A health monitoring method and system for a balanced lifting structure of an offshore wind power installation platform

Abstract

The present application relates to the technical field of offshore engineering equipment monitoring, and more particularly to a health monitoring method and system for a balanced lifting structure of an offshore wind power installation platform. The method comprises DIC image sequence acquisition and platform posture and lifting state acquisition based on DIC layout planning; rigid body motion compensation under the coupling of DIC image sequence and lifting state; full-field displacement field and strain field solving based on the compensated sequence; health index construction for the balanced lifting process based on the solved full-field displacement field and strain field; health state discrimination and risk classification in the lifting operation stage based on the health index; balanced lifting control linkage under the DIC health index feedback based on the risk level; the present application can reduce communication bandwidth occupation and remote processing delay, and is suitable for the engineering characteristics of limited on-site communication conditions, strong environmental interference and high maintenance cost of offshore wind power installation platforms.

Description

Technical Field

[0001] This invention relates to the field of marine engineering equipment monitoring technology, and in particular to a method and system for health monitoring of the balance lifting structure of an offshore wind power installation platform. Background Technology

[0002] Currently, the condition monitoring of the balancing and lifting system of offshore wind turbine installation platforms mainly relies on hydraulic pressure sensors, cylinder stroke sensors, limit switches, pin positioning detection devices, inclinometers, accelerometers, and a small number of strain gauges or fiber optic grating sensors. While these methods can reflect the operating state, load changes, and local stress response of the lifting system to some extent, they still have significant shortcomings. First, stroke, pressure, and pin condition detection primarily reflect the motion state of the lifting actuator, making it difficult to directly characterize the true deformation state of key load-bearing structures such as pile legs, pile fixing frames, lifting ring beams, and connecting nodes. Second, contact sensors such as strain gauges and fiber optic gratings typically only acquire local responses at a limited number of measuring points, making it difficult to obtain the full-field displacement, full-field strain, and local deformation gradient distribution in key areas, and easily missing stress concentration locations such as pin hole edges, weld ends, and guide contact areas.

[0003] Digital Image Correlation (DIC) is a technique that calculates the gray-level correlation of speckle images on a structural surface before and after deformation to obtain the displacement field, strain field, and local deformation evolution characteristics of the measured area. However, directly applying traditional DIC technology to the online monitoring of the balance and lifting system of offshore wind power installation platforms still faces several key technical challenges.

[0004] (1) The key structures of the balancing lifting system have strong coupled load-bearing characteristics, making the selection of monitoring objects and the layout of the field of view quite difficult. The load during the platform lifting process does not act on a single component, but is dynamically transferred between the pile legs, lifting mechanism, fixed pile frame, platform body, locking structure and hydraulic actuator. If the camera or speckle area is placed solely based on manual experience, problems such as insufficient coverage of critical load-bearing paths, redundant camera field of view, local blind spots and missed damage hotspots are likely to occur.

[0005] (2) The dynamic operating environment at sea introduces significant rigid body motion interference. The pixel displacement in the DIC image includes not only the actual deformation of the structure, but also pseudo-displacements caused by the overall platform motion, camera jitter, and changes in viewing angle. If rigid body motion and local structural deformation cannot be effectively separated, the displacement and strain fields calculated by DIC will have large errors.

[0006] (3) The health status of the balancing lifting system has obvious stage-based and process-coupled characteristics. Platform lifting typically includes different stages such as stake insertion, preloading, lifting, locking, load transfer, attitude leveling, and static operation. The stress state, risk sources, and abnormal manifestations of key structures are different in different stages.

[0007] In summary, existing monitoring methods for the balancing and lifting systems of offshore wind turbine installation platforms are insufficient to achieve non-contact, full-field, high-precision, and real-time health monitoring of critical load-bearing structures under conditions such as complex sea conditions, dynamic lifting and lowering, structural eccentric loading, and hidden damage to key nodes. There is an urgent need to propose a health monitoring method and system for critical structures of offshore wind turbine installation platform balancing and lifting systems based on DIC technology, in order to achieve real-time perception, dynamic assessment, and safety early warning of the deformation state, local damage risk, and lifting and lowering balance state of critical structures during the platform's lifting and lowering process. Summary of the Invention

[0008] To address the aforementioned problems, this invention provides a method and system for health monitoring of the balancing and lifting structure of an offshore wind power installation platform. By constructing a DIC monitoring area planning method based on load transfer paths, rigid body motion compensation and real-time calculation of full-field displacement and strain are achieved at the edge end under complex sea conditions. This enables non-contact, full-field, high-precision health monitoring and proactive safety early warning of key structures in the balancing and lifting system under dynamic offshore operating environments.

[0009] In a first aspect, the present invention provides a method for health monitoring of the balance and lifting structure of an offshore wind power installation platform, which adopts the following technical solution: A method for health monitoring of the balance lifting structure of an offshore wind power installation platform, comprising: Obtain structural drawings and operational data of the balance and lifting system for offshore wind power installation platforms; Based on the acquired structural drawings and operational data, key monitoring areas are identified and DIC layout is planned; Based on DIC layout planning, DIC image sequence acquisition and platform attitude and lifting status are obtained; Rigid body motion compensation based on the coupling effect of DIC image sequence and lifting state; The full-field displacement and strain fields are solved based on the compensated sequence. A health index for equilibrium lifting and lowering processes is constructed based on the calculated full-field displacement and strain fields. Health status assessment and risk classification during the lifting and lowering operation phase based on health indicators; Based on risk level, a balanced rise and fall control linkage is implemented under the feedback of DIC health indicators; Based on the results of the linkage control, perform edge-cloud collaborative data transmission and remote maintenance.

[0010] Secondly, a health monitoring system for the balance and lifting structure of an offshore wind power installation platform includes: The data acquisition module is configured to acquire structural drawings and operational data of the balance lifting system of the offshore wind power installation platform; The layout module is configured to identify key monitoring areas and plan the layout of DIC based on the acquired structural drawings and operational data. The acquisition module is configured to acquire DIC image sequences and obtain platform attitude and elevation status based on DIC layout planning. The compensation module is configured to compensate for rigid body motion based on the coupling effect between the DIC image sequence and the lifting state. The solver module is configured to solve the full-field displacement and strain fields based on the compensated sequence. The index module is configured to construct health indicators for the equilibrium lifting process based on the solved full-field displacement field and strain field. The judgment module is configured to determine the health status and risk classification during the lifting and lowering operation phase based on health indicators. The linkage module is configured to perform balanced lifting and lowering control linkage based on the risk level and DIC health indicator feedback. The collaboration module is configured to perform edge-cloud collaborative data transmission and remote maintenance based on the linkage control results.

[0011] Thirdly, the present invention provides a computer-readable storage medium storing a plurality of instructions adapted to be loaded and executed by a processor of a terminal device as described in the method for health monitoring of the balance lifting structure of an offshore wind power installation platform.

[0012] Fourthly, the present invention provides a terminal device, including a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, the instructions being adapted to be loaded and executed by the processor to provide a method for health monitoring of the balance lifting structure of an offshore wind power installation platform.

[0013] In summary, the present invention has the following beneficial technical effects: 1. Improved the completeness of monitoring key structures in the balance lifting system.

[0014] This invention does not rely solely on single state signals such as cylinder stroke, hydraulic pressure, pin positioning, or tilt angle. Instead, it performs full-field deformation monitoring (DIC) on key load-bearing components such as pile legs, pile fixing frames, lifting ring beams, pin holes, cylinder supports, and balancer connection nodes. This allows for the acquisition of displacement fields, strain fields, and local deformation gradients in critical areas, thereby more accurately identifying local stress concentrations, abnormal deformations, and early damage risks.

[0015] 2. It reduces the impact of the dynamic marine environment on the accuracy of DIC measurements.

[0016] This invention utilizes platform attitude, lifting stroke, camera calibration parameters, and operational status data to compensate for rigid body motion components in DIC images. It can separate the pseudo-displacements caused by the overall platform heave, roll, pitch, camera vibration, and lifting mechanism movement, while preserving the true local deformation of key structures, thus improving the stability and reliability of full-field deformation measurement under complex sea conditions.

[0017] 3. A health indicator system oriented towards the rise and fall process has been established.

[0018] This invention transforms the DIC calculation results into balance lifting health indicators such as pile leg synchronous deformation deviation, lifting ring beam torsional deformation, pin hole edge strain concentration, guide contact area wear risk, balancer connection node differential displacement and residual deformation, so that structural deformation data can directly serve to determine lifting synchronization, off-center load risk and locking safety status.

[0019] 4. The linkage between structural health monitoring and lifting control system has been realized.

[0020] This invention outputs control responses such as deceleration, pause, lock, alarm, load balancing adjustment, or lifting cycle adjustment to the lifting control system according to the risk level, so that the monitoring results can be transformed from passive display to active safety protection, reducing the risks of asynchronous platform lifting, structural load imbalance, local jamming, and locking failure.

[0021] 5. Improved the engineering adaptability of long-term online monitoring at sea.

[0022] This invention completes DIC image compensation, full-field deformation calculation, and health index extraction at the edge. Under normal conditions, it uploads a low-dimensional health state vector, and under abnormal conditions, it uploads ROI image sequences and strain cloud maps. This can reduce communication bandwidth usage and remote processing latency, and is suitable for the engineering characteristics of offshore wind power installation platforms, such as limited on-site communication conditions, strong environmental interference, and high maintenance costs. Attached Figure Description

[0023] Figure 1 is a schematic diagram of the overall structure of the key structure health monitoring method for the offshore wind power installation platform balance lifting system based on DIC technology provided in an embodiment of the present invention; Figure 2 is a schematic flowchart of the key structural health monitoring method for the offshore wind power installation platform balance lifting system based on DIC technology provided in an embodiment of the present invention; Figure 3 is a schematic diagram of the key structure of the offshore wind power installation platform balance lifting system and the layout of the DIC monitoring area provided in the embodiment of the present invention; Figure 4 is a schematic diagram of key monitoring area identification and multi-field DIC layout planning based on load transfer path provided in an embodiment of the present invention; Figure 5 is a flowchart of rigid body motion compensation processing for DIC images under the coupling effect of complex sea conditions and heave motion provided in an embodiment of the present invention. Figure 6 is a flowchart of the extraction of the full-field displacement field, strain field and equilibrium lifting health index of the key structure provided in the embodiment of the present invention; Figure 7 is a schematic diagram of the key structural health status judgment and risk classification logic provided by an embodiment of the present invention for the lifting operation stage; Figure 8 is a schematic diagram of the balance lifting control linkage based on DIC health index feedback provided in an embodiment of the present invention; Figure 9 This is a full-field displacement cloud map of the key structure provided in an embodiment of the present invention; wherein, Figure 9 (a) in the diagram is the full-field displacement contour map of the key structure under normal synchronous lifting conditions. Figure 9 (b) is the full-field displacement contour map of the key structure under eccentric loading or local jamming conditions; Figure 10 This is a full-field strain contour map of the key structure provided in an embodiment of the present invention; wherein, Figure 10 (a) in the diagram is the full-field principal strain contour plot of the key structure under the conditions of pin locking or load transfer. Figure 10 (b) is the full-field principal strain contour map of the key structure under the residual strain condition after off-center loading, lifting or unloading; Figure 11 This is a schematic diagram of the lifting state data and control response record provided in an embodiment of the present invention; wherein, Figure 11 (a) in the figure is the cylinder stroke curve of each lifting unit. Figure 11 (b) in the figure is the curve showing the maximum displacement difference versus the synchronous deformation deviation of the pile leg. Figure 11 (c) in the figure represents the curve showing the change of key components of the balanced rising and falling health state vector. Figure 11 (d) in the figure represents the control response record between the acceleration / deceleration speed command and the load balance adjustment amount; Figure 12 This is a schematic diagram of the historical health status curve of a key monitoring area provided in an embodiment of the present invention; wherein, Figure 12 (a) in the figure is the historical health status curve of the pin hole area. Figure 12 (b) in the figure is the historical curve of the risk of uneven wear in the guide contact area. Figure 12 (c) in the figure represents the historical curves of pile leg synchronization and structural response. Figure 12 (d) in the figure represents the residual deformation and maintenance review history curve; Figure 13 This is a diagram showing the balance, rise and fall health status vector and risk level provided in an embodiment of the present invention; Figure 14 These are ROI images and local strain cloud maps of abnormally rising and falling structures provided in embodiments of the present invention; wherein, Figure 14 Image (a) is an abnormal ROI image of the pin hole area. Figure 14 (b) in the diagram is a local strain contour map of the pin hole area. Figure 14 Image (c) in the diagram is an abnormal ROI image of the guide contact area. Figure 14 (d) in the diagram is a local strain contour map of the guide contact area. Detailed Implementation

[0024] The present invention will be further described in detail below with reference to the accompanying drawings.

[0025] Example 1 Reference Figure 1 This embodiment of a method for health monitoring of the balance lifting structure of an offshore wind power installation platform includes: As shown in Figure 1, this method, based on the load transfer path of the balancing lifting system, identifies key monitoring areas such as pile legs, pile fixing frame, lifting ring beam, pin holes, rack meshing area, guide contact surface, cylinder support, and balancer connection nodes. It acquires speckle image sequences of key structural surfaces through multi-field DIC image acquisition. At the edge computing end, it fuses operational data such as platform attitude, lifting stroke, hydraulic pressure, and pin status to compensate for pseudo-displacements of rigid body motion in the image sequences. Furthermore, it calculates the full-field displacement field, full-field strain field, and local deformation gradient of the key structure and constructs health indicators for the balancing lifting process. Finally, it performs risk classification based on the lifting operation stage and changes in health indicators, and feeds back abnormal results to the lifting control system, achieving linkage between structural health monitoring and lifting safety control.

[0026] Example 1 As shown in Figure 2, this embodiment provides a method for health monitoring of key structures in the balance and lifting system of an offshore wind power installation platform based on DIC technology, specifically including the following steps: Step 1: Identification of key monitoring areas and DIC layout planning based on load transfer path The balancing and lifting system of an offshore wind turbine installation platform typically consists of legs, the platform body, a pile fixing frame, a lifting drive device, a locking mechanism, a guide structure, a balancer connection structure, and a hydraulic control unit. During the platform lifting process, the load dynamically transfers between the legs, lifting mechanism, pile fixing frame, and platform body. The stress state and damage risk of critical structures differ significantly at different lifting stages. Therefore, this step first identifies the critical structural areas requiring full-field DIC monitoring and plans the camera field of view and speckle marker placement.

[0027] Step 1.1: Structural Model and Lifting Condition Data Acquisition Obtain the three-dimensional structural model, design drawings, finite element analysis results, lifting operation process, and historical operation data of the offshore wind power installation platform's balancing and lifting system. As shown in Figure 3, the three-dimensional structural model includes at least one or more of the following: pile legs, lifting ring beam, pile fixing frame, pin holes, rack and pinion engagement area, guide slider, hydraulic cylinder support, balancer connection node, and platform main body connection area.

[0028] Simultaneously, typical load conditions of the platform during stages such as pile driving, preloading, lifting, locking, load transfer, platform leveling, and static operation are obtained. These load conditions include at least one of wave load, wind load, tidal current load, hoisting load, platform self-weight load, hydraulic lifting load, and pile leg support reaction force.

[0029] Step 1.2: Extraction of Critical Bearer Path Based on the structural model and lifting condition data, a load transfer path for the balanced lifting system is established. This load transfer path includes the load path from the platform body to the lifting ring beam, the load path from the lifting ring beam to the pile legs, the load path from the hydraulic cylinder to the bearing support, the load path from the pin locking structure to the pile leg pin holes, the load balancing path from the balancer connection node to the multiple lifting units, and the contact constraint path between the guide structure and the pile legs.

[0030] Based on the results of finite element analysis or structural mechanics calculations, areas prone to localized stress concentration, localized contact deformation, fatigue crack initiation, residual deformation, or abnormal wear during lifting operations are extracted, forming a set of key monitoring areas. This set of key monitoring areas can be represented as follows: , in, Indicates the first Key monitoring areas This indicates the number of key monitoring areas. Each key monitoring area corresponds to a risk weight. The risk weights are determined based on the stress level, historical failure frequency, load variation amplitude, maintainability difficulty, and severity of failure consequences in the region.

[0031] Step 1.3: Determining the DIC speckle identification area For each key monitoring area The appropriate speckle marking area for DIC measurement is determined. This speckle marking area can be directly applied to the surface of the critical structure, or it can be applied to the surface of a detachable monitoring target plate rigidly connected to the critical structure.

[0032] Preferably, the speckle marking uses a high-contrast random speckle layer that is resistant to salt spray, water vapor, oil, UV aging, and wear. For easily worn areas such as the edges of pin holes, the rack meshing area, and the contact surface of guide sliders, rigidly connected replaceable speckle target plates can be used to avoid direct mechanical contact damage to the speckle layer. For non-direct contact areas such as the weld seams of the lifting ring beam, the cylinder support, and the connection nodes of the balancer, the speckle pattern can be applied by spraying onto the structural surface.

[0033] Step 1.4: Layout planning of multi-field image acquisition units As shown in Figure 4, the installation orientation of the multi-field image acquisition unit is determined based on the spatial location, visibility, camera installation space, protection conditions, lighting conditions, and construction interference factors of the key monitoring area. The image acquisition unit may include at least one of a binocular industrial camera, a multi-lens industrial camera, a supplementary lighting unit, an anti-salt spray protective cover, a lens defogging device, and an edge computing unit.

[0034] For critical monitoring areas requiring 3D deformation measurement, at least two cameras with a common field of view should be installed to meet the requirements of 3D DIC measurement. For areas where the primary focus is on local planar deformation, a monocular camera combined with a calibration plane can be used for 2D DIC measurement.

[0035] The goal of camera layout planning is to maximize coverage of high-risk critical monitoring areas while minimizing the number of cameras and field-of-view redundancy, provided that requirements for common field of view, spatial resolution, depth of field, and protective installation are met. Optionally, an optimization function based on risk weights can be used. , in, Indicates the first Whether the key monitoring areas are effectively covered This indicates the risk weight of the region. Indicates the number of cameras. This indicates penalties for observation degradation caused by factors such as obstruction, strong reflection, salt spray, water vapor, and construction interference. These are the weighting coefficients.

[0036] Step 2: Synchronous acquisition of multi-source data and dynamic status acquisition of the platform After completing the layout of key areas and camera installation, the system enters the online monitoring phase. This step is used to simultaneously acquire DIC image sequences and lifting system operating status data, providing a unified time reference for subsequent rigid body motion compensation and health indicator calculations.

[0037] Step 2.1: DIC Image Sequence Acquisition During the platform lifting operation, the multi-field image acquisition unit monitors the planned key monitoring areas. Continuous shooting is performed to acquire a speckle image sequence containing temporal and spatial evolution characteristics. Mathematically, this speckle image sequence is represented as a temporal set composed of a three-dimensional grayscale matrix. ,in, Represents the pixel coordinates of the image. The acquisition time is indicated. The speckle image sequence includes a reference frame and deformed frames, which are defined and separated using the following mathematical operators: (1) Reference frame Defined as the initial reference state image without significant structural deformation. The system operates according to lifting control commands. When the platform is in a static state, under standard load conditions, or in an initial locked state, the reference frame acquisition operator is triggered to extract a specific initial moment. Alternatively, the grayscale matrix after averaging multiple frames in the temporal domain can be used as a reference frame, and can be represented as follows: .

[0038] (2) Deformed frame Defined as a real-time state image during the dynamic evolution of the lifting operation. During the platform's preloading, lifting, load transfer, locking, or leveling actions, the system continuously captures the dynamic grayscale distribution at each anomaly time $t_k$ along the time axis, characterized as follows: , To ensure the accuracy and computational efficiency of digital image correlation matching under complex sea conditions, the sampling frequency of the image acquisition unit... It is not a fixed constant, but rather an adaptive multi-order adjustment based on the platform's dynamic response and structural risk status. The system constructs a dynamic adjustment operator for the sampling frequency: , Specifically, the adaptive multi-stage adjustment dynamic control strategy satisfies: , In the formula, This is the established basic sampling frequency for routine monitoring; For the relative displacement data of the pile legs The calculated real-time lifting speed, The speed threshold; For platform posture data The extracted principal vibration frequencies of the structure; A balance rise and fall health status vector fed back in real time from the edge computing end; and Let be the frequency gain coefficient, and satisfy . , This represents the rate of change of local strain between adjacent frames. The gradient threshold is set. By adjusting the above operators, data redundancy is reduced under normal conditions, and time resolution is maximized during the anomaly initiation and control linkage stages, so as to fully capture the early damage and anomaly evolution transient process of the load-bearing structure under multi-field coupling.

[0039] Step 2.2: Platform attitude and lifting status data acquisition The system synchronously acquires operational status data output from the platform attitude sensor, lifting control system, and hydraulic system. This operational status data includes at least one of the following: platform heave, roll, pitch, yaw, lateral movement, longitudinal movement, cylinder stroke, hydraulic pressure, pin locking status, lifting speed, relative displacement of the legs, platform leveling commands, and lifting control commands.

[0040] Among them, platform attitude data is used to characterize the overall motion of the platform caused by waves, wind loads and hoisting loads; lifting stroke data is used to characterize the synchronization status between different lifting units; hydraulic pressure data is used to reflect changes in lifting load; and pin status data is used to determine whether the current stage is locking, unlocking or load transfer.

[0041] Step 2.3: Unify timestamps and synchronize data The DIC image sequence and operational status data are uniformly timestamped. Preferably, image acquisition thread, attitude acquisition thread, hydraulic status acquisition thread, and lifting control command acquisition thread are established in the edge computing unit, and a multi-source synchronized data sequence is formed through hardware triggering, network time synchronization, or software time synchronization.

[0042] The synchronized data can be represented as: ,in, express DIC image data at time of moment, This represents platform posture data. This indicates the lifting stroke and the relative displacement data of the pile legs. This indicates hydraulic pressure and pin status data. This represents the lifting control command data.

[0043] Step 3: Rigid body motion compensation under the coupled effects of complex sea conditions and heave / lifting motion Offshore wind turbine installation platforms are subject to the combined effects of waves, wind loads, lifting loads, and mechanical vibrations during the lifting and lowering process. The speckle displacement in the DIC image includes not only the actual local deformation of key structures but also pseudo-displacements caused by the platform's overall heave, roll, pitch, camera vibration, and the movement of the lifting mechanism. Therefore, this step is used to separate the actual local deformation components of the structure from the total observed displacement, as shown in Figure 5.

[0044] Step 3.1: Camera calibration and coordinate system establishment Establish a global coordinate system for the platform Camera coordinate system Local coordinate system of key monitoring areas and image pixel coordinate system The camera intrinsic parameter matrix is ​​obtained through camera calibration. And distortion parameters, and obtain the extrinsic transformation matrix between the camera coordinate system and the platform global coordinate system through extrinsic parameter calibration. .

[0045] For binocular or multi-view DIC measurements, it is also necessary to calibrate the relative pose relationship between cameras and obtain the rotation matrix and translation vector between cameras in order to realize the three-dimensional reconstruction and three-dimensional displacement field calculation of the key monitoring area.

[0046] Step 3.2: Theoretical Rigid Body Displacement Estimation according to Moment platform attitude data and lifting stroke data Estimate the theoretical rigid body displacement of the key monitoring area caused by the overall motion under the condition of no local structural deformation. For any physical point within the key monitoring area... , its in The theoretical position caused solely by the overall motion of the platform at any given moment can be expressed as: , in, Indicates that the platform is The overall rotation matrix of time relative to the reference time. This represents the overall translation vector of the platform. For the monitoring area that moves with the lifting mechanism, a lifting stroke correction term can also be introduced. ,form: , Then, the camera projection model will be used to... Projecting onto the image plane yields the theoretical rigid body pixel positions.

[0047] , in express The theoretical pixel position of the rigid body at any given moment.

[0048] Step 3.3: Observation Displacement Decomposition and Compensation Location of speckle pixels observed in DIC images It includes rigid body motion components and local structural deformation components. The system is based on the theoretical rigid body pixel positions. The observed pixel positions are compensated to obtain the compensated true deformed pixel positions. : , in, This indicates the initial pixel position of the corresponding speckle in the reference frame. By performing the above compensation on each speckle sub-region or feature point, a compensated image sequence or compensated displacement data is obtained after removing the overall platform motion and camera disturbances.

[0049] Step 3.4: Reference Target Assisted Compensation In a preferred embodiment, a reference target is set within the camera's field of view, rigidly connected to the platform body but not located in a region of local deformation. The reference target is used to estimate camera shake, overall platform motion, and changes in optical viewing angle. The system performs consistency correction between the observed displacement of the reference target and the platform attitude data, further improving the accuracy of rigid body motion compensation.

[0050] If there is a deviation between the displacement observed from the reference target and the theoretical rigid body displacement obtained from the attitude estimation, the system will correct the camera extrinsic parameters or image compensation parameters online to reduce the errors caused by minor loosening, thermal drift, or vibration of the camera support during long-term monitoring.

[0051] Step 4: Calculation of the full-field displacement and strain fields of the key structure After rigid body motion compensation is completed, the edge computing end obtains a DIC image sequence that mainly reflects the true local deformation of the key structure. This step is based on the compensated image sequence to solve for full-field displacement and strain.

[0052] Step 4.1: Scattered Spot Region Division and Correlation Matching This step is based on the reference frame acquired in step 2.1 and after rigid body motion compensation in step 3. With deformed frames At the edge computing end, integer pixel matching and sub-pixel level iterative optimization are performed. The specific calculation process is as follows: Step 4.1.1: The stress gradient-based adaptive speckle sub-region window partitioning system is applied to the reference frame. The system selects the physical region to be measured and discretizes it into sub-regions along the spatial coordinate grid. To balance the measurement spatial resolution in local high-stress gradient regions with the noise resistance stability in low-stress regions, the system introduces a topography-guided adaptive sub-region partitioning operator. Let the size of the topography matrix of the speckle sub-regions to be partitioned be... The pixel's size control equation satisfies: , In the formula, This represents the prior principal strain gradient distribution of the monitoring area extracted based on the finite element model in step 1.2; This is the set stress gradient abrupt change threshold. For structural discontinuities or high-load concentration areas such as the edge of the pin hole, the end of the weld, and the rack meshing area, the system automatically converges the sub-region size to the minimum boundary value. (Preferred) (pixels), and configure dense sampling steps. Pixels are used to accurately capture displacement field features with abrupt edge changes; for non-contact areas with relatively gentle deformation, such as the fixed pile frame and the side of the lifting ring beam, the sub-region size is automatically expanded to the noise-resistant boundary value. (Preferred) (pixels) to suppress sea state noise by increasing grayscale entropy.

[0053] Step 4.1.2: Integer pixel search based on improved correlation criterion After determining the sub-region to be tested, use the center point of the reference sub-region. A local grayscale vector is constructed around the core. To combat linear grayscale drift caused by light fluctuations, salt spray, and water vapor disturbances during offshore operations, this invention employs a modified zero-mean normalized cross-correlation (ZNCOCC) criterion to calculate the similarity between the reference sub-region and the deformed sub-region. Let the spatial mapping of corresponding points within the deformed sub-region due to deformation be described by a first-order shape function: In the formula, It is an integer pixel displacement vector. The first-order spatial gradient of the displacement. The objective function for optimizing the correlation coefficient of the improved ZNCOCC. Defined as: , In the formula, and These are the grayscale values ​​of the reference frame and the deformed frame at the corresponding pixel points, respectively. and These are the average gray values ​​of the reference sub-region and the deformed sub-region, respectively. The system uses a spatiotemporal multi-scale pyramid cross search algorithm to search for the most suitable sub-region within a set search window. Reaching the maximum value (or zero-mean normalized least squares distance) The discrete coordinates of the point that reaches the minimum value are used as the initial integer pixel displacement estimate. .

[0054] Step 4.1.3: Nonlinear inverse combination Gaussian-Newton (IC-GN) subpixel iterative optimization In order to overcome the limitations of pixel physical resolution and achieve Pixel-level deformation measurement accuracy, the system estimates values ​​in integer pixels. Starting with the iteration, the Inverse Compositional Gauss-Newton (IC-GN) algorithm is introduced to solve the nonlinear least-squares manifold. The IC-GN algorithm solves the nonlinear least-squares manifold by using the deformation increment... Applying the gradient in reverse to the reference sub-region avoids the drawback of repeatedly calculating the grayscale gradient of the deformed image at each step in the conventional forward iteration, greatly reducing the computation time at the edge calculation end. Its incremental control equation for sub-pixel iteration is characterized as: , In the formula, It is a 6-DOF state vector containing displacement components and first-order deformation gradient; It is a spatial deformation mapping function; This is the grayscale gradient matrix interpolated at pixels for the reference sub-region using a bicubic spline. (Hessian matrix) Pre-calculated and kept constant in the reference image domain, defined as: In each iteration, the system obtains the deformation increment by solving a system of linear equations. And update the current state vector mapping matrix according to the inverse combination rule: The iteration termination condition is set to the incremental magnitude. Or, the maximum number of iterations can be reached (preferably 20). The first two terms of the output vector after convergence are the center points of the speckle sub-region. High-precision sub-pixel deformation displacement value after removing spurious displacement interference and , which serves as the basic input data for the reconstruction of the full-field displacement field.

[0055] Step 4.2: Reconstruction of the entire displacement field Spatial interpolation is performed on the matching results of all speckle sub-regions to obtain the continuous displacement field of the key monitoring area. For two-dimensional DIC measurements, the in-plane displacement components are output. and For three-dimensional DIC measurement, output three-dimensional displacement components. , and .

[0056] The overall displacement field is used to characterize the local deformation patterns of key monitoring areas during the lifting and lowering process. For example, discontinuities in the displacement at the edge of the pin hole can reflect abnormal local deformation of the locking structure; the displacement difference between the two sides of the lifting ring beam can reflect torsion or eccentric loading of the ring beam; and the displacement difference in key areas near different pile legs can reflect asynchronous lifting and lowering trends.

[0057] Step 4.3: Calculation of global strain field and local deformation gradient This step is based on the high-precision sub-pixel full-field displacement matrix of each spatially discrete sampling point obtained from steps 4.1 and 4.2. The spatial manifold differential and strain tensor solution are performed at the edge computing end. The specific calculation process is as follows: Step 4.3.1: Construction of Local Spatiotemporal Neighborhood Adaptive Second-Order Smooth Strain Operator Since directly performing conventional finite difference differentiation on a displacement field containing discrete, small noise will result in a severe noise amplification effect, this invention introduces an improved local spatial neighborhood adaptive second-order least squares fitting operator. This is achieved using arbitrary sampling points to be measured. Centered on, construct a system containing A neighborhood calculation window for spatial strain at each pixel is defined. Within this window, assuming the displacement field follows a second-order Taylor polynomial surface distribution, the governing equation for the local spatial evolution of displacement is: , In the formula, , The relative spatial coordinates of the sampling points within the neighborhood relative to the center point; These are the undetermined coefficients of the polynomial to be solved. The system adaptively configures the window radius based on the region category (high gradient region or flat region) to which this point belongs in step 4.1.1. : , By establishing a system of least squares overdetermined equations within a neighborhood window Furthermore, the Gram-Schmidt orthogonalization method was used to solve the matrix, directly extracting the first-order partial derivatives of the polynomial. From continuum mechanics, it is known that the center point... The displacement gradient at a point can be directly characterized as: , Step 4.3.2: Solving the Green-Lagrange large deformation strain tensor Considering that offshore wind turbine installation platforms may experience large-scale displacements or localized geometric nonlinear large deformations during preloading and load transfer, this invention employs a nonlinear Green-Lagrange strain tensor operator to solve the problem, avoiding the geometric nonlinearity errors introduced by traditional linear elastic small-deformation strain formulas. The normal strain at each physical sampling point throughout the field is... and shear strain The calculation formula is as follows: , , , Furthermore, a principal component solver for the strain tensor is introduced to calculate the maximum principal strain at each monitoring point. Minimum principal strain and principal strain azimuth : , , For the three-dimensional monitoring area using a multi-view 3D DIC branch configuration, the system introduces a 3D Green-Lagrange strain tensor. : By obtaining the eigenvalues ​​of the three-dimensional strain matrix The solution outputs the three-dimensional first principal strain, second principal strain, and third principal strain fields in the three-dimensional manifold space.

[0058] Step 4.3.3: Calculation of local deformation gradient matrix and heterogeneous strain gradient risk operator To accurately identify the evolutionary boundaries of high-risk areas, the system simultaneously extracts the deformation gradient tensor matrix. : , For harsh working conditions prone to elastoplastic transitions, such as the edge of the pin hole, the end of the weld, the guide contact area, and the connection node of the balancer, a system is constructed to identify the risk index of explicit local strain gradient abrupt changes. The aforementioned index is obtained by analyzing the maximum principal strain field. The composite representation of second derivative (i.e., calculation of the Laplace operator) and first-order gradient magnitude in the two-dimensional spatial domain: , , In the formula, The preset risk weight gain coefficient; and Directly by fitting the coefficients of the local second-order surface The linear combination space derivative is obtained, and the solution outputs the full-field strain matrix and deformation gradient risk index. This will be used as a core state variable and directly input into the health indicator vector construction in Step 5 below.

[0059] Step 4.4: Extraction of Residual Deformation After the lifting action is completed and the system enters a locked or stationary state, it continues to acquire images of key monitoring areas and compares them with reference frames or the previous safe state frame to calculate residual displacement and residual strain. If an irreversible deformation still exists in a key monitoring area after unloading or the action is completed, it is input as a residual deformation index into the subsequent health assessment module.

[0060] Step 5: Construction of health indicators for the equilibrium rise and fall process As shown in Figure 6, this step transforms the full-field displacement field, strain field, and local deformation gradient obtained from DIC calculation into health indicators that can directly reflect the equilibrium rise and fall state, avoiding the limitation of only relying on image measurement results.

[0061] Step 5.1: Pile Leg Synchronous Deformation Deviation Index For multiple pile legs or multiple lifting units, extract the characteristic displacement and characteristic strain of their corresponding key monitoring areas. Define the synchronous deformation deviation index for the pile legs. : , in, and They represent the first The and the first The characteristic deformation of the monitoring area corresponding to each pile leg. The characteristic deformation can be the average displacement, maximum displacement, peak principal strain, or residual deformation.

[0062] when An increase indicates a synchronization deviation between different legs or different lifting units, which may lead to platform instability, uneven loading of the legs, or jamming of the lifting mechanism.

[0063] Step 5.2: Torsional Deformation Index of Lifting Ring Beam DIC monitoring areas are arranged on both sides of the lifting ring beam or load-bearing frame. The vertical displacement difference, rotation angle difference, and principal strain difference between the two sides are calculated to construct the torsional deformation index of the lifting ring beam. : , in, and These represent the local rotation angles on both sides of the ring beam. and These represent the characteristic principal strains on both sides of the ring beam. This is a weighting coefficient. This index is used to determine whether the lifting ring beam has experienced eccentric torsion or local stiffness degradation.

[0064] Step 5.3: Strain Concentration Index at the Edge of the Pin Hole For pin holes, locking holes, or rack meshing areas, the system extracts the maximum principal strain, strain gradient, and local deformation discontinuity near the hole edge, and constructs a strain concentration index for the pin hole edge. : , in, Indicates the maximum principal strain. Indicates the principal strain gradient. This is a weighting factor. This index is used to identify the risk of stress concentration and crack initiation at the hole edge caused by pin locking, load transfer, or localized contact impact.

[0065] Step 5.4: Risk Indicators for Uneven Wear in the Guide Contact Area For the contact area between the pile leg and the guide structure, the system calculates the local displacement difference and strain non-uniformity on both sides or above and below the contact area, and constructs a risk index for uneven wear in the guide contact area. If this indicator continues to increase, it indicates that there may be posture deviation, uneven local contact, or a tendency for uneven wear between the pile leg and the guide structure.

[0066] Step 5.5: Differential Displacement Index of Balancer Connection Nodes For the balancer connection node, the system extracts the relative displacement difference between the two ends or multiple balancer nodes of the connection node and constructs a differential displacement index. : , in, Indicates the first The characteristic displacement of each balancer connection node. This represents the average characteristic displacement of all balancer connection nodes. This index is used to determine whether the balancer effectively distributes the load and whether there are abnormal load deviations between different lifting units.

[0067] Step 5.6: Construction of Comprehensive Health Status Vector The above indicators are combined to form a balanced rising and falling health status vector: , in, Indicates the residual deformation index. This indicates the initial risk flag at the edge. This health status vector can be updated in real time at the edge and serves as core data for risk classification, control linkage, and edge-cloud collaborative transmission.

[0068] Step 6: Health status assessment and risk classification based on the lifting operation phase The main sources of risk for a balancing lifting system differ at different operational stages. Using a uniform threshold for assessment can easily lead to false alarms or missed alarms. Therefore, this invention constructs a staged health assessment logic based on the lifting operation stage. Figure 7 As shown.

[0069] Step 6.1: Identification of Lifting Operation Phase The system identifies the current operating stage based on lifting control commands, cylinder stroke, hydraulic pressure, and pin status. The operating stage includes at least one of the following: pin unlocking stage, cylinder extension / retraction stage, synchronous lifting stage, load transfer stage, pin locking stage, platform leveling stage, and static operating stage.

[0070] Step 6.2: Selection of Stage-Specific Indicators During the unlocking or locking phase of the latch, the focus is on determining the strain concentration index at the edge of the latch hole. and residual deformation index ; During the cylinder extension and retraction phase, the key indicators to determine are the deformation of the cylinder support and the torsional deformation of the lifting ring beam. synchronous deformation deviation index of pile leg ; During the synchronous lifting phase, the key is to distinguish between different pile legs or lifting units. and ; During the load transfer phase, the key points to identify are strain concentration at the edge of the pin hole, differential displacement of the balancer connection node, and sudden changes in local strain of the fixed pile frame. During the platform leveling phase, the focus is on determining the consistency between changes in platform attitude and the deformation response of key structures. During the static operation phase, the focus is on identifying residual deformation, fatigue accumulation trends, and abnormal increases in local strain.

[0071] Step 6.3: Risk Level Determination The system is based on the health status vector With staged threshold matrix The comparison is performed, and a risk level is output. The risk levels include normal, attention, warning, and danger.

[0072] When all health indicators are within the safe threshold, it is judged to be in the normal level; When any health indicator is close to the safety threshold but does not exceed the warning threshold, it is judged to be at the level of concern; A warning level is established when any health indicator exceeds the warning threshold or shows a continuous upward trend within a consecutive time window. When any health indicator exceeds the danger threshold, or multiple health indicators show abnormal coupled growth at the same time, or when the DIC image shows obvious local deformation and discontinuity in key areas, it is judged as a dangerous level.

[0073] Step 6.4: Anomaly Area Location and Visualization Output When the risk level determined in step 6.3 reaches the warning or danger level, the system activates the reverse tracking mechanism based on abnormal health indicators (such as...). The deformation field variation characteristics contained in the image are automatically located using a spatial connected domain aggregation algorithm to identify structural anomaly sources, and visualization and low-dimensional spatiotemporal image cropping output are achieved. The specific calculation process is as follows: Step 6.4.1: Pixel set mapping and binarization determination for anomalous speckle sub-regions The system extracts the coordinates of the underlying speckle sub-regions corresponding to the abnormal health indicators that trigger the early warning level. Assume the monitoring area has a total of... Each speckle sub-region is characterized by the set of pixel coordinates of its center point as follows: The system utilizes the explicit local strain gradient abrupt change risk index of each sub-region calculated and output in step 4.3.3. or maximum principal strain Construct a spatial grayscale mapping function to generate a binary feature matrix. : , In the formula, This represents the average spatial statistical mean of damage risk indicators for all sub-regions within the current field of view. The preset dynamic amplification gain coefficient (preferably) ); For the matrix in step 6.3 for the current running stage The established phased basic strain health threshold. The coordinates of these points are identified as the core points of abnormal speckle patterns indicating localized damage, crack initiation, severe wear, or overload deformation.

[0074] Step 6.4.2: Geometric localization solution of outlier centroids based on spatial connected component aggregation To eliminate isolated noise points caused by strong sunlight reflection and water vapor refraction at sea, the system modifies the feature matrix. Perform a spatial connected component scan aggregation algorithm based on 8-neighborhood. Filter out regions whose area (i.e., the number of sub-regions they contain) is greater than the grid filtering threshold. Maximum Spatial Connectivity .

[0075] Subsequently, the system employs a centroid localization operator based on strain weighting to calculate the coordinates of the core geometric center of the critical structural anomaly region in the image pixel coordinate system. : , Furthermore, the system utilizes the camera intrinsic parameter matrix obtained in step 3.1 calibration. and extrinsic transformation matrix The pixel centroid coordinates of the image plane Back-projected onto the 3D structural model of the offshore wind power installation platform's balance and lifting system, the anomaly source is calculated and output in the platform's global coordinate system. The actual physical location in space This enables precise millimeter-level three-dimensional spatial positioning of hidden damage hotspots such as pin holes and welds.

[0076] Step 6.4.3: Adaptive Spatiotemporal ROI Matrix Clipping and Visualization Binding The system uses the calculated pixel centroid Centered on, combined with connected components circumscribed rectangle boundary width With height And introduce a safety redundancy margin factor. (Preferred) This automatically constructs a spatiotemporal region of interest (ROI) clipping window. The pixel boundaries of the ROI window are defined as follows: , Based on this boundary parameter, the system adaptively crops the deformed frame image sequence. Obtain low-dimensional, high-dynamic ROI image sequences Simultaneously, the system packages and binds the following elements at the edge computing end to form a data tuple for visualizing structural health anomalies. : , In the formula, This represents the current identified stage of platform dynamic operation. This is the pseudo-color cloud map matrix of the full-field displacement and principal strain within the corresponding time domain. This data tuple This will serve as the core visual credential, which will be directly output to step 7 below to trigger the proactive security linkage command, and input to step 8 for end-to-end cloud collaborative low-bandwidth backhaul and remote secondary security maintenance decision-making.

[0077] Step 7: Balance Lifting and Lowering Control Linkage Based on DIC Health Indicator Feedback As shown in Figure 8, this step is used to feed back the structural health monitoring results to the lifting control system, so that the monitoring results can not only be used for display and alarm, but also participate in the balance lifting safety protection.

[0078] Step 7.1: Enhanced Monitoring at the Attention Level When the risk level is designated as "Attention Level," the system sends an alert to the remote monitoring platform and increases the sampling frequency of DIC images and the calculation frequency of health indicators for the corresponding key monitoring area. Simultaneously, the system records the historical trends of the area to determine if there is a risk of continued growth.

[0079] Step 7.2: Adjustment of parameters for raising and lowering the warning level When the risk level reaches the warning level, the system sends suggestions to the lifting control system for deceleration, adjustment of the lifting cycle, or load balancing. The lifting control system can adjust the corresponding cylinder speed, pressure control parameters, or leveling strategies based on the location of the abnormal area on the pile leg or lifting unit to reduce the risk of eccentric loading and further expansion of structural deformation.

[0080] Step 7.3: Active safety protection under hazardous levels When the risk level is deemed hazardous, the system sends commands to the lifting control system to pause lifting, lock, stop operation, or issue an emergency alarm. Simultaneously, it packages and sends the ROI image sequence of the abnormal area, the overall strain cloud map, platform attitude data, hydraulic pressure data, and lifting stroke data to the remote monitoring platform for verification by maintenance personnel.

[0081] Step 7.4: Status verification after control feedback After the lifting control system executes responses such as deceleration, pause, or lock, the system continues to acquire DIC image sequences and recalculates the health status vector. If the abnormal indicators drop to a safe range, the high-level alarm will be deactivated and regular monitoring will resume; if the abnormal indicators continue to rise, the danger level will be maintained and a maintenance command will be issued.

[0082] Step 8: End-to-Cloud Collaborative Data Transmission and Remote Maintenance Decisions Considering the limited communication bandwidth on the offshore wind power installation platform, this embodiment also includes an edge-cloud collaborative data transmission step. The upload template already uses the patented approach of "low-dimensional feature extraction at the edge and abnormal state ROI image transmission". This embodiment retains the engineering concept, but changes the transmission objects to balanced rise and fall health state vectors and abnormal rise and fall structure images.

[0083] Step 8.1: Upload low-dimensional health status under normal conditions When the risk level is normal or at the level of concern, the edge computing terminal only uploads a low-dimensional health status vector. This includes timestamps, monitoring area numbers, platform elevation / relief phases, and basic operational status data. This method reduces the amount of raw image transmission and lowers the burden on maritime communication links.

[0084] Step 8.2: ROI Image Backhaul under Abnormal Conditions When the risk level is warning or danger level, the edge computing terminal adaptively crops the abnormal region ROI image sequence according to the coordinates of the abnormal speckle sub-region, and simultaneously uploads the corresponding full-field strain cloud map, full-field displacement cloud map, rise and fall status data and control response record.

[0085] Step 8.3: Cloud Trend Analysis and Maintenance Decisions After receiving data uploaded from the edge devices, the cloud platform establishes historical health status curves for each key monitoring area and analyzes the long-term trends of indicators such as pile leg synchronization deviation, ring beam torsion, pin hole strain concentration, and residual deformation. If the health indicators of a certain key area continue to approach the threshold or the frequency of abnormal occurrences increases, maintenance recommendations are generated, including at least one of the following: shutdown inspection, structural retesting, speckle target plate replacement, pin hole inspection, guide structure inspection, or cylinder support verification.

[0086] Experimental verification To verify the feasibility and effectiveness of the health monitoring method and system for the balancing and lifting structure of an offshore wind power installation platform as described in this invention, this embodiment constructs an equivalent verification platform for the key structures of the balancing and lifting system. This verification platform includes a simulated leg structure, a simulated lifting ring beam structure, a pin hole locking structure, a guide contact structure, a hydraulic loading unit, an attitude disturbance simulation unit, a multi-field-of-view (DIC) image acquisition unit, an edge computing unit, and a lifting control simulation unit. Specifically, the simulated leg structure simulates the vertical load-bearing components during the platform's lifting process; the simulated lifting ring beam structure simulates load transfer and local torsional deformation; the pin hole locking structure simulates strain concentration at the hole edges during locking, unlocking, and load transfer; and the guide contact structure simulates localized wear and uneven contact between the leg and the guide mechanism.

[0087] To verify the effectiveness of rigid body motion compensation under complex sea conditions and the coupling effect of heave and sag, low-frequency attitude disturbances such as roll, pitch, and heave were introduced into the experimental platform, superimposed with small-amplitude high-frequency vibrations, to simulate the overall motion of an offshore wind power installation platform under the action of waves, wind loads, and equipment vibrations. Experimental results show that the uncompensated DIC displacement field contains significant overall translation and pseudo-displacement due to viewpoint changes, which easily leads to local deformations in key areas being masked by the overall motion. After employing the rigid body motion compensation method described in this invention, based on platform attitude, heave and sag stroke, camera calibration parameters, and reference target observation information, the overall motion components in the image sequence are effectively weakened. The compensated displacement field can more effectively reflect the true local deformations of the pin hole edge, the side of the lifting ring beam, and the guide contact area. These results demonstrate that this invention can extract the true deformation information of key structures under conditions of overall platform motion disturbance, exhibiting good adaptability to dynamic marine environments.

[0088] The full-field displacement and full-field strain fields of the key monitoring area were calculated based on the compensated DIC image sequence. Figure 9 Figure 9 shows the overall displacement cloud map of the key structure obtained in this embodiment of the invention. As can be seen from Figure 9, under normal synchronous lifting conditions, the displacement distribution in each monitoring area is generally continuous, the displacement difference between the two sides of the lifting ring beam is small, and no obvious local abrupt changes occur in the relevant monitoring areas of the pile legs. After applying an eccentric load or simulating local jamming, a large concentration of vertical displacement occurs on one side of the lifting ring beam, and local uneven displacement occurs in the guide contact area. This result demonstrates that the present invention can intuitively reflect the synchronization deviation, eccentric load deformation, and local contact anomalies during the balanced lifting process through the overall displacement cloud map.

[0089] Figure 10 shows the full-field strain cloud map of the key structure obtained in the embodiment of the present invention. As can be seen from Figure 10, under the conditions of pin locking and load transfer, the maximum principal strain in the edge region of the pin hole is significantly higher than that in the surrounding region, and a local strain concentration zone is shown along the edge of the hole; under the condition of eccentric loading and lifting, the principal strain distribution on both sides of the lifting ring beam is asymmetrical, and a high strain gradient appears near the ends of local welds and connection nodes; after unloading, if residual strain still exists in a local area, it indicates that there may be irreversible deformation or early damage risk in that area. This result shows that the present invention can not only identify the deformation amplitude of the structure, but also locate the strain concentration area and the potential damage initiation location.

[0090] To verify the effectiveness of the health index system described in this invention, the following indices were calculated: synchronous deformation deviation of the pile leg, torsional deformation of the lifting ring beam, strain concentration at the edge of the pin hole, risk of uneven wear in the guide contact area, differential displacement of the balancer connection node, and residual deformation. A balanced lifting health state vector was then constructed. Figure 11 shows the lifting state data and control response records obtained during the experiment. As can be seen from Figure 11, during the normal lifting phase, the cylinder stroke changes smoothly, the displacement difference between different lifting units remains within the set range, and all components of the health state vector are within the normal threshold. After the application of an off-center load disturbance, the torsional deformation index of the lifting ring beam and the synchronous deformation deviation index of the pile leg first rise, and then the system enters a level of concern or warning. When the control system performs deceleration or load balancing adjustment, the above indices gradually decrease and return to a safe range. This result demonstrates that the health state vector constructed in this invention can reflect changes in the lifting operation state and provide an effective basis for lifting control linkage.

[0091] Figure 12 shows the historical health status curves in an embodiment of the present invention. In the experiment, the edge computing unit updates the health indicators of each key monitoring area at fixed time intervals and uploads the low-dimensional health status vector to the cloud platform. Figure 12 As can be seen, under normal operating conditions, the overall fluctuations of various indicators are small, and the long-term trend is stable. When there is local damage to the simulated pin hole, uneven wear of the guide contact, or asynchronous lifting, the corresponding health indicators show continuous growth or abrupt changes. After the abnormal state is resolved or control intervention is implemented, some indicators can return to the normal range, while if the residual deformation indicators remain at a high level, it indicates that the area needs further shutdown inspection or maintenance review. These results demonstrate that the present invention can support historical tracking, trend analysis, and maintenance decision-making for the health status of critical structures.

[0092] To verify the effectiveness of abnormal area localization and end-to-end cloud collaborative transmission, three typical anomalies were set up in the experiment: local strain concentration at the edge of the pin hole, unilateral eccentric loading of the lifting ring beam, and local eccentric wear in the guide contact area. Figure 13 shows the balanced lifting health status vector and risk level display results output by the embodiment of the present invention. Figure 14 shows the ROI image of the abnormal lifting structure and the corresponding local strain cloud map. The experimental results show that when the risk level is normal or of concern, the system only uploads the timestamp, monitoring area number, lifting stage, basic operating status data, and health status vector; when the risk level reaches the warning or danger level, the system can automatically crop the abnormal ROI image sequence according to the coordinates of the abnormal speckle sub-region and simultaneously upload the corresponding full-field displacement cloud map, full-field strain cloud map, lifting status data, and control response record. Compared with continuously uploading the full original image, this method can reduce redundant data transmission, allowing the remote platform to prioritize the acquisition of key evidence images and health indicators related to the anomaly, thus making it more suitable for field applications with limited communication bandwidth at sea.

[0093] In summary, the experimental results demonstrate that this invention can complete multi-field DIC image acquisition, multi-source state synchronization, rigid body motion compensation, full-field displacement and strain field calculation, health index construction, staged risk classification, abnormal area location, and lifting control linkage under the equivalent working conditions of the key structures in a balanced lifting system. The results, including full-field displacement cloud maps, full-field strain cloud maps, lifting state data, control response records, historical health state curves, balanced lifting health state vectors, and abnormal lifting structure images, show that this invention can effectively identify typical risks such as asynchronous lifting, ring beam torsion, strain concentration at the edge of the pin hole, guide contact wear, and residual deformation, proving the practical feasibility and engineering application effectiveness of this solution.

[0094] Example 2 This embodiment provides a health monitoring system for the balance lifting structure of an offshore wind power installation platform.

[0095] A computer-readable storage medium storing a plurality of instructions adapted for loading and execution by a processor of a terminal device, the method for health monitoring of the balance lifting structure of an offshore wind power installation platform.

[0096] A terminal device includes a processor and a computer-readable storage medium, the processor being used to implement various instructions; the computer-readable storage medium being used to store multiple instructions, the instructions being adapted to be loaded and executed by the processor as described in the method for health monitoring of the balance lifting structure of an offshore wind power installation platform.

[0097] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for health monitoring of the balance lifting structure of an offshore wind power installation platform, characterized in that, include: Obtain structural drawings and operational data of the balance and lifting system for offshore wind power installation platforms; Based on the acquired structural drawings and operational data, key monitoring areas are identified and DIC layout is planned; Based on DIC layout planning, DIC image sequence acquisition and platform attitude and lifting status are obtained; Rigid body motion compensation based on the coupling effect of DIC image sequence and lifting state; The full-field displacement and strain fields are solved based on the compensated sequence. A health index for equilibrium lifting and lowering processes is constructed based on the calculated full-field displacement and strain fields. Health status assessment and risk classification during the lifting and lowering operation phase based on health indicators; Based on risk level, a balanced rise and fall control linkage is implemented under the feedback of DIC health indicators; Perform edge-cloud collaborative data transmission and remote maintenance based on the results of linkage control; The health index for the equilibrium lifting process is constructed based on the calculated full-field displacement and strain fields. This includes extracting the characteristic displacement and characteristic strain of the corresponding key monitoring areas for multiple pile legs and multiple lifting units, and defining the synchronous deformation deviation index of the pile legs. : in, and They represent the first The and the first The characteristic deformation of each pile leg corresponds to the monitoring area; then, DIC monitoring areas are arranged on both sides of the lifting ring beam or bearing frame, and the vertical displacement difference, rotation angle difference, and principal strain difference of the two sides are calculated to construct the torsional deformation index of the lifting ring beam. : ,in, and These represent the local rotation angles on both sides of the ring beam. and These represent the characteristic principal strains on both sides of the ring beam. The weighting coefficients are used; for the pin hole, locking hole, or rack meshing area, the maximum principal strain, strain gradient, and local deformation discontinuity near the hole edge are extracted to construct the strain concentration index for the pin hole edge. : ,in, Indicates the maximum principal strain. Indicates the principal strain gradient. The weighting coefficients are used; for the contact area between the pile leg and the guide structure, the local displacement difference and strain non-uniformity at the two sides or above and below the contact area are calculated to construct a guide contact area wear risk index. Finally, for the balancer connection nodes, the relative displacement difference between the two ends or multiple balancer nodes of the connection node is extracted to construct a differential displacement index. : ,in, Indicates the first The characteristic displacement of each balancer connection node. This represents the average characteristic displacement of all balancer connection nodes, and finally, the indices are combined to form a balance rise and fall health state vector: ,in, Indicates the residual deformation index. This indicates the initial risk marker at the edge.

2. The method for health monitoring of the balance lifting structure of an offshore wind power installation platform according to claim 1, characterized in that, The identification of key monitoring areas and DIC layout planning based on the acquired structural drawings and operational data includes establishing the load transfer path of the balanced lifting system based on the structural model and lifting condition data. Based on the finite element analysis results, areas prone to localized stress concentration, localized contact deformation, fatigue crack initiation, residual deformation, or abnormal wear during lifting operations are extracted to form a set of key monitoring areas. This set of key monitoring areas is represented as follows: ,in, Indicates the first Key monitoring areas This indicates the number of key monitoring areas, with each key monitoring area corresponding to a risk weight. Then, for each key monitoring area The process involves identifying suitable speckle marking areas for DIC measurements; finally, based on the spatial location, visibility, camera installation space, protection conditions, lighting conditions, and construction interference factors of the key monitoring areas, determining the installation pose of the multi-field-of-view image acquisition units; the goal of camera layout planning is to maximize the coverage of high-risk key monitoring areas while reducing the number of cameras and field-of-view redundancy, under the conditions of meeting common field of view, spatial resolution, depth of field, and protective installation requirements, using an optimization function based on risk weights. ,in, Indicates the first Whether the key monitoring areas are effectively covered This indicates the risk weight of the region. Indicates the number of cameras. This indicates penalties for observation degradation caused by factors such as obstruction, strong reflection, salt spray, water vapor, and construction interference. These are the weighting coefficients.

3. The method for health monitoring of the balance lifting structure of an offshore wind power installation platform according to claim 2, characterized in that, The acquisition of DIC image sequences and the acquisition of platform attitude and lifting status based on DIC layout planning include continuously acquiring key monitoring areas using image acquisition units during platform lifting operations. A speckle image sequence, the speckle image sequence including a reference frame and deformed frames This leads to the construction of a dynamic sampling frequency adjustment operator: The adaptive multi-stage adjustment dynamic control strategy satisfies: , In the formula, This is the established basic sampling frequency for routine monitoring; For the relative displacement data of the pile legs The calculated real-time lifting speed, The speed threshold; For platform posture data The extracted principal vibration frequencies of the structure; A balance rise and fall health status vector fed back in real time from the edge computing end; and Let be the frequency gain coefficient, and satisfy . , This represents the rate of change of local strain between adjacent frames. A gradient threshold is set; then, the operating status data output by the platform attitude sensor, lifting control system, and hydraulic system are acquired synchronously; among them, the platform attitude data is used to characterize the overall platform motion caused by waves, wind load, and hoisting load; the lifting stroke data is used to characterize the synchronization status between different lifting units; the hydraulic pressure data is used to reflect changes in lifting load; and the pin status data is used to determine whether the current stage is locking, unlocking, or load transfer. Finally, the DIC image sequence and operating status data are uniformly timestamped, and image acquisition threads, attitude acquisition threads, hydraulic status acquisition threads, and lifting control command acquisition threads are established in the edge computing unit. A multi-source synchronized data sequence is formed through hardware triggering, network time synchronization, or software time synchronization. The synchronized data is represented as follows: ,in, express DIC image data at time of moment, This represents platform posture data. This indicates the lifting stroke and the relative displacement data of the pile legs. This indicates hydraulic pressure and pin status data. This represents the lifting control command data.

4. The method for health monitoring of the balance lifting structure of an offshore wind power installation platform according to claim 3, characterized in that, The rigid body motion compensation based on the coupling effect of DIC image sequence and lifting state includes first separating the actual local deformation components of the structure from the total observed displacement, and establishing the platform's global coordinate system. Camera coordinate system Local coordinate system of key monitoring areas and image pixel coordinate system Obtain the camera intrinsic parameter matrix through camera calibration. And distortion parameters, and obtain the extrinsic transformation matrix between the camera coordinate system and the platform global coordinate system through extrinsic parameter calibration. Then according to Moment platform attitude data and lifting stroke data Estimate the theoretical rigid body displacement of the key monitoring area caused by the overall motion under the condition of no local structural deformation, for any physical point within the key monitoring area. , its in The theoretical position caused solely by the overall motion of the platform at any given moment is expressed as: ,in, Indicates that the platform is The overall rotation matrix of time relative to the reference time. This represents the overall translation vector of the platform. For the monitoring area that moves with the lifting mechanism, a lifting stroke correction term is introduced. ,form: Subsequently, the model was projected using a camera. Projecting onto the image plane, we obtain the theoretical rigid body pixel positions: ,in express The theoretical rigid body pixel position at each moment, and the speckle pixel position observed in the final DIC image. Includes rigid body motion components and local structural deformation components, based on theoretical rigid body pixel positions. The observed pixel positions are compensated to obtain the compensated true deformed pixel positions. : in, This indicates the initial pixel position of the corresponding speckle in the reference frame.

5. A method for health monitoring of the balance lifting structure of an offshore wind power installation platform according to claim 4, characterized in that, The solution of the full-field displacement and strain fields based on the compensated sequence includes the acquisition of reference frames after rigid body motion compensation. With deformed frames At the edge computing end, integer pixel matching and sub-pixel level iterative optimization are performed. First, an adaptive speckle sub-region windowing system based on stress gradient is used in the reference frame. Select the physical region to be measured and discretize it into sub-regions along the spatial coordinate grid. Introduce an adaptive sub-region partitioning operator guided by morphological features. Let the size of the morphological matrix of the speckle sub-regions to be partitioned be... The pixel's size control equation satisfies: In the formula, This represents the a priori principal strain gradient distribution; Stress gradient abrupt change threshold; After determining the sub-region to be tested, use the center point of the reference sub-region. A local grayscale vector is constructed for the core. To resist linear grayscale drift, the improved zero-mean normalized cross-correlation ZNCOCC criterion is used to calculate the similarity between the reference sub-region and the deformed sub-region. It is assumed that the spatial mapping of corresponding points in the deformed sub-region due to deformation is described by a first-order shape function. In the formula, It is an integer pixel displacement vector. The objective function for optimizing the correlation coefficient of the improved ZNCOCC model is the first-order spatial gradient of the displacement. Defined as: , In the formula, and These are the grayscale values ​​of the reference frame and the deformed frame at the corresponding pixel points, respectively. and The average gray values ​​of the reference sub-region and the deformed sub-region are respectively used. A pyramid cross search algorithm based on spatiotemporal multi-scale is employed to search for a suitable region within a defined search window. The discrete coordinates point that reaches the maximum value are used as the initial integer pixel displacement estimate. .

6. The method for health monitoring of the balance lifting structure of an offshore wind power installation platform according to claim 5, characterized in that, The method for solving the full-field displacement and strain fields based on the compensated sequence also includes introducing an inverse combined Gauss-Newton IC-GN algorithm for nonlinear least-squares manifold solving, by using the deformation increment... The reverse application to the reference sub-region reduces the computation time at the edge computing end, and its incremental control equation for sub-pixel iteration is characterized as follows: , In the formula, It is a 6-DOF state vector containing displacement components and first-order deformation gradient; It is a spatial deformation mapping function; The reference sub-region's bicubic spline interpolation grayscale gradient matrix at each pixel, and the Hessian matrix. Pre-calculated and kept constant in the reference image domain, defined as: ； In each iteration, the deformation increment is obtained by solving a system of linear equations. And update the current state vector mapping matrix according to the inverse combination rule: The iteration termination condition is set to the incremental magnitude. Or reach the set maximum number of iterations; then perform spatial interpolation on the matching results of all speckle sub-regions to obtain the continuous displacement field of the key monitoring area. For two-dimensional DIC measurements, output the in-plane displacement components. and ; For three-dimensional DIC measurement, output three-dimensional displacement components. , and .

7. A method for health monitoring of the balance lifting structure of an offshore wind power installation platform according to claim 6, characterized in that, The process of solving the full-field displacement and strain fields based on the compensated sequence also includes solving for the high-precision sub-pixel full-field displacement matrix of each spatially discrete sampling point. At the edge computing end, spatial manifold differentiation and strain tensor solution are performed. First, an improved local spatial neighborhood adaptive second-order least squares fitting operator is introduced, with arbitrary sampling points to be determined. Centered on, constructing a system containing A spatial strain neighborhood calculation window for each pixel is defined. Within this window, the displacement field is assumed to satisfy a second-order Taylor polynomial surface distribution, and its local displacement spatial evolution governing equation is characterized as follows: In the formula, , The relative spatial coordinates of the sampling points within the neighborhood relative to the center point; The coefficients of the polynomial to be solved are given; and the window radius is adaptively configured according to the region category to which the point belongs. ; By establishing a system of least squares overdetermined equations within a neighborhood window The Gramm-Schmidt orthogonalization method is used to solve the matrix, and the first-order partial derivatives of the polynomial are directly extracted. From the principles of continuum mechanics, the center point... The displacement gradient at a point is directly characterized as: Finally, a nonlinear Green-Lagrange strain tensor operator was used for solution, and the maximum principal strain at each monitoring point was calculated by introducing a solution operator based on the principal components of the strain tensor. Minimum principal strain and principal strain azimuth : , For the stereo monitoring area using a multi-view 3D DIC branch configuration, a 3D Green-Lagrange strain tensor is introduced. : By obtaining the eigenvalues ​​of the three-dimensional strain matrix The algorithm calculates and outputs the three-dimensional first principal strain, second principal strain, and third principal strain fields in the three-dimensional manifold space; to accurately identify the evolution boundary of high-risk regions, the deformation gradient tensor matrix is ​​extracted simultaneously. : Meanwhile, for harsh working conditions prone to elastoplastic transitions, an explicit local strain gradient abrupt change risk index is constructed. By analyzing the maximum principal strain field Composite representation of second derivative and first-order gradient magnitude in two-dimensional spatial domain: , , In the formula, The preset risk weight gain coefficient; and Directly by fitting the coefficients of the local second-order surface The linear combination space derivative is obtained, and the solution outputs the full-field strain matrix and deformation gradient risk index. As a core state variable, it is directly input into the construction of the health indicator vector in real time.

8. A method for health monitoring of the balance lifting structure of an offshore wind power installation platform according to claim 7, characterized in that, The health status assessment and risk classification based on health indicators for the lifting operation stage includes identifying the current operation stage based on lifting control commands, cylinder stroke, hydraulic pressure, and pin status. Specifically, during the pin unlocking or locking stage, the strain concentration index at the pin hole edge is determined. and residual deformation index During the cylinder extension and retraction phase, determine the deformation of the cylinder support and the torsional deformation of the lifting ring beam. synchronous deformation deviation index of pile leg During the synchronous lifting phase, the distance between different pile legs or lifting units is determined. and During the load transfer phase, the strain concentration at the edge of the pin hole, the differential displacement of the balancer connection node, and the abrupt change in local strain of the fixed pile frame are identified. During the platform leveling phase, the consistency between platform attitude changes and the deformation response of key structures is assessed. During the static operation phase, the residual deformation, fatigue accumulation trend, and abnormal increase in local strain are identified. Then, based on the health state vector... With staged threshold matrix The risk level is compared and output; when the determined risk level reaches the warning or danger level, the coordinates of the underlying speckle sub-regions corresponding to the abnormal health indicators that triggered the warning level are first extracted. Let the total number of monitoring areas be... Each speckle sub-region is characterized by the set of pixel coordinates of its center point as follows: Using explicit local strain gradient abrupt change risk indicators for each sub-region or maximum principal strain Construct a spatial grayscale mapping function to generate a binary feature matrix. : In the formula, This represents the average spatial statistical mean of damage risk indicators for all sub-regions within the current field of view. This is the preset dynamic amplification gain coefficient; Matrix for the current running stage The established phased basic strain health threshold, The coordinates of these points are identified as the core points of abnormal speckle patterns indicating localized damage crack initiation, severe wear, or overload deformation. Then, a geometric location calculation of the abnormal centroids is performed based on spatial connected domain aggregation. To eliminate interference from isolated noise points caused by strong reflections of sunlight and water vapor refraction at sea, the binarized feature matrix is ​​modified. Execute an 8-neighborhood-based spatial connected component scanning aggregation algorithm to filter out components with an area greater than the grid filtering threshold. Maximum Spatial Connectivity Subsequently, a centroid localization operator based on strain weighting was used to calculate the core geometric center coordinates of the critical structural anomaly region in the image pixel coordinate system. : Using the camera intrinsic parameter matrix and extrinsic transformation matrix The pixel centroid coordinates of the image plane Back-projected onto the 3D structural model of the offshore wind power installation platform's balance and lifting system, the anomaly source is calculated and output in the platform's global coordinate system. The actual physical location in space This enables precise millimeter-level three-dimensional spatial positioning of hidden damage hotspots such as pin holes and welds; finally, the calculated pixel centroid is used as the basis for this positioning. Centered on, combined with connected components circumscribed rectangle boundary width With height And introduce a safety redundancy margin factor. Automatically construct a spatiotemporal region of interest (ROI) clipping window. The pixel boundaries of the ROI window are defined as follows: Based on this boundary parameter, the deformed frame image sequence is adaptively cropped. Obtain low-dimensional, high-dynamic ROI image sequences .

9. A health monitoring system for the balancing and lifting structure of an offshore wind power installation platform, comprising the health monitoring method for the balancing and lifting structure of an offshore wind power installation platform as described in claim 1, characterized in that, include: The data acquisition module is configured to acquire structural drawings and operational data of the balance lifting system of the offshore wind power installation platform; The layout module is configured to identify key monitoring areas and plan the DIC layout based on the acquired structural drawings and operational data. The acquisition module is configured to acquire DIC image sequences and obtain platform attitude and elevation status based on DIC layout planning. The compensation module is configured to compensate for rigid body motion based on the coupling effect between the DIC image sequence and the lifting state. The solver module is configured to solve the full-field displacement and strain fields based on the compensated sequence. The index module is configured to construct health indicators for the equilibrium lifting process based on the solved full-field displacement field and strain field. The judgment module is configured to determine the health status and risk classification during the lifting and lowering operation phase based on health indicators. The linkage module is configured to perform balanced lifting and lowering control linkage based on the risk level and DIC health indicator feedback. The collaboration module is configured to perform edge-cloud collaborative data transmission and remote maintenance based on the linkage control results.

Images