Canoe body posture monitoring method and monitoring system

Abstract

The invention relates to the technical field of canoeing monitoring, and discloses a canoeing body posture monitoring method and system. The method comprises the following steps: firstly, acquiring real-time environment parameters of a target water area and dynamic parameters of a kayak body when the kayak moves, and generating an initial monitoring parameter set comprising a three-dimensional attitude angle sequence and a hydrodynamic load distribution curve; then, according to the three-dimensional attitude angle sequence, a first evaluation model matched with the hull structure is determined, and the first evaluation model comprises the dynamic mapping relation between the attitude angle and the overturning risk degree; and determining an adaptive second evaluation model according to the hydrodynamic load distribution curve, wherein the adaptive second evaluation model comprises a nonlinear association rule of load distribution and a stability coefficient. And then based on the real-time attitude deviation data and the load fluctuation data, a target model is activated to generate a dynamic early warning instruction, the dynamic early warning instruction is input into an execution unit of a hull attitude adjusting device, and the deflection angle of a balance wing or water distribution of a ballast water tank is corrected so as to guarantee the running stability and safety of the kayak.

Description

Technical Field

[0001] The present invention relates to the technical field of kayak monitoring, and in particular to a kayak hull posture monitoring method and a monitoring system. Background Art

[0002] As an essential component of both watersports and transportation, kayaking's safety and stability have always been a focus of industry attention. During kayaking, even the slightest change in the boat's posture can trigger a series of chain reactions. This is especially true in complex waters, such as turbulent rivers and choppy lakes, where the boat can easily become unbalanced due to external interference or changes in its own motion.

[0003] Traditional kayak posture control relies heavily on the athlete's experience and manual adjustments, which have significant limitations. Athletes must constantly monitor the boat's swaying and rely on their sense of touch to adjust their center of gravity or paddle to maintain balance. However, in unexpected situations, such as strong crosswinds or sudden changes in current, it's often difficult for athletes to react quickly enough, significantly increasing the risk of capsizing.

[0004] With the development of kayaking technology, some simple attitude monitoring devices have begun to be used on kayaks. These devices generally only collect a single attitude parameter, such as the tilt angle, and are unable to conduct in-depth analysis and evaluation of the collected data. They only issue a simple alarm signal when the hull tilts significantly, but cannot predict potential capsizing risks in advance, nor can they provide targeted adjustment recommendations based on different water conditions and hull conditions.

[0005] Existing monitoring methods lack effective consideration of hydrodynamic loads. Hydrodynamic loads are a crucial factor affecting kayak stability. Different factors, such as current velocity and wave height, can significantly alter the distribution of hydrodynamic loads on the kayak, affecting its stability. Traditional methods often overlook the relationship between this load distribution and stability, making it difficult to fully assess the kayak's safety status.

[0006] Current kayak attitude control devices are relatively limited in functionality, mostly enabling simple balance corrections by adjusting the deflection angle of the stabilizer, without the ability to comprehensively adjust the system based on changes in hydrodynamic loads. In practice, the lack of a dynamic assessment model makes it difficult to generate accurate warning instructions based on real-time monitoring data, significantly compromising the timeliness and accuracy of attitude adjustments.

[0007] In competitive sports, a kayak's stability directly impacts an athlete's performance; even minor deviations in posture can lead to significant differences in performance. In leisure and adventure activities, kayak safety is crucial to the user's life. Therefore, achieving real-time, accurate monitoring of a kayak's posture and dynamically adjusting it based on the results has become a critical challenge in the development of kayak technology.

[0008] Among existing technical solutions, some monitoring systems incorporate sensors to collect attitude parameters, but lack specificity in model construction. The same assessment model can be applied to kayaks of varying structures, leading to significant deviations between the assessment results and actual conditions. Furthermore, these systems fail to correlate attitude angles with hydrodynamic loads, preventing them from forming a multi-dimensional risk assessment system. This results in a lack of comprehensiveness and scientificity in the generation of early warning instructions. Furthermore, during the instruction execution phase, the adjustment device responds slowly, making it difficult to quickly and accurately adjust according to dynamic early warning instructions, and unable to effectively respond to sudden changes in attitude. Summary of the Invention

[0009] The object of the present invention is to provide a kayak hull posture monitoring method and monitoring system to solve the problems raised in the above background technology.

[0010] To achieve the above object, the present invention provides a method for monitoring the posture of a kayak, the method comprising: Collecting real-time environmental parameters of the target water area and dynamic parameters of the kayak during kayaking to generate an initial monitoring parameter set, which includes a three-dimensional attitude angle sequence and a hydrodynamic load distribution curve; Determining a first evaluation model adapted to the hull structure based on the three-dimensional attitude angle sequence in the initial monitoring parameter set, wherein the first evaluation model includes a dynamic mapping relationship between attitude angles and capsizing risks; determining, based on a hydrodynamic load distribution curve in the initial monitoring parameter set, a second evaluation model adapted to the hull structure, wherein the second evaluation model includes a nonlinear association rule between load distribution and stability coefficient; activating a target model in the first evaluation model or the second evaluation model based on the real-time posture deviation data and load fluctuation data during the kayak's travel, and generating a dynamic warning instruction through the target model; The dynamic warning instruction is input into the execution unit of the hull attitude adjustment device to correct the deflection angle of the balance wing or the water distribution of the ballast water tank.

[0011] Preferably, the collecting of real-time environmental parameters of the target water area and dynamic parameters of the kayak during kayaking to generate an initial monitoring parameter set includes: Synchronously collecting environmental interference characteristic data of the target water area during the driving phase through a multimodal sensor, wherein the environmental interference characteristic data includes wave spectrum distribution and water flow velocity vector; Dividing the target water area into at least two sub-water areas according to the energy peak interval in the wave spectrum distribution, and allocating a corresponding initial attitude angle threshold and load reference value to each sub-water area; Extract historical attitude compensation parameters and load correction coefficients matching each sub-water section from the preset database; Performing spatiotemporal filtering on the historical attitude compensation parameters to generate an optimized three-dimensional attitude angle sequence corresponding to each sub-water section; Performing frequency domain decomposition processing on the load correction coefficient to generate an optimized hydrodynamic load distribution curve corresponding to each sub-water section; The initial monitoring parameter set is generated by fusing the optimized three-dimensional attitude angle sequence and the optimized hydrodynamic load distribution curve of each sub-water section.

[0012] Preferably, determining the first evaluation model and the second evaluation model includes: Inputting the optimized three-dimensional attitude angle sequence into a preset fuzzy logic model, iteratively updating the membership function through a multi-rule base, and generating a capsizing risk assessment function in the first assessment model, wherein the capsizing risk assessment function is a target model in the first assessment model; Inputting the optimized hydrodynamic load distribution curve into an extreme learning machine model, adjusting the hidden layer neuron parameters using a particle swarm algorithm, and generating a stability coefficient mapping table in the second evaluation model; the stability coefficient mapping table is the target model in the second evaluation model; After the fuzzy logic model and the extreme learning machine model converge, extracting key feature matrices from the overturning risk assessment function and the stability coefficient mapping table respectively; Performing similarity matching between the key feature matrix and the environmental interference feature data collected in real time to verify the adaptability of the first evaluation model and the second evaluation model; When the similarity matching result is lower than a preset threshold, the training sample sets of the fuzzy logic model and the extreme learning machine model are readjusted until the key feature matrix meets the adaptability condition.

[0013] Preferably, activating the target model in the first evaluation model or the second evaluation model based on the real-time posture deviation data and load fluctuation data during the kayak's travel, and generating a dynamic warning instruction through the target model includes: Real-time monitoring of the accumulated posture deviation and load fluctuation frequency during kayaking; When the accumulated amount of the posture deviation exceeds a first alarm threshold and the load fluctuation frequency is within a preset safety range, activating the overturning risk assessment function in the first assessment model; generating a dynamic deflection adjustment instruction for the stabilizer according to the risk level classification rule in the overturning risk assessment function; When the load fluctuation frequency exceeds a second alarm threshold and the accumulated amount of posture deviation is within a preset safety range, activating the stability coefficient mapping table in the second evaluation model; generating a dynamic water volume adjustment instruction for the ballast water tank according to the load distribution rule in the stability coefficient mapping table; If the accumulated attitude deviation and the load fluctuation frequency both exceed the alarm threshold, the dynamic deflection adjustment instruction generated by the first evaluation model is executed first, and the execution of the adjustment instruction of the second evaluation model is delayed until the balance wing completes the correction operation.

[0014] Preferably, the step of inputting the dynamic warning instruction to the execution unit of the hull attitude adjustment device to correct the deflection angle of the stabilizer or the water distribution of the ballast tank includes: Adjusting the instantaneous deflection angle of each control surface in the stabilizer in stages according to the angle compensation value in the dynamic deflection adjustment instruction; After each angle adjustment, real-time feedback data of the ship's attitude is collected and deviation calculation is performed with the predicted value of the capsizing risk assessment function; If the deviation value continues to decrease, the current deflection angle is maintained and the direction is adjusted until the target posture range is reached; If the deviation value shows an expanding trend, the deflection angle is adjusted in the reverse direction and the parameter iteration of the overturning risk assessment function is retriggered; Dynamically adjust the water storage capacity of the ballast water tank in different areas according to the water distribution parameters in the dynamic water volume adjustment instruction; During the water volume adjustment process, the change in the hydrodynamic load is detected in real time by the pressure sensor, and the correction coefficient in the stability coefficient mapping table is dynamically updated according to the detection result.

[0015] Preferably, the method further includes a feedback calibration phase after the dynamic warning instruction is executed, including the following operations: Collect the final posture distribution data and stability coefficient detection map after the kayak is completed; Comparing the final posture distribution data with the predicted posture range of the first evaluation model to generate a posture evaluation error signal; Performing an overlap analysis on the stability coefficient detection map and the expected coefficient template of the second evaluation model to generate a load evaluation error signal; adjusting the risk level classification rules in the first assessment model according to the system error component in the posture assessment error signal; Optimizing a stability coefficient correction coefficient in the second evaluation model according to a random error component in the load evaluation error signal; The adjusted risk level classification rules and stability coefficient correction coefficient are synchronously updated to the historical parameter library of the initial monitoring parameter set.

[0016] Preferably, the process of adjusting the risk level classification rules and the stability coefficient correction coefficient includes: Identifying a steady-state deviation component in the posture assessment error signal and calculating a steady-state compensation amount by exponential smoothing; adjusting a reference risk threshold in the capsizing risk assessment function according to the steady-state compensation amount; Identifying high-frequency interference components in the load assessment error signal and extracting effective correction components through adaptive filtering; adjusting the load distribution weight in the stability coefficient mapping table according to the effective correction component; The updated overturning risk assessment function and stability coefficient mapping table replace the original model parameters.

[0017] Preferably, the method further comprises performing the following pre-processing operations before starting the device, including: Analyze the material stiffness coefficient and center of gravity position coding in the design parameters corresponding to the hull structure to generate a structural feature description vector; Inputting the structural feature description vector into a preloaded ship type characteristic database for multi-dimensional similarity search, and screening out a set of candidate reference templates whose similarity with the current design parameters exceeds a matching threshold; For each candidate reference template in the candidate reference template set, the following operations are performed: extracting the average attitude deviation value and stability compliance rate in its historical navigation records, and calculating a comprehensive navigation effectiveness score; Prioritizing the candidate reference template set according to the comprehensive navigation effectiveness score, and selecting the candidate reference template with the highest score as the optimal reference attitude template; Extracting a hydrodynamic load reference curve set that is associated with the optimal reference attitude template from the ship type characteristic database; Verifying the synergy between load distribution and attitude parameters for each curve in the hydrodynamic load reference curve set, eliminating abnormal curves with sudden load changes or attitude conflicts, and generating an optimized load reference curve set; According to the historical navigation stability index of each optimized load reference curve in the optimized load reference curve set, the optimized load reference curve with the smallest fluctuation coefficient is selected as the optimal load reference curve; The optimal reference attitude template and the optimal load reference curve are aligned in navigation time sequence to generate a reference configuration of the initial monitoring parameter set.

[0018] Preferably, the verification of the synergy between the load distribution and the attitude parameters for each curve in the hydrodynamic load reference curve set includes: Extracting load distribution data of a single curve to be verified from the hydrodynamic load reference curve set, and synchronously obtaining a posture parameter sequence in the optimal reference posture template that is time-aligned with the curve to be verified; According to the phase nodes of the posture parameter sequence, a corresponding collaborative timestamp identifier is marked on the curve to be verified to generate a load distribution curve with a time sequence mark; Traversing each collaborative timestamp identifier in the load distribution curve with time series mark, detecting whether the load change slope in the adjacent time interval exceeds a preset mutation threshold, and identifying abnormal time intervals with load mutation; When an abnormal time interval is identified, the posture parameter value of the corresponding time node in the optimal reference posture template is traced back to determine whether the change direction of the posture parameter value generates reverse interference with the load mutation direction; If the reverse interference intensity exceeds the conflict threshold, the abnormal time interval is marked as a posture conflict area, and the starting position and duration of the posture conflict area in the load distribution curve are calculated; Delineating a load correction interval on the curve to be verified according to the starting position and duration of the posture conflict area, and generating an alternative load smoothing segment based on correction records of similar conflicts in a historical parameter library; Inserting the alternative load smoothing segment into the load correction interval to generate an optimized load distribution curve, and deleting abnormal data points in the original load distribution curve that overlap with the posture conflict area; Performing integrity check on all optimized load distribution curves that have completed the substitution insertion operation in the hydrodynamic load reference curve set, and eliminating residual curves that still contain uncorrected conflict areas; The optimized load distribution curves that have passed the verification are merged into the optimized load reference curve set.

[0019] Preferably, the present invention also includes a kayak hull posture monitoring system, the system including a memory and a processor, the memory storing a computer program that can be run on the processor, and the processor implementing the steps of the above-mentioned kayak hull posture monitoring method when executing the program.

[0020] Compared with the prior art, the present invention has the following beneficial effects: From a data acquisition perspective, the invention simultaneously collects real-time environmental parameters of the target waters and dynamic parameters of the kayak, generating an initial set of monitoring parameters consisting of a three-dimensional attitude angle sequence and a hydrodynamic load distribution curve. This comprehensive parameter collection method breaks the limitation of traditional monitoring methods that focus on a single parameter, and can more comprehensively reflect the kayak's environmental conditions and its own motion state, providing rich and detailed basic data for subsequent assessment and early warning.

[0021] In constructing the assessment model, this method determines a first and second assessment model tailored to the ship's structure based on the different parameters in the initial monitoring parameter set. The first assessment model includes a dynamic mapping between attitude angle and capsizing risk, while the second assessment model incorporates nonlinear association rules between load distribution and stability coefficient. This approach of constructing dedicated assessment models for different parameters allows the models to better adapt to the ship's structure, improving the accuracy and specificity of the assessment. Compared with traditional general models, this avoids assessment bias caused by a mismatch between the model and the ship, and more accurately reflects the ship's actual condition.

[0022] In terms of dynamic warning and adjustment, this method uses real-time attitude deviation and load fluctuation data to activate the corresponding target model and generate dynamic warning instructions, which then control the hull attitude adjustment device to make corrections. This real-time response mechanism can react immediately to any attitude deviation or load fluctuation, issuing timely warnings and making adjustments. Whether adjusting the deflection angle of the stabilizer or optimizing the water distribution in the ballast tanks, the ship can be quickly affected, effectively avoiding safety hazards caused by untimely adjustments.

[0023] This method demonstrates excellent adaptability to diverse water environments and driving conditions. In calm waters, it ensures a stable kayak through precise posture monitoring and fine-tuning. In complex waters, faced with sudden current changes or wind and wave impacts, it rapidly activates the appropriate assessment model, generates effective warning instructions, and executes adjustments to maintain hull stability. This adaptability enables the kayak to maintain optimal driving conditions in a variety of environments, expanding its applicability.

[0024] This invention directly inputs dynamic warning commands into the actuator unit of the hull attitude adjustment device, achieving an integrated process of monitoring, evaluation, warning, and adjustment. This integrated operation reduces information transmission delays and errors in intermediate links, improving the overall system's response speed and adjustment accuracy. Compared with traditional manual intervention or segmented operation, it better meets the real-time and accuracy requirements of kayaking at high speeds and in complex environments, thereby enhancing the safety and reliability of kayaking. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a working principle diagram of the kayak hull posture monitoring method according to the present invention; Figure 2 Flowchart generated for the initial set of monitoring parameters; Figure 3 Flowchart for the feedback calibration phase; Figure 4 Flowchart for model parameter adjustment; Figure 5 Flowchart of the preprocessing operation. DETAILED DESCRIPTION

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0027] See also Figure 1-Figure 5 The present invention provides a method for monitoring the posture of a kayak, the method comprising: During kayaking, real-time environmental parameters of the target water area and dynamic parameters of the kayak are collected to generate an initial set of monitoring parameters, including a 3D attitude angle sequence and a hydrodynamic load distribution curve. Specifically, a sensor array deployed at different locations on the kayak hull simultaneously collects environmental data such as waves and currents, as well as dynamic data such as the kayak's tilt angle and angular velocity. After preprocessing, these data are integrated into the initial set of monitoring parameters. The 3D attitude angle sequence reflects the real-time changes in the kayak's attitude in three-dimensional space, and the hydrodynamic load distribution curve characterizes the distribution of water forces acting on different parts of the kayak hull.

[0028] Based on the 3D attitude angle sequence in the initial monitoring parameter set, a first assessment model adapted to the ship's structure is determined. This first assessment model includes a dynamic mapping relationship between attitude angles and capsizing risk. Specifically, based on the ship's structural parameters (such as length, width, and center of gravity), a first assessment model suitable for the current ship type is selected or trained from a library of preset models. This model dynamically calculates the corresponding capsizing risk based on the input 3D attitude angle sequence.

[0029] Based on the hydrodynamic load distribution curve in the initial monitoring parameter set, a second assessment model adapted to the hull structure is determined. The second assessment model incorporates a nonlinear association rule between the load distribution and the stability coefficient. Similarly, based on the hull structural characteristics, a second assessment model is constructed or invoked. This model analyzes the variation pattern of the hydrodynamic load distribution curve and establishes a nonlinear association between the load distribution and the stability coefficient, thereby achieving a quantitative assessment of the hull stability.

[0030] Based on real-time posture deviation data and load fluctuation data during kayaking, the target model in the first or second assessment model is activated, and dynamic warning instructions are generated using the target model. During kayaking, posture deviation data (the difference between the actual and ideal posture angles) and load fluctuation data (the deviation between the actual and baseline loads) are calculated in real time. When these data meet preset activation conditions, the corresponding target model is activated, and the target model generates dynamic warning instructions based on the calculation results, including warning levels and adjustment recommendations.

[0031] The dynamic warning command is input into the actuator unit of the hull attitude adjustment device to adjust the deflection angle of the stabilizer or the water distribution in the ballast tank. After receiving the dynamic warning command, the actuator unit drives the actuator mechanism of the stabilizer to adjust its deflection angle or controls the water pump and other equipment in the ballast tank to change the water distribution, thereby achieving real-time correction of the hull attitude and maintaining the kayak's stable operation.

[0032] Example 1: The environmental interference characteristic data of the target water area during the driving phase are synchronously collected through multimodal sensors, wherein the environmental interference characteristic data includes wave spectrum distribution and water flow velocity vector. The multimodal sensor includes a high-frequency wave radar installed at the bow of the kayak, underwater acoustic Doppler current meters on both sides of the hull, and a meteorological sensor at the stern. The high-frequency wave radar captures the vertical displacement and propagation direction of the waves at a sampling frequency of 10 times per second, and generates a wave spectrum distribution after signal processing. The distribution can clearly show the energy proportion of waves of different frequencies; the acoustic Doppler current meter measures the water flow velocity from the horizontal and vertical dimensions to form a water flow velocity vector containing east, north and vertical components; the meteorological sensor assists in collecting wind speed and wind direction data as a supplementary reference for the environmental interference characteristic data.

[0033] Based on the energy peak interval in the wave spectrum distribution, the target water area is divided into at least two sub-water segments, and each sub-water segment is assigned a corresponding initial attitude angle threshold and load reference value. For example, when there are two significant energy peaks in the wave spectrum distribution, located in the 0.5-1Hz and 3-5Hz intervals, the target water area is divided into a low-frequency wave segment and a high-frequency wave segment. For the low-frequency wave segment, due to its long wave period and large amplitude, the assigned initial attitude angle threshold is set to ±5°, and the load reference value is set to 1.2 times the rated load of the hull. For the high-frequency wave segment, due to its short wave period and small amplitude, the assigned initial attitude angle threshold is set to ±3°, and the load reference value is set to 0.9 times the rated load of the hull. During the division process, by monitoring changes in the wave spectrum distribution in real time, the boundaries and number of sub-water segments are dynamically adjusted when the energy peak interval shifts.

[0034] The historical attitude compensation parameters and load correction coefficients that match each sub-water section are extracted from the preset database. The preset database stores navigation records in different water environments over the past three years. Each record contains the corresponding water characteristic parameters, attitude compensation parameters and load correction coefficients. During extraction, the K-nearest neighbor algorithm is used to calculate the similarity between the wave spectrum distribution and water velocity vector of the current sub-water section and the historical data in the database. The top 20 records with the highest similarity are selected as matching samples, and then the historical attitude compensation parameters and load correction coefficients are extracted from these samples. The historical attitude compensation parameters include the roll, pitch and bow angle compensation values ​​at different times, and the load correction coefficient covers the distribution coefficient along the length of the hull and the pressure coefficient in the vertical direction.

[0035] The historical attitude compensation parameters are subjected to spatiotemporal filtering to generate an optimized three-dimensional attitude angle sequence corresponding to each sub-water section. The spatiotemporal filtering process uses a two-dimensional Kalman filter algorithm. In the time dimension, a sliding average filter is performed on the historical attitude compensation parameters with a time window of 0.1 seconds to eliminate the interference caused by instantaneous fluctuations. In the spatial dimension, the hull is divided into three regions: the bow, the middle of the ship, and the stern, taking into account the structural dimensions of the kayak. The attitude data collected by sensors in different regions are weighted and fused. The weight is set according to the degree of influence of the region on the hull balance, and the weight of the middle of the ship is higher than that of the bow and stern. After spatiotemporal filtering, the attitude compensation parameters that originally had jumps become smooth and continuous, forming an optimized three-dimensional attitude angle sequence that includes the roll angle, pitch angle, and bow angle that change with time. The time resolution of the sequence is consistent with the sensor sampling frequency.

[0036] The load correction coefficient is subjected to frequency domain decomposition processing to generate the optimized hydrodynamic load distribution curve corresponding to each sub-water section. The frequency domain decomposition processing first converts the load correction coefficient to the frequency domain through fast Fourier transform to obtain the amplitude and phase information of different frequency components. Then, according to the energy peak interval in the wave spectrum distribution, the load frequency components that match the frequency of this interval are retained, and other high-frequency noise and low-frequency interference are filtered out. For example, for the low-frequency wave sub-water section, the load frequency components in the range of 0.5-1Hz are retained; for the high-frequency wave sub-water section, the load frequency components in the range of 3-5Hz are retained. The processed frequency domain data is then converted back to the time domain through inverse Fourier transform to obtain the optimized hydrodynamic load distribution curve corresponding to each sub-water section. The curve uses the hull length as the horizontal axis and the load size as the vertical axis to clearly show the distribution of the hydrodynamic load on the hull and its changes over time.

[0037] Based on the optimized three-dimensional attitude angle sequence and optimized hydrodynamic load distribution curve of each sub-water section, the initial monitoring parameter set is fused and generated. The fusion process adopts a timestamp-based synchronous alignment method to arrange the optimized three-dimensional attitude angle sequence and optimized hydrodynamic load distribution curve of different sub-water sections in chronological order to ensure that the attitude parameters and load parameters at the same moment correspond one to one. For the connection of sub-water sections, a smooth transition algorithm is used to handle parameter mutations so that the parameter changes of adjacent sub-water sections are continuous. The fused initial monitoring parameter set is stored in the form of data frames. Each frame of data contains the three-dimensional attitude angle value at that moment, the characteristic point data of the hydrodynamic load distribution curve, and the corresponding timestamp and sub-water section identifier, providing complete and coherent basic data for subsequent model evaluation.

[0038] Example 2: The optimized three-dimensional attitude angle sequence is input into a preset fuzzy logic model. The membership function is iteratively updated through a multi-rule base to generate a capsizing risk assessment function in the first assessment model. The fuzzy logic model contains three input variables: roll angle deviation, pitch angle deviation, and attitude angle change rate. The output variable is the capsizing risk. The multi-rule base includes a static balance rule base, a dynamic transition rule base, and an extreme attitude rule base. The static balance rule base is for scenarios where the attitude angle is stable within a small range, the dynamic transition rule base is suitable for processes where the attitude angle changes rapidly, and the extreme attitude rule base corresponds to situations approaching the danger threshold. The initial membership function adopts a triangular distribution. Based on the sample data in the optimized three-dimensional attitude angle sequence, the function's vertex coordinates and width parameters are iteratively adjusted using the least squares method. During each iteration, the deviation between the model output and the expected sample output under the current membership function is calculated. The function coverage is contracted or expanded according to the direction of the deviation until the deviation converges to the preset range. The generated capsizing risk assessment function can directly map the corresponding relationship between the attitude angle parameters and the capsizing risk.

[0039] The optimized hydrodynamic load distribution curve is input into the extreme learning machine model. The particle swarm algorithm (PSO) adjusts the parameters of the hidden layer neurons to generate a stability coefficient mapping table for the second evaluation model. The input layer nodes of the ELM model represent the characteristic values ​​of the hydrodynamic load distribution curve, including the maximum load value, the load center of gravity, and the uniformity of the load distribution. The output layer nodes represent the stability coefficient. The number of hidden layer neurons is adaptively determined based on the amount of sample data, initially set to three times the number of input layer nodes. The particle dimension of the PSO algorithm corresponds to the number of weights and bias parameters in the hidden layer neurons, with each particle representing a parameter combination. During algorithm initialization, particle positions are randomly generated. The fitness value is determined by calculating the model output error corresponding to each particle. The smaller the error, the higher the fitness value. During the iteration process, the particles adjust their flight direction and step size based on their own historical optimal position and the optimal position of the swarm, gradually approaching the optimal parameter combination. Iterations are terminated when the change in the optimal fitness value after 10 consecutive iterations is less than the set value. At this point, the correspondence between the stability coefficient output by the model and the load distribution characteristics is compiled into a stability coefficient mapping table.

[0040] After the fuzzy logic model and the extreme learning machine model converge, key feature matrices are extracted from the rollover risk assessment function and the stability coefficient mapping table, respectively. The key feature matrix extracted from the rollover risk assessment function includes the risk gradient, the coordinates of the risk mutation point, and the risk accumulation rate corresponding to different attitude angle intervals. The key feature matrix extracted from the stability coefficient mapping table covers the correlation strength between load parameters and stability coefficients, the sensitive range of coefficient changes, and the coefficient attenuation pattern under extreme loads. The dimensions of the key feature matrix are determined based on the model complexity to ensure that it fully represents the core characteristics of the model.

[0041] The key feature matrix is ​​matched against the real-time collected environmental interference feature data. This similarity matching uses the cosine similarity algorithm to calculate the cosine of the angle between the feature matrix and the environmental interference feature data in high-dimensional space. The environmental interference feature data is converted into vectors of equal dimension, whose elements include wave energy fraction, water velocity component, and water roughness. When the cosine similarity value is greater than 0.7, the model is considered suitable for the current environment. If it is less than 0.7, corrected samples from similar environments are retrieved from the historical database and added to the training sample set to retrain the model until the key feature matrix and environmental data meet the required match.

[0042] Real-time monitoring of the kayak's cumulative attitude deviation and load fluctuation frequency during travel. The cumulative attitude deviation is calculated by accumulating the difference between the continuously collected actual attitude angle and the reference attitude angle. The accumulation period is set to 5 seconds, and the cumulative result is updated every second. The load fluctuation frequency is determined by counting the number of times the load value exceeds the reference range per unit time. The reference range is set based on the load reference value of the current sub-water section with a fluctuation of 10%. The statistical period is consistent with the cumulative attitude deviation.

[0043] When the accumulated attitude deviation exceeds the first alarm threshold and the load fluctuation frequency is in the preset safety interval, the capsizing risk assessment function is activated. The first alarm threshold is determined based on the kayak's structural anti-capsulation capability and is usually set to 1.5 times the reference attitude angle range. The preset safety interval is when the load fluctuation frequency is less than 0.5 times per second. At this time, the load change has little impact on the stability of the hull, and the attitude deviation becomes the main source of risk. After activation, the capsizing risk assessment function receives the current attitude angle data in real time, outputs the corresponding capsizing risk, and generates dynamic deflection adjustment instructions for the balance wing based on the risk level. The risk levels are divided into three levels: mild, medium, and severe, corresponding to different deflection angle adjustment ranges.

[0044] When the load fluctuation frequency exceeds the second alarm threshold and the accumulated attitude deviation is within the preset safety interval, the stability coefficient mapping table is activated. The second alarm threshold is set to 1.5 times / second, and the preset safety interval is that the accumulated attitude deviation is less than 30% of the first alarm threshold. At this time, the attitude is basically stable. Frequent fluctuations in the load may cause structural fatigue or excessive local stress. After activating the stability coefficient mapping table, the dynamic water adjustment instructions for the ballast water tank are generated by querying the stability coefficient corresponding to the current load distribution characteristics. The instructions include the target water volume, water transfer rate and adjustment completion time for each compartment to ensure that the impact of load fluctuations is offset by water distribution.

[0045] If both the accumulated attitude deviation and the load fluctuation frequency exceed the alarm threshold, the dynamic deflection adjustment commands generated by the first assessment model are prioritized, and the adjustment commands from the second assessment model are delayed until the trimmer is corrected. The trimmer correction operation is clearly marked as complete when the accumulated attitude deviation falls within the safe range and the rate of change is less than 0.5° / second over three consecutive sampling periods. During this delayed execution, the adjustment commands from the second assessment model are temporarily stored in a buffer. Once the trimmer is corrected, they are executed sequentially in the order they were generated. Attitude parameters are monitored in real time during execution, and if new attitude risks emerge, execution is immediately interrupted and the first assessment model is reactivated.

[0046] Example 3: Based on the angle compensation value in the dynamic deflection adjustment command, the instantaneous deflection angle of each rudder surface in the trim wing is adjusted in stages. The trim wing consists of a bow trim wing and a stern trim wing, each of which contains two independent left and right rudder surfaces that can be deflected independently. The angle compensation value is the core parameter in the dynamic deflection adjustment command and represents the total angle correction required. The staged adjustment adopts a three-level progressive mode: the first stage adjusts the angle by 30% of the total compensation value, the second stage by 50%, and the third stage by 20%. The adjustment duration of each stage is determined by the current speed of the kayak. The faster the speed, the shorter the adjustment duration, ensuring a smooth change in the impact force of the water on the hull during the deflection of the rudder surface. For example, when the angle compensation value is 15° and the speed is relatively fast, the rudder surface deflects 4.5° in 0.5 seconds in the first stage, 7.5° in 0.8 seconds in the second stage, and 3° in 0.3 seconds in the third stage, for a total of 1.6 seconds. During the adjustment process, the encoder installed on the rudder drive motor provides real-time feedback on the actual deflection angle, which is compared with the command angle to ensure that the angle error in each stage does not exceed 0.5°.

[0047] After each angle adjustment, real-time feedback data on the ship's attitude is collected and compared with the predicted value of the capsize risk assessment function to calculate the deviation. Real-time feedback data on the ship's attitude is collected by a triaxial gyroscope and accelerometer mounted at the ship's center of gravity. It includes the real-time values ​​and rates of change of the roll and pitch angles, with a sampling frequency of 20 times per second. The predicted value of the capsize risk assessment function is the expected attitude angle calculated based on the current rudder deflection angle. This predicted value is dynamically updated during each adjustment phase. Deviation calculation uses an absolute value accumulation method, summing the differences between the real-time feedback data and the predicted value at the same time point to obtain the total deviation for that phase. For example, after the first phase of adjustment, if the differences between the real-time and predicted roll angles at five sampling points are 0.3°, -0.2°, 0.1°, -0.4°, and 0.2°, respectively, the total deviation is 1.2°.

[0048] If the deviation continues to decrease, the current deflection angle adjustment direction is maintained until the target attitude range is reached. A continuously decreasing deviation means that the total deviation value of two consecutive adjustment stages shows a decreasing trend, and the deviation value of the latter stage is less than 80% of the previous stage. This indicates that the current rudder deflection direction is consistent with the attitude correction requirements. There is no need to change the adjustment direction, and subsequent angle adjustments continue as planned. The target attitude range is determined by the kayak's design parameters. Typically, the roll angle is controlled within ±3° and the pitch angle is controlled within ±2°. When real-time feedback data remains within this range for three consecutive seconds, the target attitude range is determined to be reached and the stabilizer angle adjustment stops.

[0049] If the deviation value shows an increasing trend, the deflection angle is adjusted in the opposite direction and the parameter iteration of the capsize risk assessment function is retriggered. Deviation value expansion refers to the total deviation value of the current stage being greater than 120% of the previous stage, and the real-time deviation of three consecutive sampling points is increasing. At this time, the current adjustment stage is immediately stopped, and the rudder deflection angle is adjusted back to 50% of the previous stage. For example, if the original plan was to adjust from 4.5° to 12° (the second stage), it would be adjusted back to 2.25° if the deviation increases. At the same time, the attitude feedback data and deviation value from this adjustment process are input into the capsize risk assessment function, and the risk gradient parameter and attitude angle mapping relationship in the function are re-iterated and updated, so that the function can more accurately predict attitude changes under different deflection angles.

[0050] The ballast water tank's water storage capacity in different areas is dynamically adjusted based on the water distribution parameters in the dynamic water volume adjustment instructions. The ballast water tanks are divided into four independent compartments: the port bow compartment, the right bow compartment, the port stern compartment, and the right stern compartment. Each compartment is equipped with an independent water pump and level sensor. The water volume distribution parameters include the target liquid level for each compartment, the liquid level adjustment rate, and the water balance coefficient between adjacent compartments. During the adjustment process, the pump's operating power is dynamically adjusted based on the difference between the current liquid level and the target level. The larger the difference, the higher the power. The maximum power ensures that the full tank water volume is transferred within 5 minutes. For example, if the target liquid level is 60% of the tank volume and the current liquid level is 30%, the pump operates at 70% power. When the liquid level reaches 50%, the power is reduced to 30% to avoid overshoot.

[0051] During the water volume adjustment process, pressure sensors detect changes in hydrodynamic loads in real time, and the correction coefficients in the stability coefficient mapping table are dynamically updated based on the detection results. The pressure sensor collects the pressure of the water on the hull in real time, with a sampling frequency of 10 times per second. The change in hydrodynamic load is determined by calculating the difference between the average pressure values ​​in adjacent time periods. When the difference exceeds a certain percentage of the initial average value, it is determined that the load has changed significantly. The correction coefficients in the stability coefficient mapping table are used to adjust the stability coefficient corresponding to different load ranges. The calculation formula is:

[0052] in, represents the updated correction coefficient, represents the original correction coefficient before updating, represents the load influence factor, Represents the change in the average pressure value, Represents the initial pressure average value.

[0053] Example 4: Collecting the final attitude distribution data and stability coefficient detection map after the kayak has completed its journey. After the kayak has completed a voyage, the attitude data of the entire journey is exported through the attitude recorder built into the hull, and the roll angle, pitch angle and bow angle values ​​of the last 5 minutes are extracted from it to form the final attitude distribution data. These data are presented with time as the horizontal axis and angle value as the vertical axis, which can intuitively reflect the attitude stability before stopping. The stability coefficient detection map is generated by the data collected by the pressure sensor array installed on the bottom of the hull. The array contains 12 pressure sensors, evenly distributed on both sides of the central axis from the bow to the stern. The stability coefficients at different times are calculated based on the pressure values ​​recorded by the sensors, and then these coefficients are arranged in chronological order to form a continuous map. The map also marks the time node of each dynamic warning command execution.

[0054] The final attitude distribution data is compared with the predicted attitude range of the first assessment model to generate an attitude assessment error signal. The predicted attitude range of the first assessment model is preset before voyage based on water environment parameters and includes upper and lower limits for roll and pitch angles. For example, the predicted range for roll angle is ±4°, and for pitch angle is ±3°. During the comparison, the deviation between the angle value in the final attitude distribution data and the predicted range is calculated point by point. If the actual angle is within the predicted range, the deviation is recorded as 0; if it is outside the range, the absolute value of the excess is recorded. These deviation values ​​are combined in chronological order to form the attitude assessment error signal. The amplitude of the signal fluctuation reflects the degree of consistency between the model prediction and the actual attitude.

[0055] The stability coefficient detection graph and the expected coefficient template of the second assessment model are analyzed for overlap to generate a load assessment error signal. The expected coefficient template of the second assessment model is a standard curve generated based on historical data. This curve shows the ideal trend of the stability coefficient over time. For example, in calm waters, it should fluctuate slightly between 0.8 and 1.0. During the overlap analysis, the detection graph and the expected coefficient template are aligned on the time axis, and the coefficient difference at the same time point is calculated. The load assessment error signal is then generated based on the size and duration of the difference. If the coefficient in the detection graph is lower than the expected template for a long time, the error signal will show continuous negative fluctuations.

[0056] The risk classification rules in the first assessment model are adjusted based on the systematic error component in the attitude assessment error signal. Systematic error components refer to regular deviations in the attitude assessment error signal. For example, the actual roll angle value is consistently 1° lower than the upper limit of the predicted range throughout the entire voyage, indicating that the model's risk assessment of the roll angle is systematically biased toward overestimation. When adjusting the risk classification rules, the risk threshold corresponding to the roll angle is lowered by 1° to account for this systematic error. Originally, a roll angle of 5° triggered a medium risk, but after adjustment, it is now 4°, making the risk classification more consistent with the actual attitude.

[0057] The stability coefficient correction factor in the second assessment model is optimized based on the random error component in the load assessment error signal. Random error components are irregular fluctuations in the error signal. For example, at a certain moment, the actual value of the stability coefficient suddenly drops below the expected value by 0.2, then quickly recovers. Such fluctuations are often caused by instantaneous changes in water flow. When optimizing the stability coefficient correction factor, the frequency and magnitude of the random error component are statistically analyzed, and the correction factor is fine-tuned for time periods with frequent large errors. For example, during periods of turbulent water flow, the stability coefficient correction factor is increased by 0.1 to offset the impact of instantaneous fluctuations.

[0058] The steady-state deviation component in the attitude assessment error signal is identified and a steady-state compensation is calculated using exponential smoothing. A steady-state deviation component is a constant deviation in the error signal that persists for more than 5 minutes. For example, the actual pitch angle is consistently 0.5° higher than the predicted value. The exponential smoothing method calculates the compensation by assigning different weights to historical deviation data, with more recent data being given higher weights than older data. For example, the steady-state deviation data from the last three voyages is weighted 0.5, 0.3, and 0.2, respectively. These data are multiplied by the weights and then summed to obtain the steady-state compensation.

[0059] The baseline risk threshold in the rollover risk assessment function is adjusted based on the steady-state compensation. If the steady-state compensation is positive, indicating that the actual attitude is more stable than the model predicts, the baseline risk threshold is raised. For example, if a roll angle of 6° triggers a high risk, it may be raised to 7° after adjustment. If the compensation is negative, the baseline risk threshold is lowered, allowing the model to identify potential risks earlier.

[0060] Identify high-frequency interference components in the load assessment error signal and extract effective correction components through adaptive filtering. High-frequency interference components are rapid fluctuations in the error signal with frequencies exceeding 0.5 Hz. These fluctuations are often caused by transient disturbances such as wave impacts. Adaptive filtering adjusts filter parameters in real time to filter out high-frequency interference components while retaining low-frequency fluctuations that reflect system characteristics. These low-frequency fluctuations are effective correction components, such as the trend of a slowly decreasing stability coefficient over an extended period.

[0061] The load distribution weights in the stability factor mapping table are adjusted based on the effective correction component. If the effective correction component indicates that the effect of the load on the stability factor in a certain area is underestimated, for example, the actual effect of the change in the stern load on the factor is 20% greater than what is recorded in the mapping table, the load distribution weight in that area is increased by 20% so that the mapping table more accurately reflects the relationship between the load and the stability factor in different areas.

[0062] The adjusted risk classification rules and stability coefficient correction factors are synchronously updated to the historical parameter library of the initial monitoring parameter set. The historical parameter library is categorized by water area type and vessel type. During the update, a parameter subset matching the current kayak model and water area is found and the new rules and coefficients are replaced with the original data. The update time and reason are also recorded to facilitate subsequent tracing of the basis for adjusting different parameter versions. The updated parameter library will serve as the basis for generating the initial monitoring parameter set for the next voyage, allowing the model to continuously adapt to actual sailing conditions.

[0063] Example 5: Analyze the material stiffness coefficient and center of gravity position code in the design parameters corresponding to the hull structure to generate a structural feature description vector. The hull design parameters include the material stiffness coefficients such as the elastic modulus and Poisson's ratio of the hull material, and the center of gravity position code formed by the coordinate values ​​of the center of gravity in the three-dimensional coordinate system. For example, the material stiffness coefficient of a kayak is the elastic modulus 2.1×1 MPa, a Poisson's ratio of 0.3, and a center of gravity position encoding of (1.2m, 0m, 0.1m), corresponding to coordinates along the ship's length, width, and draft, respectively. During the analysis process, these parameters are converted to standardized values. The material stiffness coefficients are normalized to industry standard ranges, while the center of gravity position encoding retains the original coordinate values. Together, they form a five-dimensional structural feature description vector, with each element in the vector corresponding to the standardized value of a design parameter.

[0064] The structural feature description vector is input into a preloaded ship type characteristic database for multi-dimensional similarity search, which screens out candidate benchmark templates whose similarity with the current design parameters exceeds a matching threshold. The ship type characteristic database stores structural feature vectors and corresponding navigation parameter templates for hundreds of different kayak models. Multi-dimensional similarity search calculates similarity in two dimensions: material stiffness and center of gravity position. The material dimension is calculated using Euclidean distance, and the center of gravity dimension is calculated using Manhattan distance. The distance values ​​of the two dimensions are combined to obtain the overall similarity. The matching threshold is set to 0.8. When the overall similarity exceeds this value, the corresponding template is included in the candidate benchmark template set. For example, a candidate template with a similarity of 0.85 in material stiffness and 0.82 in center of gravity, for an overall similarity of 0.83, meets the screening criteria.

[0065] For each candidate benchmark template in the candidate benchmark template set, the following operations are performed: the average attitude deviation and stability compliance rate from its historical voyage records are extracted, and a comprehensive navigation effectiveness score is calculated. The average attitude deviation is the average difference between the actual and ideal attitude of the template over multiple voyages, and the stability compliance rate is the proportion of voyages during which the stability coefficient remains within the safe range. The comprehensive navigation effectiveness score is calculated using a weighted calculation, with the average attitude deviation having a weight of 0.4 and the stability compliance rate having a weight of 0.6. For example, if a candidate template has an average attitude deviation of 1.2°, corresponding to a score of 0.88 (smaller deviations give higher scores), and a stability compliance rate of 92%, corresponding to a score of 0.92, the comprehensive score is 0.88 × 0.4 + 0.92 × 0.6 = 0.904.

[0066] The candidate benchmark templates are prioritized based on their comprehensive navigation performance scores, and the highest-scoring candidate template is selected as the optimal benchmark posture template. Sorting is in descending order, with the highest-scoring template first. For example, if there are five templates in the candidate template set with scores of 0.904, 0.892, 0.876, 0.865, and 0.851, the template with a score of 0.904 is selected as the optimal benchmark posture template. This template contains the posture change curve and typical posture parameter range of this kayak model in different waters.

[0067] A set of hydrodynamic load reference curves that are navigationally correlated with the optimal baseline attitude template is extracted from the ship type characteristics database. The navigation correlation is determined based on the template's historical navigation records. A correlation is considered to exist when the matching degree of the navigation time, water environment, and attitude parameters corresponding to the two curves exceeds 0.7. For example, the attitude curve of the optimal baseline attitude template during a lake voyage is highly correlated with the hydrodynamic load curve recorded during that voyage. These correlated curves together form a set of hydrodynamic load reference curves, each of which is annotated with environmental parameters such as the corresponding water type, wind speed, and wave height.

[0068] The synergy between the load distribution and attitude parameters is verified for each curve in the hydrodynamic load reference curve set, and abnormal curves with sudden load changes or attitude conflicts are eliminated to generate an optimized load reference curve set. During the synergy verification, each load curve is aligned with the corresponding attitude parameter sequence in the optimal reference attitude template in time to check whether the load change matches the attitude adjustment. For example, when the attitude parameters show that the hull is tilting to the left, the corresponding load curve should show a trend of increasing load on the left side. If the opposite occurs, it is determined to be an attitude conflict. A sudden load change refers to a load change rate that exceeds the normal range within a certain period of time in the curve. For example, the load value suddenly increases from 500N to 1000N within 1 second. Such curves will also be eliminated, and the remaining curves will form the optimized load reference curve set.

[0069] Based on the historical stability indicators of each optimized load reference curve in the optimized load reference curve set, the optimized load reference curve with the smallest fluctuation coefficient is selected as the optimal load reference curve. Historical stability indicators include the curve's fluctuation amplitude, fluctuation frequency, and duration of stability. The fluctuation coefficient is calculated by calculating the ratio of the curve's standard deviation to its mean. A smaller ratio indicates a more stable load distribution. For example, if the optimized load reference curve set contains three curves with fluctuation coefficients of 0.12, 0.15, and 0.11, respectively, the curve with a fluctuation coefficient of 0.11 is selected as the optimal load reference curve.

[0070] The optimal baseline attitude template and the optimal load baseline curve are aligned in the sailing time sequence to generate the baseline configuration of the initial monitoring parameter set. The sailing time sequence alignment takes the kayak's departure time as the starting point, aligning the key time nodes in the attitude template (such as the start time of attitude adjustment) with the corresponding nodes in the load curve to ensure that the two are synchronized in the time dimension. The aligned baseline configuration includes the standard range of attitude parameters, the reference value of load distribution, and the corresponding relationship between the two in different sailing stages. For example, within 10 minutes after departure, the roll angle should be stable at ±2°, and the corresponding maximum hydrodynamic load should not exceed 800N. These data together constitute the benchmark of the initial monitoring parameter set.

[0071] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0072] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring the posture of a kayak, characterized in that: The method comprises: Collecting real-time environmental parameters of the target water area and dynamic parameters of the kayak during kayaking to generate an initial monitoring parameter set, which includes a three-dimensional attitude angle sequence and a hydrodynamic load distribution curve; Determining a first evaluation model adapted to the hull structure based on the three-dimensional attitude angle sequence in the initial monitoring parameter set, wherein the first evaluation model includes a dynamic mapping relationship between attitude angles and capsizing risks; determining, based on a hydrodynamic load distribution curve in the initial monitoring parameter set, a second evaluation model adapted to the hull structure, wherein the second evaluation model includes a nonlinear association rule between load distribution and stability coefficient; activating a target model in the first evaluation model or the second evaluation model based on the real-time posture deviation data and load fluctuation data during the kayak's travel, and generating a dynamic warning instruction through the target model; The dynamic warning instruction is input into the execution unit of the hull attitude adjustment device to correct the deflection angle of the balance wing or the water distribution of the ballast water tank.

2. A kayak hull posture monitoring method according to claim 1, characterized in that: The real-time environmental parameters of the target water area and the dynamic parameters of the kayak are collected during the kayaking process to generate an initial monitoring parameter set, including: Synchronously collecting environmental interference characteristic data of the target water area during the driving phase through a multimodal sensor, wherein the environmental interference characteristic data includes wave spectrum distribution and water flow velocity vector; Dividing the target water area into at least two sub-water areas according to the energy peak interval in the wave spectrum distribution, and allocating a corresponding initial attitude angle threshold and load reference value to each sub-water area; Extract historical attitude compensation parameters and load correction coefficients matching each sub-water section from the preset database; Performing spatiotemporal filtering on the historical attitude compensation parameters to generate an optimized three-dimensional attitude angle sequence corresponding to each sub-water section; Performing frequency domain decomposition processing on the load correction coefficient to generate an optimized hydrodynamic load distribution curve corresponding to each sub-water section; The initial monitoring parameter set is generated by fusing the optimized three-dimensional attitude angle sequence and the optimized hydrodynamic load distribution curve of each sub-water section.

3. The method for monitoring the kayak hull posture according to claim 2, wherein: Determining the first evaluation model and the second evaluation model includes: Inputting the optimized three-dimensional attitude angle sequence into a preset fuzzy logic model, iteratively updating the membership function through a multi-rule base, and generating a capsizing risk assessment function in the first assessment model, wherein the capsizing risk assessment function is a target model in the first assessment model; Inputting the optimized hydrodynamic load distribution curve into an extreme learning machine model, adjusting the hidden layer neuron parameters using a particle swarm algorithm, and generating a stability coefficient mapping table in the second evaluation model; the stability coefficient mapping table is the target model in the second evaluation model; After the fuzzy logic model and the extreme learning machine model converge, extracting key feature matrices from the overturning risk assessment function and the stability coefficient mapping table respectively; Performing similarity matching between the key feature matrix and the environmental interference feature data collected in real time to verify the adaptability of the first evaluation model and the second evaluation model; When the similarity matching result is lower than a preset threshold, the training sample sets of the fuzzy logic model and the extreme learning machine model are readjusted until the key feature matrix meets the adaptability condition.

4. The method for monitoring the kayak hull posture according to claim 3, wherein: The method of activating a target model in the first evaluation model or the second evaluation model based on the real-time posture deviation data and load fluctuation data during the kayak's travel, and generating a dynamic warning instruction through the target model, includes: Real-time monitoring of the accumulated posture deviation and load fluctuation frequency during kayaking; When the accumulated amount of the posture deviation exceeds a first alarm threshold and the load fluctuation frequency is within a preset safety range, activating the overturning risk assessment function in the first assessment model; generating a dynamic deflection adjustment instruction for the stabilizer according to the risk level classification rule in the overturning risk assessment function; When the load fluctuation frequency exceeds a second alarm threshold and the accumulated amount of posture deviation is within a preset safety range, activating the stability coefficient mapping table in the second evaluation model; generating a dynamic water volume adjustment instruction for the ballast water tank according to the load distribution rule in the stability coefficient mapping table; If the accumulated attitude deviation and the load fluctuation frequency both exceed the alarm threshold, the dynamic deflection adjustment instruction generated by the first evaluation model is executed first, and the execution of the adjustment instruction of the second evaluation model is delayed until the balance wing completes the correction operation.

5. The method for monitoring the kayak hull posture according to claim 4, characterized in that: Inputting the dynamic warning instruction to the execution unit of the hull attitude adjustment device to correct the deflection angle of the stabilizer or the water distribution of the ballast tank includes: Adjusting the instantaneous deflection angle of each control surface in the stabilizer in stages according to the angle compensation value in the dynamic deflection adjustment instruction; After each angle adjustment, real-time feedback data of the ship's attitude is collected and deviation calculation is performed with the predicted value of the capsizing risk assessment function; If the deviation value continues to decrease, the current deflection angle is maintained and the direction is adjusted until the target posture range is reached; If the deviation value shows an expanding trend, the deflection angle is adjusted in the reverse direction and the parameter iteration of the overturning risk assessment function is retriggered; Dynamically adjust the water storage capacity of the ballast water tank in different areas according to the water distribution parameters in the dynamic water volume adjustment instruction; During the water volume adjustment process, the change in the hydrodynamic load is detected in real time by the pressure sensor, and the correction coefficient in the stability coefficient mapping table is dynamically updated according to the detection result.

6. The method for monitoring the kayak hull posture according to claim 1, characterized in that: The method further includes a feedback calibration phase after the dynamic warning instruction is executed, including the following operations: Collect the final posture distribution data and stability coefficient detection map after the kayak is completed; Comparing the final posture distribution data with the predicted posture range of the first evaluation model to generate a posture evaluation error signal; Performing an overlap analysis on the stability coefficient detection map and the expected coefficient template of the second evaluation model to generate a load evaluation error signal; adjusting the risk level classification rules in the first assessment model according to the system error component in the posture assessment error signal; Optimizing a stability coefficient correction coefficient in the second evaluation model according to a random error component in the load evaluation error signal; The adjusted risk level classification rules and stability coefficient correction coefficient are synchronously updated to the historical parameter library of the initial monitoring parameter set.

7. A kayak hull posture monitoring method according to claim 6, characterized in that: The process of adjusting the risk classification rules and stability factor correction factors includes: Identifying a steady-state deviation component in the posture assessment error signal and calculating a steady-state compensation amount by exponential smoothing; adjusting a reference risk threshold in the capsizing risk assessment function according to the steady-state compensation amount; Identifying high-frequency interference components in the load assessment error signal and extracting effective correction components through adaptive filtering; adjusting the load distribution weight in the stability coefficient mapping table according to the effective correction component; The updated overturning risk assessment function and stability coefficient mapping table replace the original model parameters.

8. The method for monitoring the kayak hull posture according to claim 1, wherein: The method further includes performing the following pre-processing operations before the device is started, including: Analyze the material stiffness coefficient and center of gravity position coding in the design parameters corresponding to the hull structure to generate a structural feature description vector; Inputting the structural feature description vector into a preloaded ship type characteristic database for multi-dimensional similarity search, and screening out a set of candidate reference templates whose similarity with the current design parameters exceeds a matching threshold; For each candidate reference template in the candidate reference template set, the following operations are performed: extracting the average attitude deviation value and stability compliance rate in its historical navigation records, and calculating a comprehensive navigation effectiveness score; Prioritizing the candidate reference template set according to the comprehensive navigation effectiveness score, and selecting the candidate reference template with the highest score as the optimal reference attitude template; Extracting a hydrodynamic load reference curve set that is associated with the optimal reference attitude template from the ship type characteristic database; Verifying the synergy between load distribution and attitude parameters for each curve in the hydrodynamic load reference curve set, eliminating abnormal curves with sudden load changes or attitude conflicts, and generating an optimized load reference curve set; According to the historical navigation stability index of each optimized load reference curve in the optimized load reference curve set, the optimized load reference curve with the smallest fluctuation coefficient is selected as the optimal load reference curve; The optimal reference attitude template and the optimal load reference curve are aligned in navigation time sequence to generate a reference configuration of the initial monitoring parameter set.

9. The method for monitoring the kayak hull posture according to claim 8, characterized in that: The verifying the synergy between load distribution and attitude parameters for each curve in the hydrodynamic load reference curve set includes: Extracting load distribution data of a single curve to be verified from the hydrodynamic load reference curve set, and synchronously obtaining a posture parameter sequence in the optimal reference posture template that is time-aligned with the curve to be verified; According to the phase nodes of the posture parameter sequence, a corresponding collaborative timestamp identifier is marked on the curve to be verified to generate a load distribution curve with a time sequence mark; Traversing each collaborative timestamp identifier in the load distribution curve with time series mark, detecting whether the load change slope in the adjacent time interval exceeds a preset mutation threshold, and identifying abnormal time intervals with load mutation; When an abnormal time interval is identified, the posture parameter value of the corresponding time node in the optimal reference posture template is traced back to determine whether the change direction of the posture parameter value generates reverse interference with the load mutation direction; If the reverse interference intensity exceeds the conflict threshold, the abnormal time interval is marked as a posture conflict area, and the starting position and duration of the posture conflict area in the load distribution curve are calculated; Delineating a load correction interval on the curve to be verified according to the starting position and duration of the posture conflict area, and generating an alternative load smoothing segment based on correction records of similar conflicts in a historical parameter library; Inserting the alternative load smoothing segment into the load correction interval to generate an optimized load distribution curve, and deleting abnormal data points in the original load distribution curve that overlap with the posture conflict area; Performing integrity check on all optimized load distribution curves that have completed the substitution insertion operation in the hydrodynamic load reference curve set, and eliminating residual curves that still contain uncorrected conflict areas; The optimized load distribution curves that have passed the verification are merged into the optimized load reference curve set.

10. A kayak hull posture monitoring system, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the program, the steps of the kayak hull posture monitoring method according to any one of claims 1 to 9 are implemented.

Images