurements, the density of the polypropylene model ice used in the experiment was 912.2 kg / m³. Meanwhile, the density of sea ice in nature ranges from 720 to 940 kg / m³. This is because the density of sea ice increases with increasing salinity and decreasing air content. The longer the ice age, the lower the density due to the seepage of brine from the ice. The average density is 910 kg / m³, which is similar to the density of polypropylene material.

[0019] Furthermore, the coefficient of friction of polypropylene is similar to that of sea ice. In addition, the Young's modulus of polypropylene is typically 1.32–1.42 GPa, which is close to that of real sea ice. Due to its ease of processing, polypropylene is chosen for model ice. Figure 2 As shown.

[0020] Furthermore, this application also provides a general processing method, which uses a high-speed mixing algorithm to control the mixing process. Polypropylene particles and a suitable proportion of magnetic ferrite powder are fed into a high-speed mixer, and a suitable proportion of dispersant is added at the same time to ensure that the magnetic ferrite powder is uniformly dispersed in the polypropylene matrix. The mixture is then fed into an injection molding machine and injected into a mold. The addition of magnetic ferrite powder makes the ice floe model used in this application magnetic, thus providing a basis for the subsequent adjustment of the magnetorheological driven unit array in the movement of the ice floe. S1 also plans the layout of the magnetorheological drive unit array based on the size of the test water tank using a three-dimensional spatial layout algorithm, specifically: Using a three-dimensional spatial layout algorithm and considering the size of the test pool, we determined that magnetorheological drive unit arrays of appropriate specifications should be arranged at the bottom and sides of the pool. The array spacing of the magnetorheological drive unit arrays is set according to the requirement of full magnetic field coverage, so as to ensure that the magnetic field covers the entire test area without any magnetic field blind spots. The magnetorheological drive unit is fixed on a preset mounting bracket, and the wiring is completed using a waterproof wiring process. Each magnetorheological drive unit is connected to an independent control interface. The magnetic field strength calibrator is started, and a closed-loop magnetic field calibration algorithm is used to calibrate the magnetic field output of each magnetorheological drive unit. The actual output value of each magnetorheological drive unit under different magnetic field strengths is detected point by point, and the deviation is corrected by PID adjustment algorithm.

[0021] S1 also includes: Based on measured polar wave spectrum data, a spectrum fitting algorithm is used to extract the frequency distribution and energy distribution characteristics of the waves. Wave height data at different locations in the experimental pool are collected in real time, and the energy loss rate during wave propagation is calculated. Based on the energy loss rate and the extracted frequency distribution and energy distribution characteristics, the motion amplitude and frequency of the wave generator push plate are dynamically adjusted to complete the adjustment of the wave generation parameters. Furthermore, this application uses a phase synchronization algorithm to calibrate the wave phase of the wave generator with the motion phase of the magnetically sensitive model ice. By adjusting the start-up sequence of the wave generator, it is ensured that the wave phase and the floating ice movement are accurately synchronized, and the error is controlled within a preset range, which conforms to the real polar ice-wave coupling environment, thereby completing the construction of the experimental physical framework. In response to the above technical solution, this application provides specific verification steps during implementation: Because wave conditions in polar seas vary considerably depending on geographical location and meteorological conditions, a survey of wave height data from 2011 to 2019 shows that the significant wave height in polar seas ranges from 0.5m to 1.5m in summer and from 1.5m to 3m in winter. To meet the requirements of severe sea conditions, this application proposes to use winter wave parameters. For many years, the average significant wave height in various Arctic sea areas has remained stable at around 2.5m in winter. Figure 3 As shown in the figure, the solid line corresponds to winter and the dashed line corresponds to summer. In addition, the density of sea ice in the sea area is always between 20% and 60%. Based on this, the actual wave height in this experiment is set at 2.4m, and the effective wave height at the corresponding model scale is 0.03m obtained by similarity.

[0022] The average wave period for this experiment is set at 9.00 s in actual sea area. The corresponding wave frequency at the model scale is the encounter frequency in the dynamic positioning experiment. Generally, the same wave means the same wave height and the same encounter frequency. The wave period mentioned above corresponds to the scenario where a ship encounters waves during dynamic positioning operations in actual sea area. When simulating this relative motion through towing, the wave generator should produce waves with the same height but different frequencies. The conversion relationship between encounter frequency and wave frequency is as follows: In the formula ω e Indicates the frequency of encounters. ω Indicates wave frequency. V Indicates the towing speed. The wave direction angle is 0°, as the ship model in this application is always towed facing the waves. After conversion, the wavelength of the waves generated by the wave generator at the model scale is 1.79m, corresponding to a period of 1.07s. Therefore, the amplitude and frequency of the wave generator's pusher plate are dynamically adjusted.

[0023] S2. Based on the experimental physical framework, and combined with the ship model and the preset ice-facing angle, the ship model is fixedly deployed and its levelness is calibrated. After the waves are generated and stabilized, the initial position coordinates of each floating ice model in the model ice group are calibrated. Dielectric elastomer film sensors are attached to the surface of each floating ice model to complete the construction of the basic experimental scenario. S2 specifically includes: After fixing the hull model in the preset position, adjust the hull model so that the ice-facing surface is consistent with the preset ice-facing angle; In some specific practical applications, the loads on the hull model vary depending on its ice-facing angle. In hull ice load characteristic tests, if the load is too small, the load characteristics cannot be determined; if the load is too large, it will damage the hull. Therefore, the preset ice-facing angle still needs further confirmation. This application addresses this situation by providing a specific implementation process: Because the location of the peak surface load varies depending on the angle of the ship model, the pressure cloth was divided into three areas: "bow," "midship," and "stern." The pressure cloth was placed at the bow when the ship model angle was 0° and 45°, at the midship angle at 90°, and at the stern at 135° and 180°. When the pressure cloth was applied to the hull surface and subjected to compression and collision with floating ice, it generated electrical signals, which were then converted into load values ​​at various measurement points within the measurement area. The output of a single measurement was a 20×32 load matrix, where each cell was a 3cm×1cm square. Therefore, the surface load value generated by the collision of floating ice with the hull in the ice dynamic positioning test can be considered as the reading of the pressure cloth.

[0024] This application integrates the peak load measurement results at various locations on the hull surface under five different angles along the ship's length and plots them as follows: Figure 4 As shown; Figure 4 The peak loads mainly occurred at the point where the ice floe collided head-on with the model ship. The peak load values ​​were similar at ice-facing angles of 0° and 45°, but their locations differed. At 0°, the peak load was 8.166 kPa, occurring at the bow, after which it rapidly decreased along the ship's length. At 45°, the peak load was 8.250 kPa, occurring approximately 12 cm from the bow, with significant load values ​​still present in its vicinity. For example, at approximately 9 cm from the bow, the load reached 6.482 kPa, and even near the midships, there were still considerable load fluctuations.

[0025] The load peak of the model ship at an ice-facing angle of 90° exhibits a bulge that is higher in the middle and lower at both ends, meaning the load value is higher closer to the midships. The maximum load peak of 15.573 kPa occurs approximately 18 cm from the midships. This is because a large amount of ice accumulates in this area and impacts the ship's side directly, while the ice force chains near the bow and stern tilt, resulting in a smaller angle between the force lines and the ship's side. Furthermore, there are two locations in the midships with extremely low load values, only around 1 kPa. This is due to a triangular gap that forms as the model enters the ice-covered area and remains there until it leaves. This gap precisely covers the corresponding sensor units, demonstrating the relatively random nature of the ice load.

[0026] The load distribution of the model ship at ice-facing angles of 135° and 180° is similar to that at 45° and 0°. The peak load at 180° is 33.084 kPa, which occurs at the sternpost. The peak load at 135° is 23.234 kPa, which occurs about 12 cm from the bowpost.

[0027] Therefore, the ship model can be adjusted to a suitable ice-facing angle through the aforementioned pre-set tests.

[0028] After the wave generator is started and the wave field stabilizes, the initial position coordinates of each ice floe model in the stable wave field are calibrated, and a dielectric elastomer thin film sensor is attached to the surface of each ice floe model. The capacitance signal is acquired from the dielectric elastomer film sensor. High-frequency noise in the capacitance signal is filtered out by a first-order low-pass filter algorithm, and temperature drift interference in the capacitance signal is eliminated by a baseline calibration algorithm. The capacitance signal is converted into the corresponding contact pressure using a capacitance-pressure conversion formula. A unique ID is assigned to each dielectric elastomer film sensor and bound to the corresponding ice floe model.

[0029] S3 constructs a collaborative decision-making framework for the model ice swarm using a bee colony cooperation mechanism, including: The ice-facing surface of the ship model is set as the target point of the floating ice model. The three-dimensional coordinate data of the edge and center of the ice-facing surface are collected by the three-dimensional coordinate acquisition algorithm to construct the target point coordinate matrix, which serves as the motion pointing reference for each individual floating ice model. Based on the size of the test pool and the control boundary of the magnetorheological drive unit array, the motion boundary of each floating ice model is divided by the boundary division algorithm. The maximum geometric size of the floating ice model in the model ice group is estimated using a radius estimation algorithm. A neighborhood search radius is set, and the number of floating ice per unit area in the test pool is multiplied by the neighborhood search area to obtain the minimum number of floating ice models that the floating ice model with the maximum geometric size will contact within the movement boundary. The neighborhood search area is calculated by the formula for the area of ​​a circle and the neighborhood search radius. The obtained minimum number is used as the interaction threshold for each floating ice model to interact with other floating ice models in the neighborhood. Based on the preset interaction threshold, the corresponding neighborhood retrieval radius is obtained by using a radius estimation algorithm for each floating ice model in the model ice group, and combined with the initial position coordinates of each floating ice, the interactive neighborhood floating ice models within the movement boundary of each floating ice model are retrieved and each floating ice model is constructed into a unique neighborhood group. Based on the initial position coordinates and target point coordinate matrix of each ice floe model, the straight-line distance from each ice floe model to the center region of the target point is calculated. According to the motion boundary corresponding to each ice floe model, the straight-line distance value is converted into a relative distance value. The real-time distance between each ice floe model and all neighboring ice floes is calculated. Based on the difference between the average real-time distance of each ice floe model and the average distance of all ice floe models, the distance deviation value of the neighborhood pairing of each ice floe model is obtained. The fitness function is constructed using the distance deviation value and the relative distance value as input.

[0030] After obtaining the relative distance value and the spacing deviation value, according to the fitness function's preset adaptation weight, the relative distance value and the neighborhood pairing spacing deviation value of each ice floe model are multiplied by the corresponding adaptation weight and then summed to obtain the real-time fitness value of each ice floe model. In the above steps, the adaptation weights preset in this application are based on two quantified core indicators, the distance deviation value and the relative distance value. The function weight allocation algorithm is used to determine the adaptation weights of the two indicators. It is clear that the target point distance weight is higher than the distance deviation weight. The influence scenario of the two indicators on the coordinated movement of ice floes is decomposed by the factor priority analysis algorithm. The quantitative scoring is carried out in combination with the pre-experiment experience data to initially define the weight allocation range. Then, the weight adaptation verification algorithm is used to build a calibration simulation model by calling the initial position and neighborhood distribution data of the ice floe group. The weight combination is repeatedly adjusted and simulated and verified until a reasonable weight that can both ensure the movement of the ice floe group towards the target point and control the collision rate is obtained. The weight values ​​are fixed and bound to the quantification method of the two core indicators as the parameter support for the fitness value calculation. Furthermore, the lower the fitness value, the better the current motion strategy of the ice floe model, providing a basis for judgment on the motion optimization of the ice floe model; The fitness function is embedded into the bee colony algorithm. After the algorithm is deployed, the velocity of each ice floe model is initialized using a velocity initialization algorithm. Based on the size of each ice floe model, the magnetorheological drive unit array is activated, and an initial velocity is allocated. The allocation of the initial velocity is based on the size and density of each ice floe model, and its weight is obtained. Then, based on the linear relationship between the weight and the movement velocity of the ice floe model under the same magnetic field strength, the initial velocity is allocated to ensure that the initial velocity of each ice floe model is the same. The bee colony algorithm is then activated for iteration. The position coordinates of each ice floe model are collected in real time and substituted into the fitness function to calculate the corresponding fitness. The neighboring ice floe models in the dedicated neighborhood group are called to select the optimal fitness value in the neighborhood. The step size optimization algorithm is used to adjust the optimization step size based on the difference between the fitness of the ice floe model and the optimal fitness value in the neighborhood. After that, the velocity of the corresponding ice floe model is optimized based on the positional deviation between the ice floe model and the center area of ​​the target point and the optimal velocity of the neighboring ice floe models. The decision iteration is completed and continues to iterate until the model ice swarm forms a stable cooperative movement trend.

[0031] S3. A collaborative decision-making framework for the model ice swarm is constructed using a bee colony collaboration mechanism, and a dielectric elastic body deformation feedback mechanism is embedded as the control input for decision-making. Through magnetic field regulation, the magnetorheological matrix is ​​driven to regulate the motion state of the model ice swarm. A topology optimization path planning algorithm is integrated to generate the global optimal trajectory of the hull model, complete the deployment of decision-making, enable the model ice swarm to reach the target ice-facing area, and finally output the motion state and contact pressure time series data of the model ice swarm. During decision iteration, S3 embeds a dielectric elastomer deformation feedback mechanism as the control input for collaborative decision-making. This includes setting the pressure value that causes the ice floe model to deform as a pressure threshold, converting the capacitance signal collected in real time by the dielectric elastomer thin film sensor into pressure data, and comparing the pressure data with the pressure threshold that causes the ice floe model to deform. If the pressure data exceeds the pressure threshold, the following settings are also included: Obtain the unique ID, location coordinates, and pressure data of the corresponding ice floe model. Use the unique ID of the ice floe model to call the corresponding exclusive neighborhood group to obtain the range of the neighborhood ice floes. Based on the coordinates of the contact point between the corresponding ice floe model and the ship hull model and the neighborhood ice floe models, obtain the orientation of the contact point relative to the corresponding ice floe model. By changing the magnetic field strength and direction of the magnetorheological drive unit closest to the corresponding ice floe model, an opposite magnetic force is generated at the contact point relative to the orientation of the corresponding ice floe model, thereby adjusting the movement of the corresponding ice floe model, and adjusting the pressure data according to the real-time pressure data until the pressure data does not exceed the pressure threshold. After adjusting the corresponding ice floe model, restart the decision iteration. If the contact pressure is still higher than the threshold, continue to optimize and adjust the instructions until the contact is released, so as to ensure that the ice floe group can avoid collisions and model damage, and can quickly return to the cooperative motion trend.

[0032] S3 uses the wave phase distribution collected in real time in the test pool, i.e., the wave phase data collected in real time, the coordinate boundary of the ship model, the initial position coordinates of the model ice group, and the target ice-facing area as constraints. It uses complex plane analytical functions to complete environmental topology modeling and converts the ship model and the wave field influence area into topological node constraints. Multiple candidate paths with distinct topologies are generated using a graph search algorithm. Bézier curves are then used to smoothly model these candidate paths, specifically: Each candidate path is decomposed into several short paths according to the topological nodes. Each path is treated as a uniform motion segment to simplify energy consumption calculation and ensure that the magnetic field control requirements of each path can be accurately matched. Based on the correlation between the magnetic field strength and acceleration applied by the magnetorheological drive unit array, the magnetic field strength required for the ice floe model to move along each short path and the magnetorheological drive energy consumption of each short path are calculated. The total energy consumption of each candidate path is obtained by subtracting the corresponding energy consumption offset from the magnetorheological drive energy consumption of all short paths. Map the coordinates of all topological nodes, the coordinate boundaries of the ship model, and the coordinates of the neighboring ice floes along the candidate path to the same three-dimensional coordinate system. Calculate the shortest distance between each topological node and the ship model wheel, and the shortest distance between each topological node and the neighboring ice floes as the ice floes move along the candidate path. Compare these distances with the preset safe distance for ice floes to obtain two distance deviation values. Sum the distance deviation values ​​of all topological nodes to obtain the distance deviation value representing the collision risk for each candidate path. An optimization function is constructed with the objectives of minimizing the total energy consumption and the minimum spacing deviation for each candidate path. After cost optimization of the optimization function and topology classification of candidate paths, the globally optimal path is selected. After decomposing the optimal path into multiple topological nodes, a decision command is issued and a wave generator is started, so that the model ice swarm reaches the target ice-facing area after passing through multiple topological nodes. A timestamp synchronization algorithm is used to synchronize and correlate the position, velocity, attitude time series data of each floating ice model, the contact pressure time series data of the dielectric elastomer sensor, the wave phase data, and the magnetic field strength control data. Finally, the motion state and contact pressure time series data of the model ice swarm are output.

[0033] Furthermore, the control logic of this application can be described as follows: the control unit, based on the instructions output by the algorithm and combined with the real-time position and pressure data of each ice floe, generates a corresponding magnetic field control signal through magnetic field modulation and sends it to the corresponding drive unit of the three-dimensional magnetorheological drive matrix; the drive unit outputs the corresponding magnetic field strength according to the control signal, regulates the motion acceleration of the magnetically sensitive model ice, causing the model ice group to move towards the hull, and dynamically adjusts the magnetic field distribution according to the real-time data of each ice floe model to control the attitude of the ice floes. During the process, pressure data and motion data from the dielectric elastomer sensor are continuously collected, and an algorithm iteration is completed periodically. If the contact pressure of the ice floes exceeds the limit or the motion trajectory deviates from the optimal path, the motion instructions are immediately regenerated to adjust the motion state of the ice floes and ensure the effective interaction between the ice floes and the hull.

[0034] S4. Real-time acquisition of internal stress diagrams and flow velocity field data of floating ice models in the model ice group in the test pool, combined with contact pressure time series data, and obtaining multi-field coupled datasets through data segmentation and storage algorithm, performing blind source separation on the dataset, extracting ice load component data, and extracting time-frequency features from the ice load component data to obtain a parameter set reflecting the characteristics of floating ice loads.

[0035] S4 builds a blind source separation model based on the principle of maximizing negative entropy. It adopts a parameter initialization algorithm to randomly initialize the separation matrix. Then, it takes the surface load data of the ship model, the flow velocity field data, and the wave phase data as inputs and starts the independent component analysis algorithm to iteratively update the separation matrix. Specifically, it calculates the negative entropy value of each component in the output layer of the separation model in real time through the negative entropy calculation algorithm. With the goal of maximizing negative entropy, it adjusts the element values ​​of the separation matrix through the gradient descent algorithm. After each iteration, it calculates the change in negative entropy. After the iteration is completed, the ice load, flow load, and wave load components are obtained.

[0036] After obtaining the ice load component data, it is decomposed by wavelet packet transform to obtain several frequency bands. The wavelet packet coefficients of each frequency band are reconstructed to obtain the ice load signal of the corresponding frequency band, and the energy value of the ice load signal of each frequency band is calculated. For the ice load signal and energy data of each reconstructed frequency band, time-frequency features including energy ratio, peak frequency, kurtosis and skewness are extracted to construct a set of ice load time-frequency feature parameters to reflect the characteristics of floating ice load.

Claims

1. A test method for the ice-floating load characteristics of a ship hull under wave conditions, characterized in that, The method includes: S1. Based on the experimental design parameters and the scale ratio preset by the ship model, the size of the floating ice model is determined, the floating ice model is made, and a model ice group is formed in the experimental pool. Based on the size of the experimental pool, the layout of the magnetorheological drive unit array is planned through a three-dimensional spatial layout algorithm. The measured spectrum data of polar waves is introduced into the wave generator, the wave generation parameters are adjusted, and the physical framework of the experiment is completed. S2. Based on the experimental physical framework, and combined with the ship model and the preset ice-facing angle, the ship model is fixedly deployed and its levelness is calibrated. After the waves are generated and stabilized, the initial position coordinates of each floating ice model in the model ice group are calibrated. Dielectric elastomer film sensors are attached to the surface of each floating ice model to complete the construction of the basic experimental scenario. S3. A collaborative decision-making framework for the model ice swarm is constructed using a bee colony collaboration mechanism, and a dielectric elastic body deformation feedback mechanism is embedded as the control input for decision-making. Through magnetic field regulation, the magnetorheological matrix is ​​driven to regulate the motion state of the model ice swarm. A topology optimization path planning algorithm is integrated to generate the global optimal trajectory of the hull model, complete the deployment of decision-making, enable the model ice swarm to reach the target ice-facing area, and finally output the motion state and contact pressure time series data of the model ice swarm. S4. Real-time acquisition of internal stress diagrams and flow velocity field data of floating ice models in the model ice group in the test pool, combined with contact pressure time series data, and obtaining multi-field coupled datasets through data segmentation and storage algorithm, performing blind source separation on the dataset, extracting ice load component data, and extracting time-frequency features from the ice load component data to obtain a parameter set reflecting the characteristics of floating ice loads.

2. The test method for the hull ice-floating load characteristics under wave conditions according to claim 1, characterized in that, S1, based on the experimental design parameters, uses a scaling ratio conversion algorithm, taking the preset scaling ratio of the hull model as a basis, and combining the similarity principle, substitutes the actual size of the target ice floe to calculate the corresponding ice floe model thickness. Combined with the roundness parameter, the size of ice floe models with different geometric shapes is determined through a geometric parameter iteration algorithm. A biomimetic texture parametric modeling algorithm is adopted to extract texture depth, texture spacing, and texture distribution density based on the freeze-thaw texture and collision texture of the target ice surface, and to construct a three-dimensional biomimetic texture model. The three-dimensional biomimetic texture model is used as the physical basis of the ice model for processing and acquisition. S1 further plans the layout of the magnetorheological drive unit array based on the size of the test water tank using a three-dimensional spatial layout algorithm, specifically as follows: Using a three-dimensional spatial layout algorithm and considering the size of the test pool, we determined that magnetorheological drive unit arrays of appropriate specifications should be arranged at the bottom and sides of the pool. The array spacing of the magnetorheological drive unit arrays was set according to the requirement of full magnetic field coverage. The magnetorheological drive unit is fixed on a preset mounting bracket, and the wiring is completed using a waterproof wiring process. Each magnetorheological drive unit is connected to an independent control interface. The magnetic field strength calibrator is started, and a closed-loop magnetic field calibration algorithm is used to calibrate the magnetic field output of each magnetorheological drive unit. The actual output value of each magnetorheological drive unit under different magnetic field strengths is detected point by point, and the deviation is corrected by PID adjustment algorithm.

3. The method for testing the hull ice-floating load characteristics under wave conditions according to claim 1, characterized in that, S1 further includes: Based on measured polar wave spectrum data, a spectrum fitting algorithm was used to extract the frequency distribution and energy distribution characteristics of the waves. Wave height data at different locations in the test pool were collected in real time, and the energy loss rate during wave propagation was calculated. Based on the energy loss rate and the extracted frequency distribution and energy distribution characteristics, the motion amplitude and frequency of the wave generator push plate were dynamically adjusted to complete the adjustment of the wave generation parameters.

4. The test method for the hull ice-floating load characteristics under wave conditions of ice according to claim 1, characterized in that, S2 specifically includes: After fixing the hull model in the preset position, adjust the hull model so that the ice-facing surface is consistent with the preset ice-facing angle; After the wave generator is started and the wave field stabilizes, the initial position coordinates of each ice floe model in the stable wave field are calibrated, and a dielectric elastomer thin film sensor is attached to the surface of each ice floe model. The capacitance signal is acquired from the dielectric elastomer film sensor. High-frequency noise in the capacitance signal is filtered out by a first-order low-pass filter algorithm, and temperature drift interference in the capacitance signal is eliminated by a baseline calibration algorithm. The capacitance signal is converted into the corresponding contact pressure using a capacitance-pressure conversion formula. A unique ID is assigned to each dielectric elastomer film sensor and bound to the corresponding ice floe model.

5. The method for testing the hull ice-floating load characteristics under wave conditions according to claim 1, characterized in that, The S3 constructs a collaborative decision-making framework for the model ice swarm using a bee colony cooperation mechanism, including: The ice-facing surface of the ship model is set as the target point of the floating ice model. The three-dimensional coordinate data of the edge and center of the ice-facing surface are collected by the three-dimensional coordinate acquisition algorithm to construct the target point coordinate matrix. Based on the size of the test pool and the control boundary of the magnetorheological drive unit array, the motion boundary of each floating ice model is divided by the boundary division algorithm. The maximum geometric size of the floating ice model in the model ice group is estimated using a radius estimation algorithm. The neighborhood search radius is set, and the number of floating ice per unit area in the test pool is multiplied by the neighborhood search area to obtain the minimum number of floating ice models that the floating ice model with the maximum geometric size comes into contact with within the movement boundary. The obtained minimum number is used as the interaction threshold for each floating ice model to interact with other floating ice models in the neighborhood. Based on the preset interaction threshold, the corresponding neighborhood retrieval radius is obtained by using a radius estimation algorithm for each floating ice model in the model ice group, and combined with the initial position coordinates of each floating ice, the interactive neighborhood floating ice models within the movement boundary of each floating ice model are retrieved and each floating ice model is constructed into a unique neighborhood group. Based on the initial position coordinates and target point coordinate matrix of each ice floe model, the straight-line distance from each ice floe model to the center region of the target point is calculated. According to the motion boundary corresponding to each ice floe model, the straight-line distance value is converted into a relative distance value. The real-time distance between each ice floe model and all neighboring ice floes is calculated. Based on the difference between the average real-time distance of each ice floe model and the average distance of all ice floe models, the distance deviation value of the neighborhood pairing of each ice floe model is obtained. The fitness function is constructed using the distance deviation value and the relative distance value as input.

6. The test method for the hull ice-floating load characteristics under wave conditions according to claim 5, characterized in that, After obtaining the relative distance value and the spacing deviation value, according to the fitness function's preset adaptation weight, the relative distance value and the neighborhood pairing spacing deviation value of each ice floe model are multiplied by the corresponding adaptation weight and then summed to obtain the real-time fitness value of each ice floe model. The fitness function is embedded into the bee colony algorithm. After the algorithm is deployed, the velocity of each ice floe model is initialized using a velocity initialization algorithm. Based on the size of each ice floe model, the magnetorheological drive unit array is activated, an initial velocity is allocated, and the bee colony algorithm iteration is initiated. The position coordinates of each ice floe model are collected in real time and substituted into the fitness function to calculate the corresponding fitness. The neighboring ice floe models in the dedicated neighborhood group are called to select the optimal fitness value in the neighborhood. The step size optimization algorithm is used to adjust the optimization step size based on the difference between the fitness of the ice floe model and the optimal fitness value in the neighborhood. Based on the positional deviation between the ice floe model and the center area of ​​the target point and the optimal velocity of the neighboring ice floe models, the velocity of the corresponding ice floe model is optimized, and the decision iteration is completed. The iteration continues until the model ice swarm forms a stable cooperative movement trend.

7. The method for testing the hull ice-floating load characteristics under wave conditions according to claim 6, characterized in that, In the decision iteration, S3 embeds a dielectric elastomer deformation feedback mechanism as the control input for collaborative decision-making, including: setting the pressure value that causes the ice floe model to deform as a pressure threshold; converting the capacitance signal collected in real time by the dielectric elastomer thin film sensor into pressure data; comparing the pressure data with the pressure threshold that causes the ice floe model to deform; and if the pressure data exceeds the pressure threshold, the following settings are also included: Obtain the unique ID, location coordinates, and pressure data of the corresponding ice floe model. Use the unique ID of the ice floe model to call the corresponding exclusive neighborhood group to obtain the range of the neighborhood ice floes. Based on the coordinates of the contact point between the corresponding ice floe model and the ship hull model and the neighborhood ice floe models, obtain the orientation of the contact point relative to the corresponding ice floe model. By changing the magnetic field strength and direction of the magnetorheological drive unit closest to the corresponding ice floe model, an opposite magnetic force is generated at the contact point relative to the orientation of the corresponding ice floe model, thereby adjusting the movement of the corresponding ice floe model, and adjusting the pressure data according to the real-time pressure data until the pressure data does not exceed the pressure threshold. After adjusting the corresponding ice floe model, restart the decision iteration.

8. The test method for the hull ice-floating load characteristics under wave conditions according to claim 1, characterized in that, The S3 uses the wave phase distribution, ship model coordinate boundary, initial position coordinates of model ice group and target ice-facing area collected in real time in the test pool as constraints, and completes environmental topology modeling with complex plane analytical functions to convert the ship model and wave field influence area into topology node constraints. Multiple candidate paths with distinct topologies are generated using a graph search algorithm. Bézier curves are then used to smoothly model these candidate paths, specifically: Each candidate path is decomposed into several short paths according to the topological nodes. Each path is treated as a uniform motion segment. Based on the correlation between the magnetic field strength and acceleration applied by the magnetorheological driving unit array, the magnetic field strength required for the floating ice model to move along each short path and the magnetorheological driving energy consumption of each short path are calculated. The total energy consumption of each candidate path is obtained by subtracting the corresponding energy consumption offset from the magnetorheological driving energy consumption of all short paths. Map the coordinates of all topological nodes, the coordinate boundaries of the ship model, and the coordinates of the neighboring ice floes along the candidate path to the same three-dimensional coordinate system. Calculate the shortest distance between each topological node and the ship model wheel, and the shortest distance between each topological node and the neighboring ice floes as the ice floes move along the candidate path. Compare these distances with the preset safe distance for ice floes to obtain two distance deviation values. Sum the distance deviation values ​​of all topological nodes to obtain the distance deviation value representing the collision risk for each candidate path. An optimization function is constructed with the objectives of minimizing the total energy consumption and the minimum spacing deviation for each candidate path. After cost optimization and topology path classification of candidate paths for the optimization function, the globally optimal path is selected. The optimal path is decomposed into multiple topology nodes, and a decision command is issued to start the wave generator. This allows the model ice swarm to reach the target ice-facing area after passing through multiple topology nodes. A timestamp synchronization algorithm is used to synchronize and correlate the position, velocity, attitude time series data of each floating ice model, the contact pressure time series data of the dielectric elastomer sensor, the wave phase data, and the magnetic field strength control data. Finally, the motion state and contact pressure time series data of the model ice swarm are output.

9. The method for testing the ice-floating load characteristics of a ship hull under wave conditions according to claim 1, characterized in that, The S4 model is built based on the principle of maximizing negative entropy. It uses a parameter initialization algorithm to randomly initialize the separation matrix. Then, it takes the surface load data of the ship model, the flow velocity field data, and the wave phase data as inputs and starts the independent component analysis algorithm to iteratively update the separation matrix. Specifically, it calculates the negative entropy value of each component in the output layer of the separation model in real time through the negative entropy calculation algorithm. With the goal of maximizing negative entropy, it adjusts the element values ​​of the separation matrix through the gradient descent algorithm. After each iteration, it calculates the change in negative entropy. After the iteration is completed, it obtains the ice load, flow load, and wave load components.

10. The method for testing the ice-floating load characteristics of a ship hull under wave conditions according to claim 9, characterized in that, After obtaining the ice load component data, it is decomposed by wavelet packet transform to obtain several frequency bands. The wavelet packet coefficients of each frequency band are reconstructed to obtain the ice load signal of the corresponding frequency band, and the energy value of the ice load signal of each frequency band is calculated. For the ice load signal and energy data of each reconstructed frequency band, time-frequency features including energy ratio, peak frequency, kurtosis and skewness are extracted to construct a set of ice load time-frequency feature parameters to reflect the characteristics of floating ice load.

Images