A steel berthing member stress deformation self-adaptive compensation control system

Abstract

The application relates to the technical field of structural health monitoring, and discloses a steel berthing component stress deformation self-adaptive compensation control system, which comprises the following modules: a distributed stress deformation sensing module, which is used for monitoring the stress deformation of the steel berthing component in real time; a real-time rolling optimization control module, which is used for receiving the measured stress deformation data; a model parameter online correction module, which is used for receiving the measured stress deformation data and obtaining the prediction output of a component dynamic prediction model; a model structure self-adaptive setting module, which is used for receiving a model residual signal; and an internal active prestress execution module, which is used for receiving a control instruction and applying active prestress in the steel berthing component to compensate for the stress deformation. Through the construction of a double-layer adaptive framework of parameters and structures, the model mismatch problem is solved, the complex time-varying characteristics of the component have high robustness, and the long-term stability and effectiveness are ensured.

Description

Technical Field

[0001] This invention relates to the field of structural health monitoring technology, specifically to an adaptive compensation control system for stress deformation of steel berthing components. Background Technology

[0002] Steel berthing components are key load-bearing structures in marine engineering projects such as ports and docks. During their service life, they are frequently subjected to huge and complex transient impact loads generated when ships berth. To ensure their structural safety and extend their service life, existing technical solutions mainly focus on installing passive energy-absorbing devices (such as rubber fenders) and deploying structural health monitoring systems.

[0003] Passive energy absorption devices can only dissipate impact energy and cannot actively regulate the stress distribution inside the component based on the real-time stress state, making it difficult to fundamentally avoid stress concentration or suppress fatigue hotspots. Traditional structural health monitoring systems typically only monitor and control the stress deformation of components, lacking the ability to actively intervene.

[0004] To achieve active control, some approaches attempt to introduce predictive model-based controllers that apply active forces within the component to counteract external loads. However, these controllers generally rely on a static dynamic model with fixed parameters established during the design phase. As a large structure that serves in harsh marine environments for extended periods, the dynamic characteristics of steel berthing components are not static. Material aging, the accumulation of minor damage, and changes in ambient temperature can all cause a slow drift in their physical parameters (such as stiffness and damping). More seriously, when a component is subjected to unexpected impacts or cumulative damage, its overall dynamic structure may change significantly, rendering the original model order no longer fit.

[0005] Controllers using fixed models cannot adapt to such complex time-varying characteristics. When the model mismatches with the physical reality, its prediction accuracy will drop significantly, resulting in suboptimal control commands. This not only fails to effectively suppress stress concentration or delay fatigue accumulation, but may also produce erroneous control actions, endangering structural safety. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an adaptive compensation control system for stress deformation of steel berthing components. This system solves the problem that existing controllers based on fixed models cannot adapt to the time-varying dynamic characteristics of steel berthing components, leading to model mismatch and decreased control performance.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an adaptive compensation control system for stress deformation of steel berthing components, comprising:

[0008] A distributed stress deformation sensing module is used to monitor the stress deformation of the steel berthing components in real time and output measured stress deformation data.

[0009] A real-time rolling optimization control module is connected to the distributed stress deformation sensing module. It is used to receive the measured stress deformation data and perform rolling time-domain optimization based on a component dynamic prediction model and the measured stress deformation data to generate control commands.

[0010] The online model parameter correction module is connected to the distributed stress deformation sensing module and the real-time rolling optimization control module. It is used to receive the measured stress deformation data and obtain the prediction output of the component dynamic prediction model. By comparing the prediction output with the measured stress deformation data, it generates a model residual signal and updates the internal parameters of the component dynamic prediction model online based on the model residual signal.

[0011] The model structure adaptive tuning module is connected to the model parameter online correction module. It is used to receive the model residual signal, perform feature analysis on the model residual signal to diagnose the structural well-being of the component dynamic prediction model, and send a model structure switching command to the real-time rolling optimization control module when a model structure mismatch is diagnosed, so as to adjust the component dynamic prediction model used by the module.

[0012] An internal active prestressing execution module, connected to the real-time rolling optimization control module, is used to receive the control command and apply active prestress inside the steel berthing component to compensate for the stress deformation.

[0013] Preferably, the distributed stress deformation sensing module includes:

[0014] A fiber Bragg grating array is deployed on the steel berthing structure;

[0015] The signal demodulation unit is used to monitor the center wavelength drift of the fiber Bragg grating array;

[0016] The data processing unit includes a temperature compensation mechanism and is used to calculate the measured stress deformation data based on the temperature-compensated center wavelength drift.

[0017] Furthermore, the distributed stress deformation sensing module includes a fiber Bragg grating array, a signal demodulation unit, and a data processing unit. The data processing unit includes a temperature compensation mechanism and is used to calculate the temperature-compensated center wavelength drift. Calculate engineering strain :

[0018] ;

[0019] In the formula, For position In Constant engineering adaptability This represents the center wavelength shift at the corresponding position and time. The initial center wavelength of the FBG sensor. The effective elastic-optical coefficient of the optical fiber material. For the first along the direction of the fiber Bragg grating array Location of each measuring point For intermediate variables.

[0020] Preferably, the real-time rolling optimization control module has a built-in reduced-order model library unit, which stores multiple component dynamic prediction models with different orders.

[0021] The real-time rolling optimization control module is used to respond to model structure switching commands and activate a component dynamic prediction model of a specific order from the reduced-order model library unit.

[0022] Preferably, the real-time rolling optimization control module performs rolling temporal optimization by solving a cost function for stress field topology management;

[0023] The cost function includes at least one of the following:

[0024] Peak stress suppression term, used to minimize the maximum stress value at critical locations of the component;

[0025] The stress gradient smoothing term is used to minimize the spatial gradient norm of the stress field inside the component.

[0026] The fatigue accumulation penalty term is used to apply an asymmetric weighted penalty to the tensile stress occurring in the prediction time domain.

[0027] Furthermore, the real-time rolling optimization control module incorporates a reduced-order model library unit, which stores multiple models with different orders. The component dynamic prediction model. Activated The dynamic prediction model for step-by-step components can be expressed as:

[0028] ;

[0029] In the formula, The order of the model. State vector Regarding time The derivative of For the model Order state vector, This is the vector of internal parameters to be corrected in the model. Let be the control input vector to be solved. The external load input vector, The output vector of strain or stress predicted by the model. , , , They are respectively the corresponding The system matrix of the order model.

[0030] Preferably, the online model parameter correction module is specifically used for:

[0031] Construct an augmented state vector containing the state vector and internal parameters of the component dynamic prediction model, where the internal parameters are represented as an internal parameter vector;

[0032] The extended Kalman filter algorithm is used to recursively estimate the augmented state vector by taking the model residual signal as measurement information, so as to update the internal parameter vector online.

[0033] Preferably, the model structure adaptive tuning module includes a residual feature analysis unit, which uses short-time Fourier transform or wavelet transform to perform time-frequency analysis on the model residual signal to obtain the energy distribution characteristics of the model residual signal in the time domain and frequency domain.

[0034] Furthermore, the online model parameter correction module corrects the measured stress-deformation data. and the predicted output Perform vector subtraction to generate the model residual signal. :

[0035] ;

[0036] In the formula, For model residual signals, This is the measured output vector. To predict the output vector, For time variables, This represents the model order.

[0037] Preferably, the model structure adaptive tuning module further includes a mismatch diagnosis decision unit, which is used to execute diagnostic decision logic, including:

[0038] When the energy distribution characteristics show that the residual signal of the model continuously exhibits significant energy concentration in high-frequency bands not included in the current component dynamic prediction model, the component dynamic prediction model is determined to be under-order.

[0039] If the total energy of the model residual signal remains below a preset threshold for a long period of time, and the analysis of the model state shows that the contribution of some higher-order state components of the current component dynamic prediction model to the system output is negligible, then the component dynamic prediction model is determined to be over-order.

[0040] Preferably, the mismatch diagnosis decision unit also generates an upgrade instruction as a model structure switching instruction when it determines that the model is under-ordered, and generates a downgrade instruction as a model structure switching instruction when it determines that the model is over-ordered.

[0041] Preferably, the internal active prestressing execution module includes one or more execution elements and a drive control circuit;

[0042] The actuator is a high-speed electromechanical actuator or a piezoelectric ceramic stack actuator;

[0043] The actuator is integrated and installed inside the steel berthing component and is driven by a drive control circuit to apply active prestress.

[0044] Preferably, the online model parameter correction module constitutes a first-layer adaptive loop to compensate for the drift of physical parameters caused by material aging or environmental changes;

[0045] The model structure adaptive tuning module constitutes a second layer of adaptive loop. The second layer of adaptive loop is used to monitor the model residual signal of the first layer of adaptive loop in order to diagnose and adjust the model structure mismatch caused by changes in dynamic characteristics due to significant changes in the working conditions of the components or cumulative damage.

[0046] Furthermore, the model structure adaptive tuning module includes a residual feature analysis unit. This residual feature analysis unit uses short-time Fourier transform (STFT) or wavelet transform to analyze the model residual signal. Time-frequency analysis was performed to obtain the energy distribution characteristics of the model residual signal in the time and frequency domains. :

[0047] ;

[0048] In the formula, Indicates the current The model residual signal corresponding to the first-order model, The window function represents the weighting function used for local time truncation in the Short-Time Fourier Transform (STFT). Represents a time-shifted variable. Represents the angular frequency variable. Represents the complex exponential kernel function. The time-spectral energy distribution function represents the residual signal of the model.

[0049] This invention provides an adaptive compensation control system for stress deformation of steel berthing components. It has the following beneficial effects:

[0050] 1. This invention constructs a two-layer adaptive framework by setting up an online model parameter correction module and a model structure adaptive tuning module. The online model parameter correction module can compensate for the slow drift of component physical parameters caused by material aging or environmental factors, while the model structure adaptive tuning module can diagnose and adjust the significant changes in component dynamic characteristics caused by cumulative damage or drastic changes in operating conditions. The synergistic effect of the two makes the system highly robust to the complex time-varying characteristics of steel berthing components, ensuring the long-term stability and effectiveness of the control system.

[0051] 2. This invention employs a rolling time-domain optimization strategy through a real-time rolling optimization control module, and solves the problem using a cost function that includes a peak stress suppression term, a stress gradient smoothing term, and a fatigue accumulation penalty term. This allows the application of active prestress to not only compensate for instantaneous stress but also to achieve optimized management of the stress field topology, actively avoid stress concentration, and suppress damage accumulation in fatigue hotspot areas, thereby effectively extending the structural service life of steel berthing components.

[0052] 3. The adaptive tuning module of the model structure of the present invention can accurately diagnose the structural well-being of the current component dynamic prediction model by performing time-frequency feature analysis on the model residual signal, and determine whether it is in an under-order or over-order state. This ensures that the prediction model on which the real-time rolling optimization control module is based always maintains the best match with the actual dynamic characteristics of the component, avoids the decline in control performance caused by model structure mismatch, and balances model accuracy and computational efficiency. Attached Figure Description

[0053] Figure 1 This is a system framework diagram of the present invention;

[0054] Figure 2 This is a schematic diagram of the internal units of the real-time rolling optimization control module of the present invention;

[0055] Figure 3 This is a schematic diagram of the internal units of the adaptive tuning module of the model structure of the present invention;

[0056] Figure 4 This is a schematic diagram of the internal unit of the internal active prestressing execution module of the present invention. Detailed Implementation

[0057] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] Please see the appendix Figure 1 - Appendix Figure 4 This invention provides an adaptive compensation control system for stress deformation of steel berthing components, comprising:

[0059] A distributed stress deformation sensing module is used to monitor the stress deformation of the steel berthing components in real time and output measured stress deformation data.

[0060] Specifically, the distributed stress deformation sensing module in this embodiment is used to monitor the stress deformation of the steel berthing components in real time.

[0061] In one specific embodiment, the distributed stress deformation sensing module includes a sensing element array, a signal demodulation unit, and a data processing unit.

[0062] The sensing element array can specifically be a high-density multiplexed fiber Bragg grating (FBG) array. FBGs are chosen because they exhibit high sensitivity to axial strain and temperature changes, and are suitable for distributed, long-distance deployment. The physical deployment scheme of the FBG array on the steel berthing component includes: during the component manufacturing stage, pre-embedding the FBG sensor array along with the optical fiber inside the component; or after the component is manufactured, deploying the FBG sensor array on the component surface using surface fusion or high-strength bonding techniques. The deployment locations are preferentially selected from critical stress transmission paths, geometrically abrupt regions, and anticipated fatigue hotspots to ensure that the most critical mechanical responses can be captured.

[0063] The signal demodulation unit is connected to the FBG array. When a component is subjected to impact and undergoes stress deformation, the grating period of the FBG sensor at a specific location changes, causing a shift in the center wavelength of its reflection spectrum. The function of the signal demodulation unit is to monitor and acquire the center wavelength shift of all sensors in real time. .

[0064] The data processing unit is responsible for converting the original wavelength drift signal into the measured stress deformation data. The data processing unit first converts the measured wavelength drift... Calculate the corresponding position Engineering strain :

[0065] ;

[0066] In the formula, For position In Constant engineering adaptability This represents the center wavelength shift at the corresponding position and time. The initial center wavelength of the FBG sensor. The effective elastic-optical coefficient of the optical fiber material. For the first along the direction of the fiber Bragg grating array Location of each measuring point For intermediate variables.

[0067] Because the FBG sensor is sensitive to both strain and temperature, the data processing unit also includes a temperature compensation mechanism to eliminate the interference of temperature changes on the measurement results and ensure the accuracy of the calculated values. To simulate real mechanical strain, one specific method for temperature compensation is to deploy a temperature-compensating grating pair. One grating is specially encapsulated so that it only senses temperature and is unaffected by strain. Compensation is achieved by calculating the difference between the two signals. Another method is to place a separate temperature sensor (such as a thermocouple or another FBG) near the FBG sensor. After measuring the ambient temperature, it is subtracted from the calculation based on the thermistor coefficient of the optical fiber.

[0068] The data processing unit will handle all The real-time mechanical strain data from each measurement point, after temperature compensation, are aggregated to form a... dimensional strain vector In some embodiments, the data processing unit may also base its calculations on the material elastic modulus of the steel berthing component. , strain vector Further converted into stress vector :

[0069] ;

[0070] In the formula, The measured stress vector, The elastic modulus of the material. This is the strain vector.

[0071] Finally, the distributed stress-deformation sensing module outputs the measured stress-deformation data (i.e., strain vector). or stress vector .

[0072] A real-time rolling optimization control module, which is connected to a distributed stress deformation sensing module, is used to receive the measured stress deformation data and perform rolling time-domain optimization based on the model and the measured stress deformation data to generate control commands.

[0073] Specifically, in this embodiment, the real-time rolling optimization control module receives measured stress and deformation data and performs rolling time-domain optimization based on the component dynamic prediction model and the measured stress and deformation data to generate control commands.

[0074] The real-time rolling optimization control module has a built-in reduced-order model library unit, which generates and stores dynamic prediction models of components with different orders (i.e., different state vector dimensions) during the offline phase. In one embodiment, these models are obtained by applying the intrinsic orthogonal decomposition (POD) method to reduce the order of a high-fidelity finite element model of a steel berthing component.

[0075] During runtime, the real-time rolling optimization control module 200, based on an initial setting or a model structure switching instruction, activates and calls a dynamic prediction model of a specific order of components from the reduced-order model library unit. The activated model... The dynamic prediction model of a step component can be represented in the following state-space form:

[0076] ;

[0077] In the formula, The order of the model. State vector Regarding time The derivative of For the model Order state vector, This is the vector of internal parameters to be corrected in the model. Let be the control input vector to be solved. The external load input vector, The output vector of strain or stress predicted by the model. , , , They are respectively the corresponding The system matrix of the order model.

[0078] The real-time rolling optimization control module employs a model predictive control (MPC) strategy to perform rolling time-domain optimization. In each control cycle, it first receives the latest measured stress-deformation data from the distributed stress-deformation sensing module. This measured stress-deformation data is used as a feedback signal to correct the current state of the component's dynamic prediction model.

[0079] The real-time scrolling optimization control module is based on the corrected state and the currently active (and its internal parameters) The component dynamic prediction model (updated by the online model parameter correction module) solves a constrained optimization problem within a finite prediction time domain. This optimization problem is addressed by a cost function. For the goal.

[0080] Cost function Designed to implement topology management of stress fields, cost function Includes one or more of the following:

[0081] The peak stress suppression term minimizes the maximum stress value at key locations of the component within the prediction time domain.

[0082] The stress gradient smoothing term minimizes the spatial gradient norm of the stress field inside the component, so as to make the stress distribution smoother and more uniform and avoid stress concentration.

[0083] The fatigue accumulation penalty term is used to apply an asymmetric weighted penalty to the tensile stress that occurs in the prediction time domain based on the estimated fatigue damage weight at different locations of the component, so as to delay the accumulation of damage in fatigue hot spots.

[0084] The result of the optimization solution is an optimal future control sequence. The real-time rolling optimization control module outputs the first control action of this sequence as the control command for the current cycle.

[0085] The online model parameter correction module is connected to the distributed stress deformation sensing module and the real-time rolling optimization control module. It is used to receive the measured stress deformation data and obtain the prediction output of the component dynamic prediction model. By comparing the prediction output with the measured stress deformation data, it generates the model residual signal and updates the internal parameters of the component dynamic prediction model online based on the model residual signal.

[0086] Specifically, in this embodiment, the online model parameter correction module receives the measured stress and deformation data and obtains the prediction output of the component dynamic prediction model. By comparing the prediction output with the measured stress and deformation data, a model residual signal is generated, and the internal parameters of the component dynamic prediction model are updated online based on the model residual signal.

[0087] The online model parameter correction module is used to compensate for the drift of physical parameters caused by factors such as material aging, accumulation of minor damage, or changes in ambient temperature, thereby ensuring the consistency between the dynamic prediction model of the component and the physical entity.

[0088] The online model parameter correction module obtains the measured stress and deformation data from the distributed stress and deformation sensing module, denoted as... ;

[0089] Meanwhile, the online model parameter correction module obtains the currently activated parameters from the real-time rolling optimization control module. The prediction output of the dynamic prediction model for step-by-step components is denoted as... .

[0090] The online model parameter correction module generates the model residual signal by performing vector subtraction on the two signals mentioned above. ;

[0091] ;

[0092] In the formula, For model residual signals, This is the measured output vector. To predict the output vector, For time variables, This represents the model order.

[0093] The model residual signal The amplitude and dynamic characteristics directly reflect the prediction accuracy of the current component dynamic prediction model and the degree of mismatch between it and physical reality.

[0094] The online model parameter correction module employs an online parameter identification algorithm based on the model residual signal. Internal parameter vectors of the component dynamic prediction model Iterative updates will be performed.

[0095] The online model parameter calibration module constructs an augmented state vector. The augmented state vector will transform the state vector of the original component dynamic prediction model. With the internal parameter vector to be corrected The residual signals of the model are merged into a unified state vector. The extended Kalman filter algorithm then converts these residual signals into a single state vector. As measurement innovation, the augmented state vector is recursively estimated through two steps: prediction and update. In each update step, the algorithm adjusts the Kalman gain based on the magnitude and direction of the residuals and corrects the parametric components in the augmented state vector.

[0096] The output of the online model parameter calibration module consists of two parts. One is the updated intrinsic parameter vector. Updated internal parameter vector The data is fed back in real time to the real-time rolling optimization control module to correct the state matrix of the component dynamic prediction model on which it relies during rolling time-domain optimization. The second is the generated model residual signal. .

[0097] The model structure adaptive tuning module is connected to the model parameter online correction module. It is used to receive the model residual signal, perform feature analysis on the model residual signal to diagnose the structural well-being of the component dynamic prediction model, and send a model structure switching command to the real-time rolling optimization control module when a model structure mismatch is diagnosed, so as to adjust the component dynamic prediction model it uses.

[0098] Specifically, in this embodiment, the model structure adaptive tuning module receives the model residual signal, performs feature analysis on the model residual signal to diagnose the structural well-being of the component dynamic prediction model, and when a model structure mismatch is diagnosed, sends a model structure switching command to the real-time rolling optimization control module to adjust the component dynamic prediction model it uses.

[0099] The model structure adaptive tuning module is designed to handle changes in dynamic characteristics caused by significant changes in the operating conditions of components or cumulative damage.

[0100] The adaptive tuning module for the model structure includes a residual feature analysis unit and a mismatch diagnosis and decision unit.

[0101] The residual feature analysis unit is used to receive model residual signals from the online model parameter correction module. The residual signal is then subjected to time-frequency analysis to reveal its energy distribution characteristics in the time and frequency domains. One specific implementation involves the residual feature analysis unit using a short-time Fourier transform (STFT) to calculate the model residual signal. time spectrum :

[0102] ;

[0103] In the formula, Indicates the current The model residual signal corresponding to the first-order model, The window function represents the weighting function used for local time truncation in the Short-Time Fourier Transform (STFT). Represents a time-shifted variable. Represents the angular frequency variable. Represents the complex exponential kernel function. The time-spectral energy distribution function represents the residual signal of the model.

[0104] The mismatch diagnosis decision unit is connected to the residual feature analysis unit and is used to execute diagnostic decision logic based on the time-frequency analysis results to determine whether there is a model structure mismatch in the current component dynamic prediction model.

[0105] The diagnostic decision logic specifically includes:

[0106] Under-order diagnosis: If the time spectrum is... It shows that, in the absence of the current Within the high-frequency bands included in the dynamic prediction model of the step component, the model residual signal If a significant energy concentration persists, the current model is considered under-order. Under-order indicates that the physical system possesses dynamic characteristics that are not captured by the current model, resulting in insufficient model complexity.

[0107] Over-order diagnosis: If the residual signal of the model The total energy remained below a preset extremely low threshold for an extended period, and analysis of the model state indicated that the current... If the contribution of some higher-order state components of a model to the system output is negligible, then the current model is considered to be out of order. Out-of-order indicates that the model structure is too complex, causing unnecessary computational burden.

[0108] Structure well-determined: If the diagnostic conditions for under-order or over-order are not met, the current model is determined to be structurally well-determined.

[0109] When the mismatch diagnosis decision unit detects under-order or over-order, it determines that a model structure mismatch has occurred and generates a model structure switching instruction. The model structure switching instruction includes an order-up instruction (corresponding to under-order diagnosis) or an order-down instruction (corresponding to over-order diagnosis). The model structure switching instruction is sent to the real-time rolling optimization control module, instructing it to switch to and activate a more suitable order from the built-in order-down model library unit. or ) component dynamic prediction model.

[0110] An internal active prestressing execution module, which is connected to a real-time rolling optimization control module, is used to receive control commands and apply active prestress inside the steel berthing components to compensate for stress deformation.

[0111] Specifically, in this embodiment, the internal active prestressing execution module receives the control command and applies active prestress inside the steel berthing component to compensate for the stress deformation.

[0112] The active prestressing execution module includes one or more execution elements and corresponding drive control circuits.

[0113] The actuators can be high-speed electromechanical actuators (EMAs) or piezoelectric ceramic stack actuators. These actuators possess high-frequency response characteristics and high-precision displacement or force output capabilities, meeting the demands of real-time control. These actuators act as the active support in the system.

[0114] The drive control circuit receives control commands from the real-time rolling optimization control module. The control commands are a sequence of digital or analog signals, with amplitudes corresponding to the desired applied active prestress or the target displacement of the actuator. The drive control circuit parses and amplifies the control commands, converting them into the high voltage or high current required to drive the actuator.

[0115] Under the action of the drive signal, the actuator generates precise elongation or shortening, thereby applying precise tension or pressure to the structural components between its anchoring points, i.e., active prestress. Active prestress actively changes the internal force balance of the steel berthing components to offset or reconstruct the local stress concentration caused by external ship impact loads, thereby achieving adaptive compensation for stress deformation.

Claims

1. A steel berthing member stress deformation self-adaptive compensation control system, characterized in that, The application relates to a steel docking member stress deformation real-time monitoring and control system. The system comprises: a distributed stress deformation sensing module for monitoring the stress deformation of the steel docking member in real time and outputting measured stress deformation data; a real-time rolling optimization control module connected with the distributed stress deformation sensing module, for receiving the measured stress deformation data and performing rolling horizon optimization based on a component dynamic prediction model and the measured stress deformation data to generate control instructions; a model parameter online correction module connected with the distributed stress deformation sensing module and the real-time rolling optimization control module, for receiving the measured stress deformation data and obtaining the prediction output of the component dynamic prediction model, comparing the prediction output with the measured stress deformation data to generate a model residual signal, and online updating the internal parameters of the component dynamic prediction model based on the model residual signal; a model structure adaptive setting module connected with the model parameter online correction module, for receiving the model residual signal, performing feature analysis on the model residual signal to diagnose the structure adaptability of the component dynamic prediction model, and sending a model structure switching instruction to the real-time rolling optimization control module to adjust the component dynamic prediction model used by the real-time rolling optimization control module when the model structure is diagnosed as being mismatched; 2. The stress deformation self-adaptive compensation control system of a steel docking member according to claim 1, characterized in that, an internal active pre-stress execution module connected with the real-time rolling optimization control module, for receiving the control instructions and applying active pre-stress inside the steel docking member to compensate for the stress deformation. The distributed stress deformation sensing module comprises: a fiber Bragg grating array arranged in the steel docking member; a signal demodulation unit for monitoring the center wavelength drift of the fiber Bragg grating array; 3. The stress deformation self-adaptive compensation control system of a steel berthing member according to claim 1, characterized in that, a data processing unit comprising a temperature compensation mechanism and used for calculating the measured stress deformation data according to the temperature-compensated center wavelength drift. The real-time rolling optimization control module comprises a reduced-order model library unit, and the reduced-order model library unit stores a plurality of component dynamic prediction models with different orders; 4. The stress-deformation self-adaptive compensation control system of a steel docking member according to claim 1, characterized in that, The real-time rolling optimization control module is used for activating a component dynamic prediction model with a specific order from the reduced-order model library unit in response to a model structure switching instruction. The real-time rolling optimization control module performs rolling horizon optimization by solving a cost function for stress field topology management; The cost function comprises at least one of the following: a peak stress suppression term for minimizing the maximum stress value at a key position of the component; a stress gradient smoothing term for minimizing the spatial gradient norm of the stress field inside the component; 5. The stress-deformation self-adaptive compensation control system of a steel docking member according to claim 1, characterized in that, a fatigue accumulation penalty term for asymmetrically weighting and punishing tensile stress appearing in a prediction horizon. The model parameter online correction module is specifically used for: constructing an augmented state vector comprising a state vector of the component dynamic prediction model and internal parameters, and the internal parameters are expressed as an internal parameter vector; and adopting an extended Kalman filter algorithm to take the model residual signal as measurement innovation to recursively estimate the augmented state vector so as to online update the internal parameter vector.

6. The stress-deformation self-adaptive compensation control system of a steel docking member according to claim 1, characterized in that, The model structure adaptive setting module comprises a residual characteristic analysis unit, which uses short-time Fourier transform or wavelet transform to perform time-frequency analysis on the model residual signal to obtain the energy distribution characteristics of the model residual signal in time domain and frequency domain.

7. The stress-deformation self-adaptive compensation control system of a steel docking member according to claim 1, characterized in that, The model structure adaptive setting module further comprises a mismatch diagnosis decision unit, which is configured to execute a diagnosis decision logic, which comprises: When the energy distribution characteristics show that the model residual signal continuously has significant energy concentration in a high-frequency band not contained in the current component dynamic prediction model, it is determined that the component dynamic prediction model is under-ordering. When the total energy of the model residual signal continuously falls below a preset threshold for a long time, and the analysis of the model state shows that the contribution of part of the high-order state components of the current component dynamic prediction model to the system output is negligible, it is determined that the component dynamic prediction model is over-ordering.

8. The stress-shaping self-adapting compensation control system for a steel docking member according to claim 7, characterized in that, The mismatch diagnosis decision unit further generates an order-raising instruction as the model structure switching instruction when it is determined to be under-ordering, and generates an order-reducing instruction as the model structure switching instruction when it is determined to be over-ordering.

9. The stress-forming self-adaptive compensation control system of a steel docking member according to claim 1, characterized in that, The internal active prestress execution module comprises one or more execution elements and a driving control circuit; The execution element is a high-speed electro-mechanical actuator or a piezoelectric ceramic stack actuator; The execution element is integrated and installed inside the steel berthing component and is driven by the driving control circuit to apply active prestress.

10. The stress-forming self-adaptive compensation control system of a steel docking member according to claim 1, characterized in that, The model parameter online correction module constitutes a first layer adaptive loop, which is used to compensate for the drift of physical parameters caused by material aging or environmental changes; The model structure adaptive setting module constitutes a second layer adaptive loop, which is used to monitor the model residual signal of the first layer adaptive loop to diagnose and adjust the model structure mismatch caused by the change of the dynamics characteristics of the component caused by significant working condition changes or cumulative damage.

Images