Dynamic wind-resistant stability control system and method in roll-on platform ramp adjustment process

Abstract

This invention relates to the field of adaptive actuation technology for ramps, specifically to a dynamic wind-resistant stability control system and method for the ramp adjustment process of a roll-on / roll-off platform. In this invention, a fuzzy neural network is introduced to compare the extreme pressure values ​​of the accumulator, mitigating the phase lag caused by simple feedback control. By calling Gaussian process regression to deduce the state prediction sequence, boundary values ​​are extracted, constraint compression operations are performed, and the pump station's output rating is compared, avoiding the risk of trajectory deviation caused by extreme gusts leading to hydraulic components approaching saturation dead zones. The matrix is ​​decomposed to extract the hinge deformation residual vector, dynamically isolating the amplification effect of high-frequency measurement noise, improving the smoothness of state estimation and the robustness against disturbances. By calling the slope matrix to convert the duty cycle and adjust the servo valve opening, the large-scale articulated mechanism maintains a strict passage threshold range under the combined effects of severe wind and waves, preventing fatigue fracture failure of the mechanical structure due to local sudden load changes.

Description

Technical Field

[0001] This invention relates to the field of adaptive ramp actuation technology, and more particularly to a dynamic wind-resistant stability control system and method for the ramp adjustment process of a roll-on / roll-off platform. Background Technology

[0002] The field of ramp adaptive actuation technology specifically studies dynamic compensation schemes for large-scale rigid body posture in marine operating environments. It encompasses multiple related technology execution modules, including a hydraulic servo closed-loop drive module, a multi-body system kinematics calculation module, and a nonlinear disturbance suppression module. The execution technology standards define the steady-state error of system displacement control as limited to within ±5 mm, and the attitude angle adjustment overshoot as less than 2%. The core execution steps utilize a closed-loop control law constructed based on Lyapunov stability theory to calculate the Jacobian matrix of the spatial kinematics of multiple rigid ramp segments under nonholonomic constraints. This results in the output of multiple pulse width modulation duty cycle signals to control the opening degree of multiple hydraulic proportional valves, offsetting wave loads and applying multi-directional displacement deviations to wind disturbances.

[0003] The dynamic wind-resistant stability control system for the ramp adjustment of the roll-on / roll-off platform is specifically designed to construct a closed-loop attitude correction architecture for the dynamic articulation mechanism between the moving floating body and the fixed berthing end. The core objective of the system is to maintain the slope variable of the roll-on / roll-off platform and the working surface within the ±4 degree passage threshold range under Beaufort level 8 wind conditions. The system's operational effect is defined as reducing the amplitude of low-frequency roll and pitch oscillations induced by wind loads by more than 85%, and controlling the peak stress variable at the articulation joint to be lower than 30% of the material's yield strength quantification reference value, thereby preventing sudden gust wind loads from causing plastic deformation or even fracture failure of the platform's rigid locking mechanism.

[0004] Existing technologies heavily rely on Lyapunov stability theory to construct simple feedback closed-loop control laws. They depend on solving the Jacobian matrix of the kinematics in nonholonomic constrained spaces to output duty cycle signals, driving multi-channel hydraulic proportional valves to offset displacement deviations caused by wave and wind loads. However, the hysteresis error compensation operation mode exhibits significant response delays to sudden high-frequency wind pressure excitations in complex marine environments. The rigid kinematic solution framework fails to deeply analyze the large-scale physical hinge components, resulting in localized microscopic deformation residuals due to dynamic load intervention. Treating the platform as an absolutely rigid body leads to frequent full-opening and full-closing of the hydraulic servo closed-loop drive mechanism under extreme conditions, generating strong high-frequency vibrations. Vibration and transient high-load impacts can easily break through the ±5 mm steady-state error control line. During the period from the issuance of the command to the physical displacement of the hydraulic cylinder push rod, the huge wind pressure directly acts on the dynamic hinge joint between the platform and the dock end, causing a sharp increase in structural stress and generating low-frequency rolling oscillations. Relying solely on hydraulic actuation lacks a coordinated control mechanism with the tension distribution of flexible mooring cables, limiting the platform's multi-dimensional attitude correction process. Long-term exposure to alternating impact loads can easily cause metal fatigue accumulation in the rigid locking pin components, increasing the engineering risk of local plastic deformation and even overall hinge fracture failure, and reducing the continuity and safety margin of roll-on / roll-off operations in harsh sea conditions. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and to propose a dynamic wind-resistant stability control system and method for the adjustment process of the roll-on / roll-off platform ramp.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: the dynamic wind-resistant stability control system for the roll-on / roll-off platform ramp adjustment process includes: Spatial Deformation Analysis Module: Extracts joint pulse values ​​from the tilt angle signals of multiple rigid planks, performs spatial transformation calculations, accumulates the projection variables of the hinge points, calculates the extreme values ​​of the lap lip plate deviation and guide column slope, aggregates multi-axis displacement scalars, and generates a spatial pose state array. Wind load disturbance estimation module: Based on the spatial pose state array, extract the valve core displacement residual integral to solve the disturbance lumped deviation term, compare the accumulator pressure extreme value through a fuzzy neural network, determine the central numerical calculation torque compensation component and superimpose multiple state residual variables to obtain the feedforward yaw torque parameter; Steady-state trajectory optimization module: Based on the feedforward yaw moment parameter, it calls Gaussian process regression to calculate the state prediction sequence, extracts boundary values ​​to perform constraint contraction operation, compares the rated output of the pump station, solves the polynomial to extract the push rod displacement increment, and generates the contraction boundary thrust command flow; The deviation scaling module calculates the following: based on the feedforward yaw moment parameter and the shrinkage boundary thrust command flow, it decomposes the matrix to extract the hinge shaft deformation residual vector, matches the state interval to calculate the pad attenuation factor, extracts the constant and multiplies it by the factor to perform scaling adjustment, and constructs a gain multiplier array. The ramp servo stabilization module: Based on the gain multiplier array, it calls the slope matrix to convert the duty cycle, adjusts the servo valve opening, drives the pin to insert into the lock hole to perform rigid locking, outputs torque to adjust the mooring cable tension value and accumulates the coordinate difference to obtain multi-directional wind-resistant posture parameters.

[0007] As a further embodiment of the present invention, the spatial pose state array includes the overlap lip plate deviation, the guide post slope extreme value, and the multi-axis displacement scalar; the feedforward yaw moment parameter includes the torque compensation component, the projection vector spatial deviation, and the yaw axis coordinate difference scalar; the contraction boundary thrust command flow includes the state prediction sequence, the action threshold scalar, and the push rod displacement increment; the gain multiplier array includes the pad attenuation factor, the dynamic scaling ratio, and the servo gain weight; and the multi-directional wind-resistant pose parameter includes the drive locking displacement scalar, the mooring cable tension value, and the longitudinal and transverse tilt angle deviation.

[0008] As a further aspect of the present invention, the spatial deformation analysis module includes: Vector space projection submodule: Based on the tilt angle signals of multiple rigid planks, extract the joint pulse values ​​and multiply them by the angle cosine components to construct a spatial coordinate basis. Accumulate multiple projection differences and merge the displacement coordinate values ​​of the hinge points. Transform the displacement offset components of each three-dimensional axis to generate the node displacement mapping. Scalar aggregation and deduction submodule: Based on the node displacement mapping, the axial parameters are stripped and the reference coordinates are subtracted to split the multi-dimensional vector, the endpoint vector deviation value is calculated, the overlap lip plate deviation value is extracted and multiplied by the extreme value of the guide post slope, and multiple axial displacement scalars are merged to generate a spatial pose state array.

[0009] As a further aspect of the present invention, the wind load disturbance estimation module includes: The residual integral solving submodule: Based on the spatial pose state array, extract the servo valve core displacement residual values, multiply them by the time step span, accumulate the multinomial state residual gradient, calculate the cumulative area of ​​the multidimensional parameter time domain integral, merge the multinomial partial derivative constants, and obtain the lumped disturbance deviation term. Extreme value comparison and judgment submodule: Introduces a fuzzy neural network to compare the extreme values ​​of accumulator pressure, extract the effective values ​​within the interval, assign deviation weight parameters to select the central node variables, extract the equivalent cross-sectional area and lever arm length components of the node scalar multiplication, and generate the torque compensation benchmark value. Compensation superposition calculation submodule: Based on the torque compensation benchmark value, extract the attitude angle deviation values, accumulate various state residual variables, solve the projection vector space bias and merge the yaw axis coordinate difference scalar, superimpose multiple torque correction bases in the feedback channel, and obtain the feedforward yaw torque parameters.

[0010] As a further aspect of the present invention, the fuzzy neural network extracts the lumped disturbance deviation term and inputs it into the network mapping node. It then performs fuzzy calculations using multiple membership functions to generate a fuzzy variable matrix. The fuzzy variable matrix is ​​passed to the network hidden layer to trigger logical reasoning rules. Multiple node activation parameters are calculated, and the multiple node activation parameters are compared with the extreme boundary conditions of the accumulator pressure. Out-of-bounds parameters are removed, and valid values ​​within the extreme value range are extracted.

[0011] As a further aspect of the present invention, the steady-state trajectory optimization module includes: State sequence prediction submodule: Based on the feedforward yaw moment parameter, Gaussian process regression is called to extract the cylinder extreme value multiplication control parameter, the pose deflection value within the step size is accumulated and the inference set is split into discrete nodes, the trajectory inference vector scalar is extracted, and the state prediction evolution sequence is generated. Extreme value constraint compression submodule: Based on the state prediction evolution sequence, extract the extreme values ​​of the prediction residuals to extract discrete invariant set constants, shrink the boundary conditions of the input torque constraint space, superimpose the limit values ​​of the lifting mechanism to limit the action threshold scalar, and obtain the displacement constraint boundary parameters; Target command calculation submodule: Based on the displacement constraint boundary parameters, compare the rated flow output of the pump station to eliminate out-of-bounds deviation terms, calculate the valley value deviation by solving the first derivative of the objective function, extract the push rod displacement increment and merge the drive sequence to generate the contraction boundary thrust command flow.

[0012] As a further aspect of the present invention, the Gaussian process regression extracts the feedforward yaw moment parameter into the process regression prior mean function, calls the kernel function to calculate the feature covariance matrix, extracts the cylinder extreme value multiplication control parameter to construct the observation data sample, performs posterior probability distribution calculation and deduction and accumulates the pose deflection value within the step size, solves the multidimensional joint probability function to generate the deduction set, and decomposes the covariance feature array to split discrete nodes.

[0013] As a further aspect of the present invention, the measurement deviation scaling module includes: Matrix residual decomposition submodule: Based on the feedforward yaw moment parameter and the contraction boundary thrust command flow, decompose and extract the hinge shaft deformation residual vector, match the interval separation boundary extreme values, accumulate and merge the hinge error, and generate the state interval residual vector; Multiplier scaling adjustment submodule: Based on the residual vector of the state interval, extract the pad block residual matching attenuation factor, multiply it by the weighted constant to adjust the dynamic scaling ratio, correct multiple servo gain weights, and construct a gain multiplier array.

[0014] As a further aspect of the present invention, the ramp servo stabilization module includes: Servo pin locking submodule: Based on the gain multiplier array, extract the slope matrix to convert the voltage duty cycle, adjust the opening of the proportional servo valve, drive the guide pin to penetrate the hole for rigid locking, separate the multi-axis displacement deviation values, and generate the driving locking displacement scalar. Mooring tension adjustment submodule: Based on the drive locking displacement scalar, output torque constant to adjust the mooring cable tension value, accumulate the displacement difference of each axis in multi-dimensional space, merge the longitudinal and transverse tilt angle deviations, extract the attitude space offset component, and obtain multi-directional wind-resistant attitude parameters.

[0015] A dynamic wind-resistant stability control method for the ramp adjustment process of a roll-on / roll-off platform, wherein the dynamic wind-resistant stability control method for the ramp adjustment process of a roll-on / roll-off platform is executed based on the aforementioned dynamic wind-resistant stability control system for the ramp adjustment process of a roll-on / roll-off platform, and includes the following steps: S1: Based on the tilt angle signals of multiple rigid springboards, read the joint pulses and substitute them into the matrix for calculation. Add the differences of multiple hinge projections, derive the lip plate deviation and guide column tilt parameters, merge the multi-directional linear displacement scale, and generate a spatial pose state array. S2: Based on the spatial pose state array, extract the displacement difference integral to calculate the interference residual, use a fuzzy neural network combined with the limit to extract the effective segment, select nodes to multiply the lever arm to derive the torque, summarize multiple residuals, and obtain the feedforward yaw torque parameter. S3: Based on the feedforward yaw moment parameter, call Gaussian process regression to calculate the motion trend, separate the extreme values ​​to reduce the boundary range, combine the flow to remove out-of-bounds terms, differentiate the function to extract the push rod extension length, and generate the shrinking boundary thrust command flow. S4: Based on the feedforward yaw moment parameter and the contraction boundary thrust command flow, decompose and peel off the hinge shaft deformation error, determine the operating condition mapping pad reduction coefficient, retrieve constants for multiplication to adjust the amplification ratio, and construct a gain multiplier array; S5: Based on the gain multiplier array, the waveform is mapped by the transformation matrix to change the cross section of the proportional valve, the through hole position of the push pin is fixed, the torque is issued to change the tension of the mooring cable, and the three-dimensional axial relative displacement difference is summarized to obtain the multi-directional wind-resistant attitude parameters.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by introducing a fuzzy neural network to compare the extreme values ​​of the accumulator pressure, the torque compensation component is determined by the central numerical calculation and the state residual variable is superimposed to obtain the feedforward yaw torque parameter. The feedforward compensation mechanism intervenes in the wind load disturbance evolution cycle in advance, weakening the phase lag drawback caused by simple feedback control. In this invention, by calling the Gaussian process regression to deduce the state prediction sequence, extracting boundary values ​​to perform constraint compression operations and comparing them with the rated output of the pump station, the risk of trajectory deviation caused by extreme gusts inducing hydraulic components to approach the saturation dead zone is avoided. The matrix is ​​decomposed to extract the hinge shaft deformation residual vector, dynamically isolating the high-frequency measurement noise amplification effect, and improving the smoothness of state estimation and the robustness against disturbances. In this invention, by calling the slope matrix to convert the duty cycle and control the opening of the servo valve, the pin is driven to insert into the lock hole to perform rigid locking. The torque is output simultaneously to adjust the mooring cable tension value and accumulate the coordinate difference value to obtain multi-directional wind-resistant posture parameters. This ensures that the large-scale articulated mechanism maintains a strict passage threshold range under the superposition of severe wind and waves, and prevents fatigue fracture failure of the mechanical structure caused by local sudden load. Attached Figure Description

[0017] Figure 1 This is a system flowchart of the present invention; Figure 2 This is a schematic diagram of the method steps of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] Example 1 Please see Figure 1 The present invention provides a technical solution: a dynamic wind-resistant stability control system for the adjustment process of a roll-on / roll-off platform ramp includes...

[0020] Spatial Deformation Analysis Module: Extracts joint pulse values ​​from the tilt angle signals of multiple rigid planks, performs spatial transformation calculations, accumulates the projection variables of the hinge points, calculates the extreme values ​​of the lap lip plate deviation and guide column slope, aggregates multi-axis displacement scalars, and generates a spatial pose state array. Wind load disturbance estimation module: Based on the spatial pose state array, extract the valve core displacement residual integral to solve the disturbance lumped deviation term, compare the accumulator pressure extreme value through fuzzy neural network, determine the central numerical calculation torque compensation component and superimpose multiple state residual variables to obtain the feedforward yaw moment parameter; Steady-state trajectory optimization module: Based on the feedforward yaw moment parameter, it calls Gaussian process regression to calculate the state prediction sequence, extracts boundary values ​​to perform constraint contraction operation, compares the rated output of the pump station, solves the polynomial to extract the push rod displacement increment, and generates the contraction boundary thrust command flow; The deviation scaling module calculates the following: Based on the feedforward yaw moment parameter and the shrinkage boundary thrust command flow, it decomposes the matrix to extract the hinge shaft deformation residual vector, matches the state interval to calculate the pad attenuation factor, extracts the constant and multiplies it by the factor to perform scaling adjustment, and constructs a gain multiplier array. The ramp servo stabilization module is based on a gain multiplier array. It calls the slope matrix to convert the duty cycle, adjusts the servo valve opening, drives the pin to insert into the lock hole to perform rigid locking, outputs torque to adjust the mooring cable tension value and accumulates the coordinate difference to obtain multi-directional wind-resistant posture parameters.

[0021] The spatial pose state array includes the overlap lip plate deviation, the extreme value of the guide post slope, and the multi-axis displacement scalar. The feedforward yaw moment parameters include the torque compensation component, the projection vector spatial deviation, and the yaw axis coordinate difference scalar. The contraction boundary thrust command flow includes the state prediction sequence, the action threshold scalar, and the push rod displacement increment. The gain multiplier array includes the pad attenuation factor, the dynamic scaling ratio, and the servo gain weight. The multi-directional wind-resistant pose parameters include the drive locking displacement scalar, the mooring cable tension value, and the longitudinal and transverse tilt angle deviation.

[0022] The spatial deformation analysis module includes: Vector space projection submodule: Based on the tilt angle signals of multiple rigid planks, extract the joint pulse values ​​and multiply them by the angle cosine components to construct a spatial coordinate basis. Accumulate multiple projection differences and merge the displacement coordinate values ​​of the hinge points. Transform the displacement offset components of each three-dimensional axis to generate the node displacement mapping. Scalar aggregation and deduction submodule: Based on the node displacement mapping, the axial parameters are stripped and the reference coordinates are subtracted to split the multi-dimensional vector, the endpoint vector deviation value is calculated, the overlap lip plate deviation value is extracted and multiplied by the extreme value of the guide post slope, and multiple axial displacement scalars are merged to generate a spatial pose state array. The vector space projection submodule, based on the tilt angle signals of multiple rigid scaffold segments, employs the Rodriguez rotation formula mapping algorithm to perform coordinate system alignment transformation on the tilt angle signals to be processed. The input variables include the unit vector of the scaffold rotation axis and a preset deflection angle. The preset deflection angle is obtained by reading a pre-programmed decimal value from the hardware motherboard's read-only memory during the system reset phase, and is specifically fixed at 45 degrees. The rotation mapping matrix dimension parameter is set to a 3x3 interlaced data grid. The cosine lookup table instruction within the underlying digital signal processor is invoked to match the corresponding cosine value component of the input signal. A 64-bit wide space is allocated to the cache register to temporarily store the cosine value component. The controller's global pulse equivalent parameter is set to 4096 pulses per revolution. A single-instruction multiple-data-stream multiplier is invoked to perform the underlying numerical multiplication operation, and the connector pulse value is extracted and multiplied by the angle cosine component. A spatial coordinate basis is constructed, and a polygon vector closure summation algorithm is used to perform cumulative and merged calculations on multiple projected coordinates. The state constant variable of the central processing unit's internal data summation accumulator is initialized to 0, the system clock sampling time step constant is configured to 10 microseconds, the memory direct access channel transmission block size parameter is configured to 256 bytes, the microcontroller global interrupt control switch is enabled to block external input, and the hardware data pointer is started to circulate through the memory physical address index of the multiple projection difference array. The underlying difference variable values ​​are extracted and superimposed one by one according to the sampling time step constant. The loop stop mechanism is triggered by comparing the current address cursor with the set boundary constant 512. The cumulative sum data is stored at the memory start address of the summation accumulator end, and the multiple projection difference merged hinge point displacement coordinate values ​​are accumulated. The displacement offset components of each three-dimensional axis are converted to generate the node displacement mapping. The scalar aggregation and deduction submodule, based on node displacement mapping, employs a tensor slicing dimensionality reduction and separation algorithm to perform multidimensional data matrix reduction, cutting, and stripping operations. The input data tensor angular dimension parameter constant is set to 4, the starting node index constant for the slicing operation is configured to be 0, and the ending node index constant is set to 3. Data cutting instructions are invoked to extract the bottom-level elements of the array within the node index range. Preset reference coordinate parameters are extracted. These preset reference coordinate parameters are obtained through the system power-on initialization process, starting the boot program and reading the fixed floating-point values ​​of a specific sector of the read-only memory chip. The horizontal axis value is explicitly configured as 10 mm, and the vertical axis value as 20 mm. Hardware matrix subtraction arithmetic logic unit instructions are executed to subtract the specified reference parameters, and the axial parameters are stripped by subtracting the reference coordinates to split the multidimensional vector. Double-precision floating-point modulo operation units are called to calculate the absolute distance between the corresponding two points in space. The components are calculated, and the endpoint vector deviation values ​​are calculated. The bubble sort extreme value search algorithm is used to perform the extreme value locking operation on the slope state dataset. The threshold parameter of the outer data traversal loop control variable is configured to be equal to the constant of the total length of the underlying data (50). The inner judgment loop is configured to perform logical comparison of the numerical values ​​of adjacent memory address storage elements. The hardware data exchange instruction is called to swap the physical pointers of memory address data according to the ascending order sorting rule. The highest value item corresponding to the last highest index sequence is directly extracted and assigned to the extreme value variable. The multiply-add instruction is called to directly multiply the extracted deviation value by the extreme value variable parameter, and the deviation value of the overlapping lip plate is extracted and multiplied by the extreme value of the guide column slope. The three-dimensional axial continuous memory block traversal reading loop is started to perform the aggregation action and merge multiple axial displacement scalars. The aggregated data is written into the physical address space of the continuous register inside the central processing unit to generate a spatial pose state array.

[0023] The wind load disturbance estimation module includes: The residual integral solution submodule is based on the spatial pose state array. It extracts the servo valve core displacement residual values, multiplies them by the time step span, accumulates the multinomial state residual gradient, calculates the cumulative area of ​​the multidimensional parameter time domain integral, merges the multinomial partial derivative constants, and obtains the lumped disturbance deviation term. Extreme value comparison and judgment submodule: Introduces a fuzzy neural network to compare the extreme values ​​of accumulator pressure, extract the effective values ​​within the interval, assign deviation weight parameters to select the central node variables, extract the equivalent cross-sectional area and lever arm length components of the node scalar multiplication, and generate the torque compensation benchmark value. Compensation superposition calculation submodule: Based on the torque compensation benchmark value, extract the attitude angle deviation values, accumulate various state residual variables, solve the projection vector space bias and merge the yaw axis coordinate difference scalar, superimpose multiple torque correction bases in the feedback channel, and obtain the feedforward yaw torque parameters; The residual integral solving submodule, based on the spatial pose state array, uses the trapezoidal rule discrete numerical integration algorithm to perform time-domain cumulative calculations on discrete servo variables. It calls the system timer interrupt service routine to set a preset time step span. This preset time step span is obtained by reading the fixed configuration parameters of the external quartz crystal oscillator frequency divider register on the motherboard through the system initialization boot program, with a specific constant set to 5 milliseconds. It configures the memory direct access controller to map the underlying physical address block of the dual-port random access memory as a data buffer. It then initiates a hardware data pointer loop to extract the residual values ​​of the servo valve core displacement at discrete moments. Finally, it calls the arithmetic logic unit inside the central processing unit to perform a summation operation on two adjacent discrete displacement values ​​and shifts them to the right by 1 bit for scaling. Finally, it drives the floating-point multiplier to convert the scaled values. The values ​​are multiplied by a preset time step span constant. A 64-bit wide global accumulator register is allocated to receive the multiplier output data in a loop to perform a summation operation. The residual values ​​of the servo valve core displacement are extracted and multiplied by the time step span. A multi-threaded parallel computing unit is allocated to extract multi-dimensional array pointers and distribute them to different logic cores. The accumulated values ​​in the registers are read and multi-state residual gradients are accumulated. The memory array boundary pointers are set to execute a loop superposition addition instruction and calculate the accumulated area of ​​the multi-dimensional parameter time domain integral. The preset partial derivative coefficient solidified constant value of 10 is extracted from a specific address segment of the read-only memory chip. The single instruction multiple data stream multiply-add unit is called to combine the area variable with the solidified constant value to perform a merging operation and merge multiple partial derivative constants. The results are transmitted to the result register group through the internal high-speed bus to obtain the lumped disturbance deviation term. The extreme value comparison and determination submodule: Based on the lumped disturbance deviation term, a fuzzy neural network forward inference mapping algorithm is used to perform nonlinear evaluation mapping of the time-domain integral parameter. The operating system file system interface is invoked to read the device configuration file mapping table to establish a preset accumulator pressure extreme value range. This preset accumulator pressure extreme value range is obtained by the system power-on self-test program calling the underlying driver interface to mount non-volatile storage media and reading specific sector boundary threshold variables, explicitly configuring a lower limit of 15 MPa and an upper limit of 25 MPa. A 5-layer feedforward network topology is constructed, configuring input layer nodes to receive lumped disturbance input variables, transmitting the input variables to the fuzzification layer, calling the Gaussian membership center node constant with a constant of 20 and a width constant of 5 to extract feature measurement parameters, activating 64 conditional judgment statements in the rule inference layer and performing logical AND operations, transmitting intermediate variables to the defuzzification layer. The system uses the center of gravity to obtain the output data of the instruction operation node, and introduces a fuzzy neural network to compare the extreme values ​​of the accumulator pressure and extract the effective values ​​within the range. It configures the network error backpropagation initialization allocation parameters, sets the initial weight of the center node to a fixed constant of 0.6 and the initial weight of the edge nodes to a fixed constant of 0.4, calls the hardware multiplier accumulator to perform the summation of the weight and node data, and allocates the deviation weight parameter to select the center node variable. It reads the hardware configuration information database to extract the preset equivalent cross-sectional area constant of 50 square centimeters and the preset lever arm length constant of 120 millimeters. It calls the double-precision floating-point multiplication calculation unit to execute the arithmetic instruction of multiplying the scalar node data with the preset equivalent cross-sectional area constant and the preset lever arm length constant, and extracts the equivalent cross-sectional area and lever arm length components of the node scalar multiplication. It transmits the final product result to the output control interface register mapping address space to generate the torque compensation reference value. The compensation superposition calculation submodule, based on the torque compensation benchmark value, employs a multidimensional tensor element-wise affine superposition algorithm to perform matrix fusion and assembly operations on spatial vector data. It extracts a preset torque correction base variable, which is obtained by writing decimal parameter values ​​to a read-only memory chip via a universal serial bus communication debugging interface during the factory online calibration phase. This preset torque correction base variable is specifically configured to be 150 N·m. A 1024-byte single-precision floating-point array is constructed in the system's main memory as a temporary storage matrix. Direct memory access (DMI) instructions are initiated to push attitude deviation data collected by external sensors into the specified memory stack's starting physical address. A loop control structure is called, setting the step increment parameter to 1, to read the memory stack data line by line. The CPU's underlying matrix addition instructions are executed to accumulate multiple state residual variables in the memory storage block. The algorithm extracts attitude angle deviation values ​​and accumulates multiple state residual variables. It then coordinates the coprocessor unit to perform multiplication of the three-dimensional coordinate system rotation and translation transformation matrix to calculate the projection difference parameters. It extracts the reference coordinate origin, sets the three-dimensional initial variables to 0, calculates the actual offset distance in multi-dimensional space, and solves the projection vector space offset and merges the yaw axis coordinate difference scalar. It then opens the feedback data interrupt receiving channel to acquire the real-time parameters of the system feedback status register. It calls the single instruction multiple data stream extended instruction set to perform batch accumulation operations on the underlying data in multiple register addresses within a single clock cycle. It introduces the aforementioned preset torque correction base constant as the basic offset input of the adder and superimposes multiple torque correction bases in the feedback channel. Finally, it outputs the summarized value to the driver digital isolated communication transmit buffer address space to obtain the feedforward yaw torque parameter.

[0024] A fuzzy neural network is used to extract lumped disturbance deviation terms and input them into the network mapping nodes. Fuzzy calculations are performed using multiple membership functions to generate a fuzzy variable matrix. The fuzzy variable matrix is ​​then passed to the network hidden layer to trigger logical reasoning rules. Multiple node activation parameters are calculated, and the multiple node activation parameters are compared with the extreme boundary conditions of the accumulator pressure. Out-of-bounds parameters are removed, and valid values ​​within the extreme value range are extracted.

[0025] The steady-state trajectory optimization module includes: The state sequence prediction submodule: Based on the feedforward yaw moment parameter, it calls Gaussian process regression to extract the cylinder extreme value multiplication control parameter, accumulates the pose deflection value within the step size and solves the deduction set to split discrete nodes, extracts the trajectory deduction vector scalar, and generates the state prediction evolution sequence. Extreme value constraint compression submodule: Based on the state prediction evolution sequence, extract the extreme values ​​of the prediction residuals to extract discrete invariant set constants, shrink the boundary conditions of the input torque constraint space, superimpose the limit numerical limit action threshold scalar of the lifting mechanism, and obtain the displacement constraint boundary parameters; The target command calculation submodule: Based on the displacement constraint boundary parameters, it compares the rated flow output of the pump station to eliminate out-of-bounds deviation terms, calculates the valley value deviation by solving the first derivative of the objective function, extracts the push rod displacement increment and merges the drive sequence to generate the contraction boundary thrust command flow. The state sequence prediction submodule, based on the feedforward yaw torque parameter, uses a Gaussian process regression prior prediction algorithm to perform state deduction calculations on the dynamic yaw torque tensor. It retrieves the preset cylinder extreme constant from the hardware configuration interface. This preset cylinder extreme constant is obtained by measuring the physical travel upper limit using a laser rangefinder during the factory assembly stage and burning it into a read-only memory sector via a serial peripheral interface bus; its specific fixed configuration value is 850 mm. It configures the input data channel mapping variable memory address, calls the underlying floating-point multiplication calculation unit to execute the arithmetic instruction of multiplying the torque input array with the constant scalar, and calls Gaussian process regression to extract the cylinder extreme multiplication control parameter. It also calls the system's built-in clock module to extract the time synchronization signal, registers the data acquisition period constant of 10 milliseconds, and enables the multidimensional matrix accumulation coprocessor module. The rotation and translation step increment constant of the three-dimensional spatial coordinate system is 2. The CPU executes the auto-incrementing loop read instruction of the underlying memory pointer to merge the data of 50 consecutive sampling points within the step size span, and accumulates the pose deflection value within the step size. The Cholsky decomposition instruction in the matrix operation function library is called to decompose the eigenvector of the covariance matrix. The direct memory access transport instruction is initiated to cut the decomposed output continuous memory block into discrete blocks according to the preset node step size constant of 128 bytes and store it in the address segment of the independent register group. The derivation set is solved and the discrete nodes are split. The data cursor is configured to read the continuous floating-point value set stored in the register group. The hardware comparator is called to remove zero values ​​and keep non-zero floating-point data items and store them in the target sequence array. The trajectory derivation vector scalar is extracted and written to the coprocessor output cache bus channel in time sequence to generate the state prediction evolution sequence. The extreme value constraint compression submodule: Based on the state prediction evolution sequence, it uses an invariant set boundary space constraint scaling algorithm to perform boundary stripping and constraint operations on the evolution sequence data matrix. It configures the memory direct access controller to map the underlying physical address block of the dual-port random access memory to extract the sequence array. It calls the arithmetic logic unit inside the central processing unit to use bubble sort instructions to search for floating-point data extreme values ​​in the array. It retrieves the preset discrete invariant set constant from the system configuration file interface. This preset discrete invariant set constant is obtained by reading the fixed value of a specific reserved sector on the non-volatile storage medium of the device motherboard through the system's underlying initialization firmware, and its value is explicitly configured as 15. It calls the logical subtraction operator to subtract the set of constants within the reserved interval that exceeds the extreme value item, and extracts the predicted residual extreme value to truncate the discrete invariant set constant. It also extracts the preset boundary scaling ratio constant, which is obtained through the system... The system power-on initialization process starts the boot program and reads the read-only memory chip to obtain the fixed floating-point value of 0.8. The hardware multiplier unit is called to perform element-wise multiplication of the multiple elements in the input constraint tensor with the proportional constant, and shrinks the boundary conditions of the input torque constraint space. The hardware configuration information library is read to extract the preset lifting mechanism limit value. The preset lifting mechanism limit value is obtained by writing an integer to the control motherboard through the general serial bus communication debugging interface during the factory online calibration stage, and the specific value is 600 mm. The arithmetic logic addition unit is called to merge the boundary scaling tensor result and the limit parameter item in the hardware accumulator. The output limiting instruction drives the comparator to shield the signal channel that is greater than the given limiting threshold, and superimposes the lifting mechanism limit value to limit the action threshold scalar. The limited limiting value is stored in the physical address space of the controller global variable register to obtain the displacement constraint boundary parameters. The target instruction calculation submodule, based on displacement constraint boundary parameters, employs a first-order derivative gradient valley-finding algorithm to perform derivative search operations on the constraint boundary tensor array. It retrieves preset rated flow parameters from the operating system driver interface. These preset rated flow parameters are obtained by measuring feedback values ​​from an external flow meter during the equipment trial operation phase and writing them into a specific register of the programmable read-only memory; the specific fixed value is 200 liters per minute. The hardware comparator unit is called to compare the input flow data stream bit-by-bit with parameter constants. The logic circuit selector switch is set to only allow instruction data packets less than or equal to a constant threshold and discard overflow data frames. The rated flow rate output of the pump station is compared to eliminate out-of-bounds deviation terms. The central processing unit's floating-point derivative coprocessor core is activated to configure the discrete flow field data column vector with a preset step size of constant 2 and calls the forward differential function to perform the calculation. The system performs batch differentiation to obtain a derivative vector column, calls the array minimum search subroutine to locate the minimum index of the derivative vector column, maps the difference tensor using a cursor, and calculates the first derivative of the objective function to calculate the valley value bias. It reads the digital signal interface communication protocol frame of the position sensor to obtain the decimal value of the push rod's instantaneous absolute coordinates, calls the addition calculation unit to superimpose the above valley value mapping difference parameter to obtain the target displacement command quantity, allocates system main memory space to construct a 1024-byte instruction temporary storage queue array, pushes the displacement command packet, configures direct memory access transport instructions to transfer the pulse sequence constant set in the external digital logic drive module register to the tail address segment of the temporary storage queue for merging operation, extracts the push rod displacement increment and merges the drive sequence, encapsulates and converts the bit stream to output to the lower-level execution bus communication interface, and generates a contraction boundary thrust command stream.

[0026] Gaussian process regression is used to extract the feedforward yaw moment parameter into the process regression prior mean function. The kernel function is called to calculate the feature covariance matrix. The cylinder extreme value multiplication control parameter is extracted to construct the observation data sample. The posterior probability distribution operation is performed to deduce and accumulate the pose deflection value within the step size. The multidimensional joint probability function is solved to generate the deduction set. The covariance feature array is decomposed to split the discrete nodes.

[0027] The deviation scaling module includes: The matrix residual decomposition submodule: Based on the feedforward yaw moment parameter and the contraction boundary thrust command flow, it decomposes and extracts the hinge shaft deformation residual vector, matches the interval separation boundary extreme values, accumulates and merges the hinge error, and generates the state interval residual vector. Multiplier scaling adjustment submodule: Based on the state interval residual vector, extract the pad block residual matching attenuation factor, multiply the weighted constant to adjust the dynamic scaling ratio, correct multiple servo gain weights, and construct a gain multiplier array; The matrix residual decomposition submodule, based on the feedforward yaw moment parameter and the contraction boundary thrust command stream, employs a singular value tensor order reduction decomposition algorithm to perform orthogonal basis projection stripping analysis on the dual-channel input data stream. It allocates the central processing unit's underlying direct memory access controller to map the cache shared physical address segment, reads the preset decomposition matrix dimension constant (obtained by calling the basic input / output system firmware during system power-on initialization, reading specific sector configuration parameters of the motherboard's non-volatile memory, with a specific constant defined as 128 rows and 128 columns), calls the system's built-in linear algebra operation coprocessor to invoke double-precision floating-point singular value solving hard-wired instructions, strips the corresponding basis column vector arrays from the left singular orthogonal feature matrix whose singular values ​​are greater than the preset noise floor constant 0.05, and decomposes and extracts the hinge deformation residual vector. It configures a field-effect transistor logic comparator array to perform interval threshold truncation filtering on the underlying floating-point elements of the stripped column vectors, and reads the preset dynamic boundary extreme value interval. The isolation constant, the preset dynamic boundary extreme value isolation constant, is obtained by calling the general serial bus communication debugging interface during the overall equipment debugging stage and written to the static random access memory. The specific fixed configuration value is 15 mm. The underlying hardware data gating gate circuit is set to forcibly cut off the memory address reading channel corresponding to the absolute value exceeding the isolation constant, and the interval separation boundary extreme value is matched. The 64-bit dedicated accumulation register group inside the central processing unit is initialized and a zero-filling operation is forcibly performed. The clock beat synchronous loop control cursor pointer is enabled to extract the normal residual data component of the memory stack retention segment one by one. The memory read instruction step length constant is configured to 8 bytes to execute the double-precision aligned addressing read action. The arithmetic logic unit is called to execute the multi-cycle accumulation addition micro-instruction and accumulate and merge the hinge error. The peripheral device interconnect bus data burst transmission instruction is called to push the accumulated output result sequence into the first-in-first-out buffer storage block inside the field programmable gate array according to a specific timing sequence to generate the state interval residual vector. The multiplier scaling adjustment submodule, based on the state interval residual vector, employs a discrete space lookup table interpolation scaling algorithm to perform proportional compensation coefficient addressing and matching adjustment operations on the error parameter set. It enables the microcontroller's high-speed memory direct read channel to point to the underlying physical base address of a pre-built two-dimensional mapping lookup table within the read-only memory chip. It calls hardware indirect addressing microinstructions to read the preset pad nonlinear attenuation mapping constant. This preset pad nonlinear attenuation mapping constant is obtained by reading the external serial flash device's factory static calibration table through the system's underlying driver's file system mount, with the center node constant configured to 0.85. It calls the arithmetic logic unit to calculate the absolute value of the input residual, compares it with the physical address index node of the lookup table, extracts the corresponding compensation coefficient, and extracts the pad residual matching attenuation factor. Finally, it reads the preset weighted scaling reference constant, which is sent by the control motherboard's underlying reset boot program to the designated... The physical register address is written with a fixed decimal value of 3. The floating-point multiplication calculation hardline unit is called to perform the aforementioned compensation coefficient and the preset weighted scaling reference constant in a concurrent multiplication arithmetic operation. Simultaneously, the register logic shift instruction is called to shift 2 bits to the left to perform the underlying bit-level data amplification and conversion operation, and the dynamic scaling ratio is adjusted by multiplying by the weighting constant. The basic closed-loop proportional-integral-differential servo parameter set is extracted from a specific section of the memory control loop. The single instruction multiple data stream extended instruction architecture is called to perform batch parallel multiplication and addition correction operation micro-instruction operations on multiple parameter addresses within a single clock cycle, and correct multiple servo gain weights. The memory page continuous write mode is configured to write the corrected double-precision floating-point data set in batches into the field-programmable gate array front-end transmission communication register buffer queue space according to the continuous physical address allocation rule, and a gain multiplier array is constructed.

[0028] The ramp servo stabilizer module includes: Servo pin locking submodule: Based on the gain multiplier array, it extracts the slope matrix to convert the voltage duty cycle, adjusts the opening of the proportional servo valve, drives the guide pin to penetrate the hole for rigid locking, separates the multi-axis displacement deviation values, and generates the driving locking displacement scalar. Mooring tension adjustment submodule: Based on the drive lock displacement scalar, the output torque constant is used to adjust the mooring cable tension value, accumulate the displacement difference of each axis in multi-dimensional space, merge the longitudinal and transverse tilt angle deviations, extract the attitude space offset components, and obtain multi-directional wind-resistant attitude parameters. Servo Pin Locking Submodule: Based on a gain multiplier array, it uses a pulse width modulation waveform duty cycle digital generation algorithm to perform low-level electrical signal modulation and conversion operations on matrix data. It retrieves the low-level floating-point variables from the array's storage address mapped by the main control chip's internal direct memory access controller, and calls a floating-point to fixed-point multiplication microinstruction to multiply a preset reference clock cycle constant. This preset reference clock cycle constant is obtained by reading the high and low level states of the external crystal oscillator frequency divider's physical pins during the system's low-level firmware startup phase, and its specific value is fixed at 20 microseconds. It is written into the microcontroller's internal timer comparison capture register physical address space, and the slope matrix conversion voltage duty cycle is extracted. The digital-to-analog conversion coprocessor is then called to output a specified amplitude electrical signal to the external push-pull amplifier circuit, configuring the output drive current peak. The parameter constant is 5 amps. The proportional electromagnet displacement coil is driven to perform electromagnetic thrust conversion action, adjust the opening of the proportional servo valve, drive the hydraulic pump station to output rated pressure oil to propel the hydraulic cylinder physical push rod, set the push rod displacement sensor detection distance constant to 150 mm, touch the physical limit switch to trigger a hard-wire interrupt level flip signal, and drive the guide pin to penetrate the hole for rigid locking. The multiplexer channel is called to switch the hard-wire instruction to isolate the locking state before the dynamic data stream, configure the memory data to truncate the cursor pointer to offset 4 bytes to the right to cross the useless data segment and remove the static reference coordinate variable, and separate the multi-axis displacement deviation value. The above parameters are transmitted in batches to the displacement monitoring independent register group storage physical block through the internal integrated circuit interconnect bus to generate the drive locking displacement scalar. The mooring tension adjustment submodule, based on the drive-locking displacement scalar, employs a multi-dimensional spatial tensor difference accumulation and orthogonal merging algorithm to perform physical field variable reconstruction and reorganization calculations on multi-channel state data. It reads the preset mooring torque constant, which is obtained by calling the programmable logic controller's serial communication protocol through the winch equipment online calibration program and written to the non-volatile holding register physical address, with a specific configuration value of 1200 N·m. It then calls the central processing unit's digital signal processing arithmetic logic unit to execute basic multiply-add hardwired instructions to calculate the given electrical signal bias and outputs the torque constant to adjust the mooring cable tension. It activates the coprocessor's three-dimensional coordinate difference operation instruction internal queue, sets the bottom-level sampling sliding window length constant to 64 consecutive data packets, calls the hardware memory loop pointer to extract the three-dimensional absolute value hardware encoder feedback pulse count values ​​one by one, and uses the bottom-level digital subtractor hardwired logic to subtract the preceding buffer frames according to clock timing to temporarily store data in an independent storage block. Finally, it activates the global arithmetic accumulator register to continuously receive the output stream and accumulate it. The displacement difference of each axis in multidimensional space is used to extract a preset tilt angle compensation constant. The preset tilt angle compensation constant is obtained by running the low-level calibration program of the inertial navigation unit during the factory sea trial and is burned into the encrypted protection sector of the read-only memory. The specific value is fixed at 2 degrees. The logic operation unit calls the bit-by-bit OR splicing instruction to merge the lateral tilt angle data bits and the longitudinal tilt angle data bits. The aforementioned tilt angle compensation constant is directly added to the head address of the splicing segment in the physical memory space to directly overwrite and replace the value. The longitudinal and lateral tilt angle offsets are also merged. The direct memory access channel controller is called to transmit the merging instruction block to the independent temporary storage block inside the feature map. The micro-instruction of bit alignment of specific bytes in the memory page is executed to cross the header descriptor mask bit segment to extract the low-level hexadecimal variable of the specific address segment and extract the attitude space offset component. The low-level protocol stack driver function of the motherboard Ethernet controller is called to perform packet conversion and formatting operations. The extracted parameters are filled into the Ethernet application layer data payload area and sent to the upper computer communication interface through the bus physical layer to obtain multi-directional wind-resistant attitude parameters.

[0029] Please see Figure 2 The dynamic wind resistance stability control method during the adjustment process of the roll-on / roll-off platform ramp includes the following steps: S1: Based on the tilt angle signals of multiple rigid springboards, read the joint pulses and substitute them into the matrix for calculation. Add the differences of multiple hinge projections, derive the lip plate deviation and guide column tilt parameters, merge the multi-directional linear displacement scale, and generate a spatial pose state array. S2: Based on the spatial pose state array, the displacement difference integral is extracted to calculate the interference residual. The effective segment is extracted by combining the fuzzy neural network with the limit, the torque is derived by selecting the node multiplication arm, and multiple residuals are summarized to obtain the feedforward yaw moment parameter. S3: Based on the feedforward yaw moment parameter, call the Gaussian process regression to calculate the motion trend, separate the extreme values ​​to reduce the boundary range, combine the flow to remove out-of-bounds terms, derive the function to extract the push rod extension length, and generate the shrinking boundary thrust command flow; S4: Based on the feedforward yaw moment parameter and the contraction boundary thrust command flow, decompose and peel off the hinge shaft deformation error, determine the operating condition mapping pad reduction coefficient, retrieve constants for multiplication to adjust the amplification ratio, and construct a gain multiplier array; S5: Based on the gain multiplier array, the waveform is mapped by the transformation matrix to change the cross section of the proportional valve, the through hole position is fixed by the push pin, the torque is issued to change the tension of the mooring cable, and the three-dimensional axial relative displacement difference is summarized to obtain the multi-directional wind-resistant attitude parameters.

[0030] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A dynamic wind-resistant stability control system for the adjustment process of a roll-on / roll-off platform ramp, characterized in that, The system includes: Spatial Deformation Analysis Module: Extracts joint pulse values ​​from the tilt angle signals of multiple rigid planks, performs spatial transformation calculations, accumulates the projection variables of the hinge points, calculates the extreme values ​​of the lap lip plate deviation and guide column slope, aggregates multi-axis displacement scalars, and generates a spatial pose state array. Wind load disturbance estimation module: Based on the spatial pose state array, extract the valve core displacement residual integral to solve the disturbance lumped deviation term, compare the accumulator pressure extreme value through a fuzzy neural network, determine the central numerical calculation torque compensation component and superimpose multiple state residual variables to obtain the feedforward yaw torque parameter; Steady-state trajectory optimization module: Based on the feedforward yaw moment parameter, it calls Gaussian process regression to calculate the state prediction sequence, extracts boundary values ​​to perform constraint contraction operation, compares the rated output of the pump station, solves the polynomial to extract the push rod displacement increment, and generates the contraction boundary thrust command flow; The deviation scaling module calculates the following: based on the feedforward yaw moment parameter and the shrinkage boundary thrust command flow, it decomposes the matrix to extract the hinge shaft deformation residual vector, matches the state interval to calculate the pad attenuation factor, extracts the constant and multiplies it by the factor to perform scaling adjustment, and constructs a gain multiplier array. The ramp servo stabilization module: Based on the gain multiplier array, it calls the slope matrix to convert the duty cycle, adjusts the servo valve opening, drives the pin to insert into the lock hole to perform rigid locking, outputs torque to adjust the mooring cable tension value and accumulates the coordinate difference to obtain multi-directional wind-resistant posture parameters.

2. The dynamic wind-resistant stability control system for the roll-on / roll-off platform ramp adjustment process according to claim 1, characterized in that, The spatial pose state array includes the overlap lip plate deviation, the guide post slope extreme value, and the multi-axis displacement scalar; the feedforward yaw moment parameter includes the torque compensation component, the projection vector spatial deviation, and the yaw axis coordinate difference scalar; the contraction boundary thrust command flow includes the state prediction sequence, the action threshold scalar, and the push rod displacement increment; the gain multiplier array includes the pad attenuation factor, the dynamic scaling ratio, and the servo gain weight; and the multi-directional wind-resistant pose parameter includes the drive locking displacement scalar, the mooring cable tension value, and the longitudinal and transverse tilt angle deviation.

3. The dynamic wind-resistant stability control system for the roll-on / roll-off platform ramp adjustment process according to claim 1, characterized in that, The spatial deformation analysis module includes: Vector space projection submodule: Based on the tilt angle signals of multiple rigid planks, extract the joint pulse values ​​and multiply them by the angle cosine components to construct a spatial coordinate basis. Accumulate multiple projection differences and merge the displacement coordinate values ​​of the hinge points. Transform the displacement offset components of each three-dimensional axis to generate the node displacement mapping. Scalar aggregation and deduction submodule: Based on the node displacement mapping, the axial parameters are stripped and the reference coordinates are subtracted to split the multi-dimensional vector, the endpoint vector deviation value is calculated, the overlap lip plate deviation value is extracted and multiplied by the extreme value of the guide post slope, and multiple axial displacement scalars are merged to generate a spatial pose state array.

4. The dynamic wind-resistant stability control system for the roll-on / roll-off platform ramp adjustment process according to claim 1, characterized in that, The wind load disturbance estimation module includes: The residual integral solving submodule: Based on the spatial pose state array, extract the servo valve core displacement residual values, multiply them by the time step span, accumulate the multinomial state residual gradient, calculate the cumulative area of ​​the multidimensional parameter time domain integral, merge the multinomial partial derivative constants, and obtain the lumped disturbance deviation term. Extreme value comparison and judgment submodule: Introduces a fuzzy neural network to compare the extreme values ​​of accumulator pressure, extract the effective values ​​within the interval, assign deviation weight parameters to select the central node variables, extract the equivalent cross-sectional area and lever arm length components of the node scalar multiplication, and generate the torque compensation benchmark value. Compensation superposition calculation submodule: Based on the torque compensation benchmark value, extract the attitude angle deviation values, accumulate various state residual variables, solve the projection vector space bias and merge the yaw axis coordinate difference scalar, superimpose multiple torque correction bases in the feedback channel, and obtain the feedforward yaw torque parameters.

5. The dynamic wind-resistant stability control system for the roll-on / roll-off platform ramp adjustment process according to claim 1, characterized in that, The fuzzy neural network extracts the lumped disturbance deviation term and inputs it into the network mapping node. It then performs fuzzy calculations using multiple membership functions to generate a fuzzy variable matrix. This matrix is ​​then passed to the network hidden layer to trigger logical reasoning rules. Multiple node activation parameters are calculated, and the activation parameters are compared with the extreme boundary conditions of the accumulator pressure. Out-of-bounds parameters are removed, and valid values ​​within the extreme value range are extracted.

6. The dynamic wind-resistant stability control system for the roll-on / roll-off platform ramp adjustment process according to claim 1, characterized in that, The steady-state trajectory optimization module includes: State sequence prediction submodule: Based on the feedforward yaw moment parameter, Gaussian process regression is called to extract the cylinder extreme value multiplication control parameter, the pose deflection value within the step size is accumulated and the inference set is split into discrete nodes, the trajectory inference vector scalar is extracted, and the state prediction evolution sequence is generated. Extreme value constraint compression submodule: Based on the state prediction evolution sequence, extract the extreme values ​​of the prediction residuals to extract discrete invariant set constants, shrink the boundary conditions of the input torque constraint space, superimpose the limit values ​​of the lifting mechanism to limit the action threshold scalar, and obtain the displacement constraint boundary parameters; Target command calculation submodule: Based on the displacement constraint boundary parameters, compare the rated flow output of the pump station to eliminate out-of-bounds deviation terms, calculate the valley value deviation by solving the first derivative of the objective function, extract the push rod displacement increment and merge the drive sequence to generate the contraction boundary thrust command flow.

7. The dynamic wind-resistant stability control system for the roll-on / roll-off platform ramp adjustment process according to claim 1, characterized in that, The Gaussian process regression extracts the feedforward yaw moment parameter into the process regression prior mean function, calls the kernel function to calculate the feature covariance matrix, extracts the cylinder extreme value multiplication control parameter to construct the observation data sample, performs posterior probability distribution calculation and deduction and accumulates the pose deflection value within the step size, solves the multidimensional joint probability function to generate the deduction set, and decomposes the covariance feature array to split discrete nodes.

8. The dynamic wind-resistant stability control system for the ramp adjustment process of the roll-on / roll-off platform according to claim 1, characterized in that, The measurement deviation scaling module includes: Matrix residual decomposition submodule: Based on the feedforward yaw moment parameter and the contraction boundary thrust command flow, decompose and extract the hinge shaft deformation residual vector, match the interval separation boundary extreme values, accumulate and merge the hinge error, and generate the state interval residual vector; Multiplier scaling adjustment submodule: Based on the residual vector of the state interval, extract the pad block residual matching attenuation factor, multiply it by the weighted constant to adjust the dynamic scaling ratio, correct multiple servo gain weights, and construct a gain multiplier array.

9. The dynamic wind-resistant stability control system for the ramp adjustment process of the roll-on / roll-off platform according to claim 1, characterized in that, The ramp servo stabilization module includes: Servo pin locking submodule: Based on the gain multiplier array, extract the slope matrix to convert the voltage duty cycle, adjust the opening of the proportional servo valve, drive the guide pin to penetrate the hole for rigid locking, separate the multi-axis displacement deviation values, and generate the driving locking displacement scalar. Mooring tension adjustment submodule: Based on the drive locking displacement scalar, output torque constant to adjust the mooring cable tension value, accumulate the displacement difference of each axis in multi-dimensional space, merge the longitudinal and transverse tilt angle deviations, extract the attitude space offset component, and obtain multi-directional wind-resistant attitude parameters.

10. A dynamic wind-resistant stability control method during the adjustment process of a roll-on / roll-off platform ramp, characterized in that, The dynamic wind-resistant stability control system for the roll-on / roll-off platform ramp adjustment process according to any one of claims 1-9 includes the following steps: S1: Based on the tilt angle signals of multiple rigid springboards, read the joint pulses and substitute them into the matrix for calculation. Add the differences of multiple hinge projections, derive the lip plate deviation and guide column tilt parameters, merge the multi-directional linear displacement scale, and generate a spatial pose state array. S2: Based on the spatial pose state array, extract the displacement difference integral to calculate the interference residual, use a fuzzy neural network combined with the limit to extract the effective segment, select nodes to multiply the lever arm to derive the torque, summarize multiple residuals, and obtain the feedforward yaw torque parameter. S3: Based on the feedforward yaw moment parameter, call Gaussian process regression to calculate the motion trend, separate the extreme values ​​to reduce the boundary range, combine the flow to remove out-of-bounds terms, differentiate the function to extract the push rod extension length, and generate the shrinking boundary thrust command flow. S4: Based on the feedforward yaw moment parameter and the contraction boundary thrust command flow, decompose and peel off the hinge shaft deformation error, determine the operating condition mapping pad reduction coefficient, retrieve constants for multiplication to adjust the amplification ratio, and construct a gain multiplier array; S5: Based on the gain multiplier array, the waveform is mapped by the transformation matrix to change the cross section of the proportional valve, the through hole position of the push pin is fixed, the torque is issued to change the tension of the mooring cable, and the three-dimensional axial relative displacement difference is summarized to obtain the multi-directional wind-resistant attitude parameters.

Images