A method for intelligent optimization of the hoisting path of a hinged sinker unit

Abstract

This invention relates to the field of underwater engineering construction technology and discloses an intelligent optimization method for the hoisting path of a hinged submerged raft unit. The method includes dividing the expected total hoisting cycle into multiple discrete time slices and constructing a three-dimensional predictive environment model based on multi-level water flow profile data; calculating the predicted tension of each grid node of the submerged raft by combining the hydrodynamic resistance model and the three-dimensional predictive environment model, generating a flexible deformation energy state matrix; inspecting the flexible deformation energy state matrix to determine local tension peaks; introducing rigid hanger tilt angle control variables and assigning non-equivalent rate commands to slices with peaks, generating a globally optimal equipment workflow scheduling table; executing according to the scheduling table, extracting the total predicted tension, verifying the deviation between the actual physical load equivalent mean calculated from the frequency converter driver feedback data and the total predicted tension; and truncating commands and triggering resampling to replan the remaining time slices when the deviation exceeds the limit.

Description

Technical Field

[0001] This invention relates to the field of underwater engineering construction technology, specifically to an intelligent optimization method for the hoisting path of a hinged submerged unit. Background Technology

[0002] In water conservancy and waterway improvement projects, hinged sluice gates are commonly used for riverbed protection and levee reinforcement. The hoisting and lowering process of the sluice gates requires traversing water of a certain depth. During the lowering phase, the sluice gates are subjected to the combined effects of water flow erosion and their own weight, resulting in fluid-structure interaction.

[0003] In actual underwater operating environments, water bodies typically exhibit multi-level variations in velocity and direction, resulting in flow shear. Current submersible hoisting methods usually rely on fixed lowering rates or manual experience, making it impossible to acquire and model the flow field changes at different underwater depths in advance.

[0004] This control method makes it difficult to predict the local stress state of each node of the raft at different water depths. When the raft is subjected to water flow impact and generates a peak tension, the system cannot actively adjust the raft's water-facing posture to disperse the impact force, resulting in excessive stress at the raft's joints and physical tearing.

[0005] Meanwhile, under the continuous thrust of the water flow, the submersible will experience lateral slippage, deviating from the predetermined lowering trajectory. Existing operational control methods often disconnect force control from trajectory control.

[0006] Maintaining a horizontal lowering posture during hoisting increases the risk of exceeding stress limits; however, altering the raft's posture solely to avoid water flow impact makes it difficult to meet spatial displacement requirements. This can lead to deviations in the raft's positioning during placement, failing to meet the project's positioning requirements.

[0007] Furthermore, existing lifting and hoisting equipment lacks the ability to respond to and avoid sudden water flow loads in its control logic. When the underwater flow environment changes abruptly and the water flow impact force exceeds the yield limit of the discharge material, the drive equipment such as winches still maintain the conventional lowering or locking state, and cannot achieve active force release.

[0008] This lack of a controlled yielding mechanism in operation means that the hoisting process lacks a safety margin when facing sudden changes in the water flow environment, leading to equipment overload or damage to the vent. Summary of the Invention

[0009] To address the shortcomings of existing technologies, this invention provides an intelligent optimization method for the hoisting path of a hinged submerged raft unit. This method solves the problems of existing underwater submerged raft hoisting operations, which struggle to predict local stress states in complex water flow environments and are unable to actively release force in the face of sudden water flow shear forces, leading to physical tearing of the raft body and deviation from the predetermined landing trajectory.

[0010] To achieve the above objectives, the present invention provides the following technical solution: an intelligent optimization method for the hoisting path of a hinged sinker unit, comprising:

[0011] The expected total lifting cycle is divided into multiple discrete time slices of equal length, and a three-dimensional predictive environment model for the corresponding discrete time slice is constructed based on the multi-level water velocity and direction profile data fed back by the acoustic Doppler current profiler.

[0012] The predicted tension of each grid node of the hinged submerged liner is calculated by combining the hydrodynamic resistance model and the three-dimensional prediction environment model, and the flexible deformation energy state matrix corresponding to the discrete time slice is generated.

[0013] The flexible deformation energy state matrix is ​​inspected to determine the local tension peak. For discrete time slices with local tension peaks, rigid hanger tilt angle control variables are introduced and non-equivalent rate commands are assigned to generate a globally optimal equipment workflow scheduling table.

[0014] The actions are executed according to the global optimal equipment workflow scheduling table. The predicted tension sum of the corresponding slice in the flexible deformation energy state matrix is ​​extracted. The actual physical load equivalent mean calculated by nonlinear mechanical loss compensation based on the frequency converter feedback data is checked against the predicted tension sum. When the deviation exceeds the limit, the current instruction is truncated and resampling is triggered to re-plan the remaining time slice.

[0015] Preferably, the steps of dividing the expected total lifting cycle into multiple discrete time slices of equal length and constructing a three-dimensional predictive environment model for the corresponding discrete time slices based on multi-level water velocity and direction profile data fed back by an acoustic Doppler current profiler specifically include:

[0016] The expected total lifting cycle is obtained by using an anti-overflow calculation logic that divides the vertical space span by the larger of the rated lowering line speed and the lower limit of the safe line speed. The expected total lifting cycle is then divided into equal lengths according to the preset slice time resolution to generate discrete time slices.

[0017] The multi-level water velocity and direction profile data fed back by the acoustic Doppler current profiler are analyzed. The water body from the water surface to the bottom of the target bed is divided into multiple discrete water depth layers. The measured water velocity and measured water direction values ​​of each discrete water depth layer at the current sampling time are extracted. The acquisition timestamp and the corresponding discrete time slice are synchronized and aligned on the time axis to synthesize an initial velocity vector.

[0018] The historical velocity vector sequence of each discrete water depth is retrieved and concatenated with the initial velocity vector as the input sequence for time series prediction. The input sequence is then used to calculate the predicted velocity vector for each discrete water depth.

[0019] A multidimensional mapping relationship is established using the corresponding predicted flow velocity vector as the state feature value, and a three-dimensional prediction environment model corresponding to the discrete time slice is constructed based on the multidimensional mapping relationship.

[0020] In this step, the system uses historical measurement information and current environmental characteristics to extrapolate future states and establish temporal and spatial references for flow field changes.

[0021] Preferably, the steps for calculating the predicted tension of each grid node of the hinged submerged liner by combining the hydrodynamic resistance model and the three-dimensional predictive environment model specifically include:

[0022] The physical dimensions of the hinged submerged slab to be hoisted are retrieved to establish a topological constraint model. The continuous physical slab is discretized into a two-dimensional grid system consisting of rows and columns according to the inherent hinge connection characteristics, and the grid node coordinates corresponding to each physical connection block are assigned.

[0023] Based on the expected uniform descent depth, the coordinates of the grid nodes are mapped to the reference spatial coordinate system, and the spatial depth parameters of each node in different time dimensions are calculated.

[0024] Interpolation addressing is performed in the three-dimensional prediction environment model based on spatial depth parameters, the predicted velocity vector at the corresponding depth is extracted, and the magnitude of the predicted velocity vector is taken as the predicted velocity scalar.

[0025] Physical parameters related to the hydrodynamic resistance model are extracted, and the predicted tension of each grid node of the hinged submerged liner is calculated by combining the hydrodynamic resistance model with the predicted flow velocity scalar and using the node tension iterative derivation formula.

[0026] The nodal tension iterative derivation formula in this step integrates the physical gravity of a single node, the buoyancy of water, and the hydrodynamic resistance term determined based on the upstream projected area and the flow resistance coefficient. By accumulating the load by transferring it upward along the longitudinal topology of adjacent nodes, the nodal force values ​​under fluid-structure interaction are obtained.

[0027] Preferably, the step of generating the flexible deformation energy state matrix corresponding to the discrete time slice specifically includes: using the row and column identifiers of the two-dimensional grid system as two-dimensional indices, arranging and reorganizing the predicted tensions of all grid nodes calculated under the same discrete time slice in an array, generating a flexible deformation energy state matrix that reflects the stress mapping distribution under underwater shear flow impact in the form of a digital array and corresponds to the discrete time slice.

[0028] Preferably, the steps for inspecting and judging local tension peaks in the flexible deformation energy state matrix specifically include: retrieving a preset safety tension threshold, performing a preliminary comparison between the predicted tension of each grid node in the flexible deformation energy state matrix and the safety tension threshold, and filtering out grid nodes whose predicted tension is greater than the safety tension threshold; introducing a connected component statistical algorithm to evaluate the number of spatial clusters adjacent to grid nodes whose predicted tension is greater than the safety tension threshold to form a spatial clustering degree; if the spatial clustering degree exceeds a preset spatial clustering degree tolerance parameter, the corresponding region is marked to determine the local tension peak; wherein, the safety tension threshold is determined based on the material yield strength of the hinge sinker physical connection component multiplied by a safety reduction factor.

[0029] Preferably, the steps of introducing a rigid hanger tilt angle control variable and assigning non-symmetrical rate commands to discrete-time slices with local tension peaks specifically include:

[0030] For discrete time slices with local tension peaks, a rigid hanger tilt angle control variable is introduced to break the rigid posture of horizontal suspension and form a relative water-facing angle between the sinking plane and the water flow impact direction;

[0031] Trigger an iterative program for shear force dispersion based on changes in the angle of attack, and use the tension peak reduction compensation formula to calculate the non-equivalent rate bias for the slice with local tension peaks.

[0032] The left and right auxiliary winches are independently decoupled based on the non-equivalent rate bias. The expected rate of the left auxiliary winch is set as the rated lowering line speed minus the non-equivalent rate bias, and the expected rate of the right auxiliary winch is set as the rated lowering line speed plus the non-equivalent rate bias, thereby allocating non-equivalent rate commands.

[0033] The tension peak reduction compensation formula used in this step utilizes the product relationship between the fluid-structure interaction adjustment coefficient and the overload ratio to establish the correspondence between the overload stress and the difference in the lowering rate control. By changing the angle of attack to reduce the projected area, the active dispersion of local flow resistance is achieved.

[0034] Preferably, the steps for generating the globally optimal device workflow scheduling table specifically include:

[0035] The cumulative lateral slip distance generated by each discrete time slice is physically integrated on the time axis. When the cumulative lateral slip distance reaches the set allowable value of spatial trajectory deviation, the rigid hanger tilt angle control variable is adjusted back to restore the horizontal attitude and the overall lowering rate of the main winch is reduced simultaneously, generating an asymmetric adjustment command under spatial trajectory constraints.

[0036] By combining conventional symmetrical release commands within the safety tension threshold without peak values ​​with asymmetric adjustment commands under spatial trajectory constraints, a globally optimal device workflow scheduling table covering the entire lifecycle is compiled and generated.

[0037] This step adjusts the posture to meet the requirements for force safety and bed positioning accuracy.

[0038] Preferably, the steps of extracting the sum of predicted tensions of the corresponding slices in the flexible deformation energy state matrix and calculating the equivalent mean of the actual physical load from the frequency converter feedback data through nonlinear mechanical loss compensation specifically include:

[0039] The variable frequency drive feedback data, including the armature current and the linear speed of the main winch drive motor during operation, is obtained. A dynamic electromechanical conversion function is introduced into the variable frequency drive feedback data to perform nonlinear mechanical loss compensation. The equivalent average value of the actual physical load is calculated using the actual physical load conversion formula.

[0040] The predicted tension of the top-level grid node in the flexible deformation energy state matrix is ​​summed laterally to extract the total predicted tension for subsequent benchmarking with the equivalent mean of the actual physical load.

[0041] The above formula for converting actual physical load introduces the equivalent conversion constant of the motor and eliminates the nonlinear mechanical friction consumption factor that varies with speed, thereby calculating the effective load borne by the lifting system.

[0042] Preferably, the step of verifying the deviation between the equivalent mean of the actual physical load and the sum of the predicted tensions specifically includes: calculating the difference between the calculated equivalent mean of the actual physical load and the extracted sum of predicted tensions to perform deviation verification and obtain the real-time dynamic deviation; determining whether the absolute value of the real-time dynamic deviation exceeds the set load verification deviation tolerance, and starting an internal timer to monitor the duration of the over-limit state when it exceeds the tolerance; when the duration continuously exceeds the preset filtering time window, confirming that the current actual stress state substantially deviates from the predicted environmental benchmark, and determining that the deviation exceeds the limit.

[0043] Preferably, the step of truncating the current instruction and triggering resampling to replan the remaining time slice when it is determined that the deviation exceeds the limit specifically includes:

[0044] When the deviation is determined to be excessive, an interrupt instruction is immediately sent to the programmable logic controller to truncate the current instruction and trigger the follow-up discharge mechanism.

[0045] The control hoisting mechanism releases the brake lock-up state through the follow-up force relief mechanism, allowing the drive motor to perform reverse retreating action under the fluid impact tension exceeding the safety tension threshold, and relocks the hoisting mechanism brake when the accumulated reverse retreating distance reaches the maximum retreating stroke limit to complete the avoidance action;

[0046] After completing the avoidance maneuver, a high-frequency wake-up signal is sent to the underwater sensing system to trigger the acoustic Doppler current profiler to resample and collect the latest environmental data of the current water area.

[0047] The three-dimensional predictive environment model is reconstructed using the latest environmental data as initial input, and the remaining time slices are replanned using the new environmental feature matrix.

[0048] This invention provides an intelligent optimization method for the hoisting path of a hinged recessed unit. It has the following beneficial effects:

[0049] 1. This invention divides the total hoisting cycle into discrete time slices, constructs a three-dimensional predictive environment model by combining water flow profile data, generates a flexible deformation energy state matrix based on the hydrodynamic resistance model, and calculates the predicted tension of each grid node. On this basis, the local tension peak is inspected, and a rigid hanger tilt angle control variable is introduced for slices with peak values ​​to adjust the water-facing angle of the raft, thereby actively dispersing the impact force of underwater shear flow. This solves the problem of difficulty in predicting the local stress state in complex water flow environments, which leads to physical tearing of the raft body.

[0050] 2. In the process of generating the globally optimal equipment workflow scheduling table, this invention performs physical integration monitoring on the cumulative lateral sliding distance generated by each time slice. When the sliding distance reaches the allowable value of spatial trajectory deviation, the horizontal attitude is restored by adjusting the tilt angle control variable and the lowering speed of the main winch is reduced simultaneously. This adjustment path, which integrates force distribution and spatial displacement, ensures that the submerged raft can overcome the water flow thrust and complete the landing and positioning according to the predetermined trajectory while changing its attitude to avoid the impact of the water flow.

[0051] 3. This invention calculates the equivalent average value of the actual physical load by extracting feedback data from the frequency converter driver, and performs dynamic deviation verification between it and the predicted tension sum in the energy state matrix. When the deviation exceeds the limit, the command is cut off and the follow-up force relief mechanism is triggered, so that the hoisting mechanism will reverse and yield in a controlled manner when the water flow impact force exceeds the limit. This mechanism provides active leakage and avoidance at the mechanical level when facing sudden water flow shear force, and after the yielding is completed, the remaining hoisting steps are planned by re-collecting environmental data, which improves the safety margin of the overall operation in dealing with unknown water flow environment. Attached Figure Description

[0052] Figure 1This is a system operating environment architecture diagram of an intelligent optimization method for the hoisting path of a hinged sinker unit according to an embodiment of the present invention;

[0053] Figure 2 This is a flowchart of an intelligent optimization method for the hoisting path of a hinged sinker unit according to the present invention;

[0054] Figure 3 This is a comparison diagram of the local tension peak changes in the exhaust body of the present invention;

[0055] Figure 4 This is a comparison diagram of the lateral offset of the hoisting trajectory according to the present invention. Detailed Implementation

[0056] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] See attached document Figure 1 This invention provides an intelligent optimization method for the hoisting path of a hinged submerged unit. The execution of this method depends on the system hardware architecture. The system hardware architecture includes an edge computing server, an underwater sensing system, and a hoisting execution system.

[0058] The underwater sensing system, including an acoustic Doppler current profiler, is deployed in the construction area. The acoustic Doppler current profiler's probe is oriented towards the underwater work area to acquire multi-level profile data of water velocity and direction from the water surface to the bottom of the target bed.

[0059] The lifting execution system is installed on a surface crane vessel and includes a main winch, a left auxiliary winch, a right auxiliary winch, a frequency converter, and a programmable logic controller (PLC). Each of the main winch, left auxiliary winch, and right auxiliary winch is equipped with a drive motor. The frequency converter is electrically connected to each drive motor. The PLC is communicatively connected to the frequency converter. The edge computing server is equipped with a data processing unit and a storage unit. The edge computing server is communicatively connected to the PLC.

[0060] The underwater sensing system transmits data packets containing depth, current velocity, and direction to the edge computing server via industrial Ethernet at a preset sampling frequency. The variable frequency drive collects the armature current of the main winch drive motor and the linear speed of the main winch in real time, and feeds this data back to the edge computing server at fixed time intervals. The edge computing server then issues equipment workflow scheduling instructions to the programmable logic controller.

[0061] See attached document Figure 2This invention provides an intelligent optimization method for the hoisting path of a hinged recessed unit, comprising the following steps:

[0062] S1, Time Resource Slice Initialization and Environment Reconstruction: The edge computing server divides the expected total hoisting cycle into multiple discrete time slices of equal length, and constructs a three-dimensional predictive environment model corresponding to each time slice based on the multi-level water velocity and water direction profile data fed back by the acoustic Doppler current profiler.

[0063] S2, Forward extrapolation of flexible deformation energy state: The edge computing server calculates the predicted tension of each grid node of the hinged submerged liner in each time slice by combining the hydrodynamic resistance model and the three-dimensional prediction environment model, and generates the flexible deformation energy state matrix corresponding to each time slice.

[0064] S3, hoisting workflow scheduling and planning: The edge computing server inspects the flexible deformation energy state matrix to determine the local tension peak. For time slices with tension peaks, rigid hanger tilt angle control variables are introduced and non-equivalent rate instructions are assigned to generate a globally optimal equipment workflow scheduling table.

[0065] S4, Low-level feedback verification and dynamic reset: The lifting execution system executes actions according to the globally optimal equipment workflow scheduling table. The edge computing server extracts the predicted tension sum of the corresponding time slice based on the flexible deformation energy state matrix. The edge computing server extracts the armature current and running linear speed fed back by the frequency converter driver to perform nonlinear mechanical loss compensation calculation to obtain the equivalent mean of the actual physical load. The edge computing server performs deviation verification between the equivalent mean of the actual physical load and the predicted tension sum, and truncates the current instruction to trigger resampling to re-plan the remaining time slice when the deviation exceeds the limit.

[0066] The following section will elaborate on the specific implementation process of each step in the intelligent optimization method for hoisting path of a hinged sinking unit provided by this invention, taking into account the system operating environment.

[0067] In this embodiment, S1, time resource slice initialization and environment reconstruction: The edge computing server divides the expected total hoisting cycle into multiple discrete time slices of equal length, and constructs a three-dimensional predictive environment model corresponding to each time slice based on multi-level water velocity and direction profile data fed back by an acoustic Doppler current profiler. Specifically, to transform the continuous physical lowering process into a discrete computational state and establish environmental prior conditions, the following sub-steps are included:

[0068] S11. To plan the global equipment action scheduling sequence, a time-dimensional evaluation benchmark needs to be established in advance. The edge computing server obtains the vertical spatial span from the hinge sinking point to the bottom of the target bed, and simultaneously obtains the rated lowering linear speed of the main winch in the lifting execution system. In actual engineering waters, to avoid overflow of the computing platform caused by the denominator approaching zero due to extremely low-speed operation or short-term shutdown of the lifting equipment, the system is configured with a safety linear speed lower limit. This safety linear speed lower limit is calibrated based on the lowest stable operating frequency of the main winch drive motor, and its value range is usually set to 0.5 m / min to 1.0 m / min. The edge computing server uses an overflow prevention calculation logic that divides the vertical spatial span by the larger of the rated lowering linear speed and the safety linear speed lower limit to obtain the expected total lifting cycle.

[0069] After determining the estimated total lifting cycle, the edge computing server extracts the preset slice time resolution from its internal storage unit. This slice time resolution is determined comprehensively based on the evolution rate of the flow field in the working water area and the communication execution cycle of the programmable logic controller, typically ranging from 5 to 15 seconds in engineering projects. The edge computing server divides the estimated total lifting cycle into equal-length slices according to the slice time resolution, generating a time resource sequence composed of multiple consecutive discrete time slices. Each discrete time slice serves as the basic time scale for the subsequent allocation of lifting equipment workflow instructions. The discretization operation in the time domain can be implemented by those skilled in the art based on conventional digital signal sampling and time-division multiplexing principles, and will not be elaborated upon here.

[0070] S12. After obtaining the time reference, it is necessary to further synchronize and collect hydrological environmental information in the spatial dimension. The acoustic Doppler current profiler in the underwater sensing system dynamically collects hydrological echo signals from the construction area and converts them into data packets containing acoustic frequency shift and depth mapping relationships. The edge computing server receives these data packets via industrial Ethernet and performs protocol decoding. To match the gridded spatial scale of the hoisting operation, the edge computing server divides the water body from the surface to the bottom of the target landing bed into multiple discrete water depth layers according to a preset water depth stratification step size. The water depth stratification step size here depends on the longitudinal detection resolution of the acoustic Doppler current profiler and is typically set between 0.5 meters and 2.0 meters.

[0071] Subsequently, the edge computing server parses and decodes the data packets, extracting the measured water velocity and direction values ​​for each discrete water depth at the current sampling time. Considering the time difference between the sensor's physical sampling delay and the control center, the edge computing server synchronizes the acquisition timestamp of the hydrological echo signal with the corresponding discrete time slice, and synthesizes the measured water velocity and direction values ​​for the same discrete water depth into an initial velocity vector. As a preliminary environmental perception method, the multi-level water velocity and direction profile data generated in this step provides the current physical state input for subsequent environmental reconstruction.

[0072] S13. Relying solely on the physical state of the current cross-section is insufficient to address the spatiotemporal evolution of complex layered flow field environments. Therefore, forward state extrapolation based on historical trends is necessary. The edge computing server retrieves historical velocity vector sequences for each discrete water depth from the storage unit, concatenates them end-to-end with the initial velocity vector at the current sampling moment in the time dimension, and fuses them as the input sequence for time-series prediction.

[0073] This embodiment employs a Long Short-Term Memory (LSTM) network algorithm to construct a time-series prediction model. The model's internal network topology includes an input layer, hidden layers, and an output layer. The feature node dimensions of the input layer correspond to the number of discrete water depths. The hidden layer utilizes internal forget gates and input gate structures to regulate the retention and dynamic updating of historical hydrological information. The output layer maps to the predicted flow velocity features in the future time domain. During the model construction and training phases, acoustic Doppler detection datasets from past historical cycles of the construction operation area are extracted as training samples. The measured flow velocity at corresponding sequential times is used as the true label. Mean squared error is used as the loss function, and the network node weights are iteratively updated using the Adam optimizer combined with the backpropagation algorithm. Based on the pre-trained forward computation logic, the edge computing server inputs the aforementioned time-series prediction input sequence into this time-series prediction model to calculate the predicted flow velocity vectors corresponding to each discrete water depth within each discrete time slice.

[0074] Ultimately, the edge computing server establishes a multidimensional mapping relationship using each discrete time slice as the time axis coordinate, each discrete water depth layer as the spatial depth coordinate, and the corresponding predicted flow velocity vector as the state feature value. Based on this multidimensional mapping relationship, a three-dimensional prediction environment model covering the entire expected hoisting cycle is constructed.

[0075] In this embodiment, S2, forward extrapolation of flexible deformation energy state: The edge computing server calculates the predicted tension of each grid node of the hinged submerged structure within each discrete time slice by combining the hydrodynamic resistance model and the three-dimensional prediction environment model, generating the flexible deformation energy state matrix corresponding to each discrete time slice. Based on the conventional physical principles of multibody dynamics and fluid-structure interaction, in order to quantify the influence of complex flow fields on the local stress distribution of the flexible working entity, it is necessary to convert the continuous physical force into two-dimensional array data. This process includes the following sub-steps:

[0076] S21. Establishing a discretized force analysis platform is a prerequisite for hydrodynamic deduction. Considering that the nonlinear deformation of continuous media in complex flow fields is difficult to solve directly analytically, the edge computing server retrieves the preset physical dimension data of the hinged submerged raft to be lifted from its internal storage unit to establish a topological constraint model. In this topological constraint model, the continuous physical raft is discretized into a two-dimensional grid system consisting of m rows and n columns according to its inherent hinge connection characteristics, and each physical connection block is assigned a corresponding grid node coordinate. Based on the expected uniform lowering depth of the lifting execution system in each discrete time slice, the edge computing server maps the above grid node coordinates to a reference spatial coordinate system and calculates the spatial depth parameters of each node in different time dimensions. For the mapping and transformation of spatial coordinates, those skilled in the art can use conventional affine transformations and coordinate system translation and rotation algorithms, which are well-known technologies in the field and will not be elaborated here.

[0077] S22, after obtaining the node spatial depth reference, it is necessary to convert the flow field environmental disturbance into the mechanical stress borne by the grid nodes. For any given discrete-time slice, the edge computing server performs interpolation addressing in the three-dimensional prediction environment model based on the aforementioned spatial depth parameters, extracts the predicted flow velocity vector at the corresponding depth, and takes its amplitude as the predicted flow velocity scalar. Simultaneously, the edge computing server pre-retrieves the stored density and single-piece volume parameters of the submerged aqueduct material to calculate the inherent load basis. Combining the characteristics of the water medium and the physical properties of the submerged aqueduct itself, the edge computing server uses a node tension iterative derivation formula to derive the predicted tension of each grid node of the hinged submerged aqueduct. The node tension iterative derivation formula is as follows:

[0078] ;

[0079] in, Indicates the first Nodes in a discrete-time slice Predicted tension; Indicates the first Nodes in a discrete-time slice The predicted tension represents the cumulative load transmitted from the adjacent nodes below along the longitudinal topology. Represents physical gravity at a single node; This represents the buoyancy of a single node. This represents the density of the water medium, and its value is dynamically determined based on the sediment content of the water in the operating area. The conventional range is 1000 kg / m³. 3 Up to 1050kg / m 3 between; This represents the fluid flow resistance coefficient, which is calibrated based on the surface roughness of the submerged material and the fluid Reynolds number, and typically ranges from 1.2 to 1.8. This represents the projected area of ​​a single node facing the water. Indicates the first Nodes in a discrete-time slice The predicted velocity scalar corresponding to the location; This represents the square operation after taking the absolute value of the predicted flow velocity scalar.

[0080] The physical design logic of this iterative deduction formula is that the upper node of the flexible submerged raft, which is in a suspended and descending state, not only bears the orthogonal vector combination of its own inherent net weight and horizontal hydrodynamic resistance, but also must bear the tensile load transmitted upward by all the lower nodes through the hinge topology.

[0081] To ensure the physical causality of dynamic load derivation, the formula introduces a multi-dimensional vector coupling of gravity, buoyancy, and hydrodynamic drag, reconstructing the actual stress transmission process under fluid-structure interaction through bottom-up, step-by-step recursion. As a preferred approach, for the bottom-level mesh nodes, i.e., those satisfying the condition... At that time, because there are no adjacent rows connected below, the edge computing server transfers its initial tension term, i.e., the first... Nodes in a discrete-time slice The predicted tension is directly set to zero for calculation, providing closed boundary conditions to ensure the logical completeness of the matrix iterative deduction process.

[0082] S23. After completing the force analysis of all single-point dimensions, a data structure needs to be constructed for subsequent global action planning and multi-dimensional state determination. Edge computing servers are identified by rows. and column identifiers For two-dimensional indexing, the predicted tensions of all grid nodes calculated under the same discrete time slice are arrayed and rearranged. Through this rearrangement operation, the edge computing server generates a flexible deformation energy state matrix corresponding to each discrete time slice. This flexible deformation energy state matrix, in the form of a quantized digital array, reflects the stress mapping distribution of the flexible submerged raft under underwater shear flow impact over the time span. This not only reflects the stress state of a single node, but also characterizes the overall deformation trend of the submerged raft through the spatial distribution characteristics and gradient changes of the array elements. Relying on this underlying data support, the system can effectively identify local anomaly areas in subsequent steps and provide an algorithmic decision basis for triggering asymmetric spatial attitude scheduling.

[0083] In this embodiment, S3, hoisting workflow scheduling planning: The edge computing server inspects the flexible deformation energy state matrix to determine local tension peaks. For time slices with tension peaks, a rigid gantry tilt angle control variable is introduced and asymmetric rate commands are assigned to generate a globally optimal equipment workflow scheduling table. Specifically, to overcome the engineering bottleneck of being unable to resolve local water flow shear forces in conventional symmetrical lowering operations, the system establishes an active force release mechanism based on the principle of spatial attitude reconstruction, including the following sub-steps:

[0084] S31. Establishing a quantitative and environmentally robust safety assessment benchmark is a prerequisite for triggering intelligent scheduling. The edge computing server performs a traversal search of the flexible deformation energy state matrix corresponding to each discrete time slice. During this data inspection process, the edge computing server retrieves a preset safety tension threshold from its internal storage unit. This safety tension threshold is determined based on the material yield strength of the hinged submerged slab's physical connection components multiplied by a safety reduction factor. Considering the uncontrollable disturbances in the underwater environment, the conventional safety reduction factor is usually set between 0.7 and 0.85.

[0085] To avoid biased judgments based solely on a single extreme value due to instantaneous noise in hydrological sensor data, the edge computing server employs multi-dimensional weighted logic for state assessment. This involves not only initially comparing the predicted tension of each grid node in the flexible deformation energy state matrix with a safe tension threshold, but also introducing a connected component statistical algorithm to evaluate the spatial clustering degree of adjacent nodes exceeding this threshold. If the predicted tension of all grid nodes in the matrix is ​​not greater than the safe tension threshold, the discrete-time slice is determined to be in a safe energy state. Conversely, if the predicted tension of a grid node exceeds the safe tension threshold, and its spatial clustering degree exceeds a preset spatial clustering degree tolerance parameter (this parameter is based on the local tearing limit of the drainage body and is typically set to 3 to 5 grid nodes), it is marked as a local tension peak, and the corresponding discrete-time slice is thus determined to be in a high-risk state.

[0086] S32, as a preferred approach, for peakless discrete-time slices determined to be in a safe energy state, the edge computing server employs a stationarity-priority control strategy, allocating conventional symmetrical lowering rate commands. Specifically, the commands instruct the main winch in the lifting system to maintain its original rated lowering linear speed, and assign synchronous rate commands to the left and right auxiliary winches, matching the speed of the main winch. This symmetrical physical lowering action maintains the horizontal posture of the rigid gantry, ensuring a smooth water entry process for the hinged submersible in a low-resistance flow field environment.

[0087] S33. If the hydrodynamic conditions of the environmental flow field change abruptly, the conventional synchronous lowering method would cause irreversible physical tearing of the raft body in certain areas. To address this situation, for discrete-time slices deemed high-risk, the edge computing server introduces a rigid hanger tilt angle control variable to forcibly disperse local shear forces. The core physical principle of this mechanism is to artificially create a relative angle of attack between the raft plane and the direction of water flow impact by breaking the rigid posture of the horizontal suspension. Changing the angle of attack effectively reduces the effective projected area of ​​the raft in the direction of hydrodynamic impact, thereby weakening the accumulation of structural tension on the raft body caused by fluid flow resistance from the root of physical stress.

[0088] S34, after introducing the rigid hanger tilt angle control variable, based on proportional control theory, it is necessary to solve for the specific control parameters for implementing this attitude intervention through closed-loop feedback. The edge computing server triggers an iterative program for shear force dispersion based on the change in the angle of attack, and uses the tension peak reduction compensation formula to calculate the non-uniform rate bias for the current high-risk slice. The tension peak reduction compensation formula is as follows:

[0089] ;

[0090] in, Indicates the first Non-equivalent rate bias under discrete time slices; This represents the fluid-structure interaction adjustment coefficient, the value of which is determined by a combination of the Reynolds number of the flow field environment and the moment of inertia of the hanger itself, and the conventional value range is 0.1 to 0.3; Indicates the first Local tension peaks under discrete time slices; This represents the safety tension threshold. Because it characterizes the inherent positive yield physical property of the material, its value in actual engineering is a constant that is always greater than zero. This avoids the singularity risk of calculation overflow caused by the denominator approaching zero in the division operation at the algorithm level. This indicates the rated lowering linear speed of the main winch.

[0091] The physical purpose of this tension peak reduction compensation formula is to establish a quantitative compensation relationship between the over-limit stress and the difference in mechanical control rate; that is, the greater the proportion of over-limit stress, the greater the rate deviation correction amount needs to be applied. Subsequently, the edge computing server independently decouples the left and right auxiliary devices based on the asymmetric rate bias. The expected rate of the left auxiliary winch is set as the rated lowering line speed minus the asymmetric rate bias, and the expected rate of the right auxiliary winch is set as the rated lowering line speed plus the asymmetric rate bias. After establishing the tilt attitude with the new asymmetric rate command, the edge computing server recalculates the mesh node tension under this attitude. If the calculated new maximum tension is still greater than the safe tension threshold, the asymmetric rate bias is increased in fixed steps for iterative looping until the force safety condition is met.

[0092] S35. After the allocation and iteration of the aforementioned asymmetric rate commands, although the local stress risk was mitigated within the time slice, the continuous tilting posture could cause the submerged raft to slip laterally under the thrust of the water flow, increasing the risk of deviating from the target landing area. To address this, the edge computing server sets a spatial trajectory deviation tolerance value (this value is based on the construction tolerance margin at the bottom of the landing area, typically between 0.5 meters and 1.5 meters), and performs physical integration on the cumulative lateral slip distance generated in each slice on the time axis. When the cumulative lateral slip distance approaches the spatial trajectory deviation tolerance value, the edge computing server forcibly reverts the rigid hanger tilt angle control variable to restore the horizontal attitude, simultaneously reducing the overall lowering rate of the main winch. By delaying the lowering process, it gains time for spatial trajectory correction, generating asymmetric adjustment commands under spatial trajectory constraints. Finally, by combining the conventional symmetric lowering commands under a safe energy state without peak values ​​and the asymmetric adjustment commands under spatial trajectory constraints, the edge computing server compiles and generates a globally optimal equipment workflow scheduling table covering the entire lifecycle. The globally optimal device workflow scheduling table is encapsulated and distributed in an industrial communication protocol format that can be parsed by the programmable logic controller, providing complete data-driven input for the actual execution of subsequent underlying device actions.

[0093] In this embodiment, S4, bottom-level feedback verification and dynamic reset: The lifting execution system executes actions according to the globally optimal equipment workflow scheduling table; the edge computing server extracts the predicted tension sum of the corresponding time slice based on the flexible deformation energy state matrix; the edge computing server extracts the armature current and operating linear velocity fed back by the frequency converter driver to perform nonlinear mechanical loss compensation calculation to obtain the equivalent mean of the actual physical load; the edge computing server verifies the deviation between the equivalent mean of the actual physical load and the predicted tension sum, and truncates the current instruction to trigger resampling to re-plan the remaining time slice when the deviation exceeds the limit. Specifically, to form a control closed loop and prevent the environmental prediction model from decoupling from the actual physical state, the following sub-steps are included:

[0094] S41, after receiving the globally optimal equipment workflow scheduling table, the programmable logic controller (PLC) parses it into underlying control signals and drives each winch of the hoisting execution system to perform the corresponding lowering action. During physical equipment operation, obtaining real load feedback is the foundation for forming closed-loop verification. The variable frequency drive (VFD) dynamically collects the armature current of the main winch drive motor and the linear speed of the main winch during operation. Considering the timing synchronization logic of multi-source data, this VFD feedback data, including the armature current of the main winch drive motor and the linear speed of the main winch, is fed back to the edge computing server via the industrial bus at fixed time intervals with timestamps, serving as the underlying hardware parameter support for evaluating the actual physical load. To balance the response sensitivity of the control system with the communication bandwidth constraints of the industrial Ethernet, this fixed time interval is typically set to a communication cycle of 10 to 50 milliseconds.

[0095] S42. To align with the actual physical state fed back from the underlying layer, a theoretical benchmark needs to be extracted from the previous deduction results. For the currently executing discrete-time slice, the edge computing server retrieves the flexible deformation energy state matrix corresponding to the time generated in the preceding steps. Based on the principle of static equilibrium, since the main winch bears the vertical lowering load of the entire sump, the edge computing server performs a lateral summation operation on the predicted tension of the top-level grid nodes in this matrix (i.e., all column nodes representing the directly stressed elements at the sump head, with the row identifier being the first row), extracting the total predicted tension for the corresponding discrete-time slice. This sum represents the macroscopic dynamic load in the sling direction that the lifting equipment should bear under the theoretical flow field environment, providing a theoretical benchmark for subsequent deviation verification.

[0096] S43. Considering the inherent friction and energy dissipation in mechanical transmission systems, directly equating the motor current to an external load would result in significant evaluation bias. Based on the principle of electromechanical energy conversion conservation, the edge computing server introduces a dynamic electromechanical conversion function to compensate for nonlinear mechanical losses in the extracted armature current and operating linear velocity, calculating the equivalent average value of the actual physical load. The edge computing server uses the actual physical load conversion formula to calculate the equivalent average value of the actual physical load. The actual physical load conversion formula is as follows:

[0097] ;

[0098] in, Indicates the first The equivalent mean of the actual physical load within a discrete time slice; Indicates the first Start time of each discrete time slice Until the end time Definite integral operations within; It represents the equivalent conversion constant of the drive motor, which is determined by the inherent factory parameters of the lifting execution system. Its physical function is to linearly convert the electromagnetic torque into the tension in the direction of the sling through the drum radius mapping relationship, so as to unify the energy calculation dimensions. express The armature current fed back by the frequency converter driver at all times; Indicates corresponding to The dynamic mechanical friction coefficient of the constant operating linear velocity is physically characterized as a nonlinear friction curve that varies with speed. This curve data is obtained in advance through the no-load calibration stage of the equipment. express The linear velocity of operation is constantly being fed back; This represents the equivalent gravity of the transmission components within the lifting machinery system itself.

[0099] The physical meaning of this actual physical load conversion formula is that, since the current operation is a lowering operation, the mechanical friction force and the motor pulling force jointly resist the external load. Therefore, by superimposing the mechanical friction loss compensation that changes dynamically with speed into the electromagnetic driving force, the actual effective load applied by the sinking raft to the lifting equipment can be restored.

[0100] This compensation calculation eliminates current fluctuations caused by system no-load wear, thereby eliminating the interference of the equipment's own transmission noise on environmental stress verification.

[0101] S44. After acquiring the data from both the theoretical prediction and the actual extraction, closed-loop tolerance verification is crucial for assessing the safety status. To avoid biased judgments based solely on a single extreme value due to instantaneous pulses or mechanical oscillations in the underlying electrical control data, the edge computing server employs multi-dimensional sliding verification logic for status evaluation. The edge computing server calculates the difference between the calculated equivalent average of the actual physical load and the extracted sum of predicted tensions to obtain the real-time dynamic deviation between the two.

[0102] Considering the latency error of hydrological sensor data and the inherent deviation caused by model simplification, the system internally sets a load verification deviation tolerance. This tolerance is determined based on a comprehensive evaluation of the statistical variance of the total weight of the submerged pontoon and historical operating conditions, and is typically set between 10% and 15% of the total predicted tension. The edge computing server not only determines whether the absolute value of this real-time dynamic deviation exceeds the set load verification deviation tolerance, but also starts an internal timer to monitor the duration of the over-limit state. Only when the over-limit state continues for more than a preset filtering time window (usually 3 to 5 communication cycles) can the system confirm that the current actual stress state has substantially deviated from the originally set predicted environmental benchmark, and determine that the deviation has exceeded the limit.

[0103] S45, when the actual environmental flow field experiences sudden surges or abnormal undercurrents far exceeding the prediction model's assessment, the absolute value of the aforementioned real-time dynamic deviation will continuously exceed the load verification deviation tolerance. As a preferred approach, in the face of such deviation exceeding limits, to prevent equipment overload or physical tearing of the submersible, the edge computing server immediately sends an interrupt command to the programmable logic controller, forcibly truncating the currently executing drop-down action command.

[0104] Simultaneously, a follow-up force relief mechanism is triggered, which controls the hoisting mechanism of the lifting execution system to release the brake from its locked state, allowing the drive motor to reverse and yield under the fluid impact force exceeding the safe tension threshold. This follow-up force relief mechanism dissipates the destructive energy generated by the fluid through active yielding at the physical level, avoiding the breakage of the lifting point caused by rigid resistance. To ensure the controllability of the yielding process, the edge computing server is internally set with a maximum yielding travel limit. When the cumulative reverse yielding distance reaches this maximum yielding travel limit, a command is resent to lock the hoisting mechanism brake, preventing the sump from completely losing control and falling, thus completing the emergency avoidance action.

[0105] S46. After completing the emergency evacuation maneuver, the original prediction model could no longer accurately reflect the current hydrological environment evolution trend. Based on the above-mentioned deviation exceeding the limit, the edge computing server sent a high-frequency wake-up signal to the underwater sensing system, triggering the acoustic Doppler current profiler to re-acquire the real hydrological echo signal of the current water area at a higher sampling frequency.

[0106] Subsequently, the edge computing server, using the latest environmental data as initial input, re-executes the time resource slice initialization and environmental reconstruction steps, as well as the subsequent flexible deformation energy state forward deduction steps. Utilizing the new environmental feature matrix, it re-plans the closed-loop scheduling table for the remaining time slices within the entire expected hoisting cycle. Through a deep integration of underlying feedback, emergency avoidance, and dynamic reconstruction, the system constructs a closed-loop control link with adaptive error correction capabilities, ensuring the safe operation of the flexible submerged conveyor in complex and variable water flow environments.

[0107] Specific application examples:

[0108] To further clarify the technical approach of this invention, a specific application embodiment of hinged submerged raft hoisting based on inland waterway improvement projects is provided. The operating waterway in this embodiment exhibits typical vertical velocity shear phenomena, with the bottom flow velocity displaying significant nonlinear characteristics due to topographic influence. Under these conditions, the system hardware architecture and control logic provided by this invention are fully deployed.

[0109] In the early stages of the project, the edge computing server determined that the vertical distance from the hinge sinking point to the bottom of the target bed was 20 meters. Simultaneously, it confirmed that the rated lowering linear speed of the main winch in the lifting system was 2.0 meters per minute. The system's set safety linear speed lower limit was 0.5 meters per minute.

[0110] Based on the overflow prevention calculation logic, the edge computing server divides 20 meters by 2.0 meters / minute to obtain the estimated total hoisting cycle of 10 minutes, or 600 seconds.

[0111] Subsequently, the edge computing server extracts a preset 10-second time resolution from its internal storage unit and divides the estimated total hoisting cycle of 600 seconds into 60 discrete time slices of equal length.

[0112] After the time dimension benchmark is established, the acoustic Doppler current profiler in the underwater sensing system begins to continuously feed data packets containing the acoustic frequency shift and depth mapping relationship back to the edge computing server at a longitudinal detection resolution of 0.5 meters.

[0113] The edge computing server parses the data packets into corresponding measured water flow velocity and measured water flow direction values. Through forward extrapolation of the time-series prediction model, a three-dimensional prediction environment model covering the entire expected hoisting cycle is constructed.

[0114] With the discretization of time resources and the reconstruction of the environmental model completed, the system begins to perform force analysis on each grid node. After establishing the two-dimensional grid system, the edge computing server locks a specific grid node at the bottom layer of the submerged surface for computation. It is known that the single-node physical weight of this grid node is 500N, the single-node buoyancy is 200N, and the density of the water medium is taken as 1000kg / m³ based on the sediment content. 3 The fluid flow resistance coefficient is calibrated to 1.5, and the projected area of ​​a single node facing the water is 1 square meter. In the 15th discrete-time slice, the edge computing server extracts the predicted flow velocity scalar of the corresponding depth of the node from the 3D prediction environment model, which is 1.2 m / s.

[0115] Because this node is the lowest level node, the tensile load term transmitted by its adjacent nodes below it is set to zero. The edge computing server calculates the predicted tension of this node using the node tension iterative derivation formula. Substituting the specific values ​​mentioned above into this formula:

[0116] ;

[0117] After calculation, the edge computing server determined that the predicted tension of the underlying mesh node was approximately 1120.9 N. This value was then filled into the flexible deformation energy state matrix of the corresponding discrete-time slice.

[0118] During matrix generation and scheduling planning, the edge computing server iterates through the flexible deformation energy state matrix corresponding to the 15th discrete-time slice. The system's internal preset safety tension threshold is 1000N. Since the calculated 1120.9N exceeds the 1000N safety tension threshold, and the number of adjacent grid nodes exceeding the limit in this region meets the spatial clustering tolerance parameter requirements, the edge computing server determines that a local tension peak has occurred in this node region and marks the current slice as a high-risk state.

[0119] To mitigate this excessive stress, the edge computing server introduces a rigid hanger tilt angle control variable and employs a tension peak reduction compensation formula to calculate the asymmetric rate offset. Given a preset fluid-structure interaction adjustment coefficient of 0.2, the corresponding value is substituted into the formula:

[0120] ;

[0121] The calculation yielded an asymmetric rate bias of 0.04836 m / min for this slice. Based on this bias, the edge computing server independently decoupled the two auxiliary winches, setting the expected rate of the left auxiliary winch to 1.95164 m / min and the expected rate of the right auxiliary winch to 2.04836 m / min. This asymmetric instruction breaks the rigid posture of the horizontal suspension, artificially creating a water-facing angle of attack on the submerged surface at the physical level, reducing the effective water-facing projected area and fundamentally weakening the accumulation of fluid flow resistance. The relevant instructions were compiled into the globally optimal equipment workflow scheduling table and sent to the programmable logic controller for execution. During the lifting system's operation, the frequency converter driver fed back the operation of the main winch motor to the edge computing server. The system then used the actual physical load conversion formula to strip away frictional losses, completed tolerance checks, and formed a closed loop.

[0122] To assess the practical engineering effectiveness of the aforementioned control logic, this embodiment supplements the verification with long-cycle simulation experiments. Under the same hydrological environment profile input, the experiment compares the scheduling method of this invention with the traditional synchronous uniform descent method, and extracts two core indicators: the local maximum tension of the dredger and the lateral spatial trajectory deviation of the system.

[0123] See attached document Figure 3 In the diagram, the solid black line represents the preferred scheduling method of this invention, and the dashed dark gray line represents the traditional symmetrical decentralization method. From... Figure 3The data trajectory shows that in the initial stage of descent, due to the low water flow velocity, the tension values ​​of the two methods were basically the same. As the descent depth increased and the material entered a high-velocity shear layer, the traditional symmetrical descent method (represented by the dark gray dashed line), which maintained a horizontal and rigid water-facing posture throughout, experienced a rapid increase in local tension peaks, exceeding 1400N, far surpassing the material's safe yield limit and posing a significant risk of tearing. In contrast, the preferred scheduling method of this invention (represented by the black solid line) successfully triggered a spatial attitude reconstruction mechanism when the tension approached the safe threshold of 1000N. Through the asymmetrical adjustment of the left and right auxiliary devices, the maximum local force was effectively suppressed and maintained in dynamic equilibrium near the safe threshold, avoiding the generation of destructive stress.

[0124] See attached document Figure 4 In the diagram, the solid black line represents the preferred scheduling method of this invention, while the dashed dark gray line represents the traditional symmetrical lowering method. Because the traditional symmetrical lowering method (represented by the dashed dark gray line) faces loss of control under the immense lateral water flow thrust, the sprue exhibits severe drifting, ultimately resulting in a lateral offset exceeding 2.5 meters, leading to a failure in landing accuracy. In contrast, the preferred scheduling method of this invention (represented by the solid black line) not only reduces overall fluid resistance through tilt-angle force relief but also sets an allowable spatial trajectory deviation value within the system. When lateral slippage approaches the set upper limit due to the tilted attitude, the edge computing server forcibly corrects the attitude and reduces the overall lowering rate. Therefore, the solid black line is consistently and strictly controlled within a safe tolerance margin of 0.8 meters throughout the entire lowering cycle. Both experimental data jointly confirm that this scheme possesses excellent spatial landing point control capabilities while ensuring the integrity of the flexible structure.

Claims

1. A method for intelligent optimization of the hoisting path of a hinged recessed unit, characterized in that, include: The expected total lifting cycle is divided into multiple discrete time slices of equal length, and a three-dimensional predictive environment model corresponding to the discrete time slice is constructed based on the multi-level water velocity and water direction profile data fed back by the acoustic Doppler current profiler. The predicted tension of each grid node of the hinged submerged liner is calculated by combining the hydrodynamic resistance model and the three-dimensional prediction environment model, and the flexible deformation energy state matrix corresponding to the discrete time slice is generated. The steps for calculating the predicted tension of each grid node of the hinged submerged dam by combining the hydrodynamic resistance model and the three-dimensional predicted environment model specifically include: The physical dimensions of the hinged submerged slab to be hoisted are retrieved to establish a topological constraint model. The continuous physical slab is discretized into a two-dimensional grid system consisting of rows and columns according to the inherent hinge connection characteristics, and the grid node coordinates corresponding to each physical connection block are assigned. Based on the expected uniform descent depth, the coordinates of the grid nodes are mapped to the reference spatial coordinate system, and the spatial depth parameters of each node in different time dimensions are calculated. Interpolation addressing is performed in the three-dimensional prediction environment model based on the spatial depth parameters, the predicted velocity vector at the corresponding depth is extracted, and the magnitude of the predicted velocity vector is taken as the predicted velocity scalar. Physical parameters related to the hydrodynamic resistance model are extracted, and the predicted tension of each grid node of the hinged submerged liner is calculated by combining the hydrodynamic resistance model with the predicted flow velocity scalar using the node tension iterative derivation formula. The flexible deformation energy state matrix is ​​inspected to determine the local tension peak. For the discrete time slice with local tension peak, a rigid hanger tilt angle control variable is introduced and a non-equivalent rate command is assigned to generate a globally optimal equipment workflow scheduling table. The step of inspecting and determining the local tension peak value of the flexible deformation energy state matrix specifically includes: A preset safety tension threshold is retrieved, and the predicted tension of each grid node in the flexible deformation energy state matrix is ​​initially compared with the safety tension threshold to filter out grid nodes whose predicted tension is greater than the safety tension threshold. A connected component statistical algorithm is introduced to evaluate the spatial clustering degree of the number of adjacent grid nodes whose predicted tension is greater than the safety tension threshold. If the spatial aggregation degree exceeds the preset spatial aggregation degree tolerance parameter, the corresponding region will be marked to determine the local tension peak. The safety tension threshold is determined by multiplying the material yield strength of the hinge sinker physical connection component by a safety reduction factor. The actions are executed according to the global optimal device workflow scheduling table. The predicted tension sum of the corresponding slice in the flexible deformation energy state matrix is ​​extracted. The actual physical load equivalent mean calculated by nonlinear mechanical loss compensation of the frequency converter feedback data is checked against the predicted tension sum. When the deviation exceeds the limit, the current instruction is truncated and resampling is triggered to re-plan the remaining time slice.

2. The intelligent optimization method for the hoisting path of a hinged recessed unit according to claim 1, characterized in that, The steps of dividing the expected total lifting cycle into multiple discrete time slices of equal length and constructing the three-dimensional predictive environment model corresponding to the discrete time slices based on multi-level water velocity and direction profile data fed back by an acoustic Doppler current profiler specifically include: The expected total lifting cycle is obtained by using an anti-overflow calculation logic that divides the vertical space span by the larger of the rated lowering line speed and the lower limit of the safe line speed. The expected total lifting cycle is then divided into equal lengths according to the preset slice time resolution to generate discrete time slices. The multi-level water velocity and direction profile data fed back by the acoustic Doppler current profiler are analyzed. The water body from the water surface to the bottom of the target bed is divided into multiple discrete water depth layers. The measured water velocity and measured water direction values ​​of each discrete water depth layer at the current sampling time are extracted. The acquisition timestamp and the corresponding discrete time slice are synchronized and aligned on the time axis to synthesize an initial velocity vector. The historical velocity vector sequence of each discrete water depth is retrieved and concatenated with the initial velocity vector as the input sequence for time series prediction. The input sequence is then used to calculate the predicted velocity vector for each discrete water depth. A multidimensional mapping relationship is established using the corresponding predicted flow velocity vector as the state feature value, and a three-dimensional prediction environment model corresponding to the discrete time slice is constructed based on the multidimensional mapping relationship.

3. The intelligent optimization method for the hoisting path of a hinged recessed unit according to claim 1, characterized in that, The step of generating the flexible deformation energy state matrix corresponding to the discrete-time slice specifically includes: Using the row and column identifiers of the two-dimensional grid system as two-dimensional indices, the predicted tensions of all grid nodes calculated under the same discrete time slice are arrayed and rearranged to generate a flexible deformation energy state matrix that reflects the stress mapping distribution under underwater shear flow impact in the form of a digital array and corresponds to the discrete time slice.

4. The intelligent optimization method for the hoisting path of a hinged sinker unit according to claim 1, characterized in that, The steps of introducing a rigid hanger tilt angle control variable and assigning asymmetrical rate commands to the discrete-time slice with local tension peaks specifically include: For the discrete time slice where the local tension peak exists, a rigid hanger tilt angle control variable is introduced to break the rigid posture of the horizontal suspension and artificially create a relative water-facing angle between the sinking plane and the water flow impact direction. Trigger a shear force dispersion iteration program based on the change in the angle of attack, and use the tension peak reduction compensation formula to calculate the non-equivalent rate bias for the current high-risk slice; The left auxiliary winch and the right auxiliary winch are independently decoupled based on the non-equivalent rate bias. The expected rate of the left auxiliary winch is set to the rated lowering line speed minus the non-equivalent rate bias, and the expected rate of the right auxiliary winch is set to the rated lowering line speed plus the non-equivalent rate bias, thereby allocating non-equivalent rate commands.

5. The intelligent optimization method for the hoisting path of a hinged recessed unit according to claim 4, characterized in that, The steps for generating the globally optimal device workflow scheduling table specifically include: The cumulative lateral slip distance generated by each discrete time slice is physically integrated on the time axis. When the cumulative lateral slip distance approaches the set allowable value of spatial trajectory deviation, the rigid hanger tilt angle control variable is forcibly reverted to restore the horizontal attitude and simultaneously reduce the overall lowering rate of the main winch, generating an asymmetric adjustment command under spatial trajectory constraints. By combining the conventional symmetrical decommissioning instructions under the safe energy state with no peak value and the asymmetric adjustment instructions under the spatial trajectory constraints, a globally optimal device workflow scheduling table covering the entire lifecycle is compiled and generated.

6. The intelligent optimization method for the hoisting path of a hinged recessed unit according to claim 4, characterized in that, The steps of extracting the sum of predicted tensions of the corresponding slices in the flexible deformation energy state matrix and calculating the equivalent mean of the actual physical load from the frequency converter feedback data through nonlinear mechanical loss compensation specifically include: The variable frequency drive feedback data, including the armature current and the linear speed of the main winch during the operation of the main winch drive motor, is obtained. The dynamic electromechanical conversion function is introduced into the variable frequency drive feedback data to perform nonlinear mechanical loss compensation. The equivalent average value of the actual physical load is calculated using the actual physical load conversion formula. The predicted tension of the top-level grid node in the flexible deformation energy state matrix is ​​summed laterally to extract the total predicted tension for subsequent comparison with the equivalent mean of the actual physical load.

7. The intelligent optimization method for the hoisting path of a hinged recessed unit according to claim 6, characterized in that, The step of verifying the deviation between the equivalent mean of the actual physical load and the sum of the predicted tensions specifically includes: The difference between the calculated equivalent mean of the actual physical load and the extracted sum of the predicted tensions is used to perform deviation verification and obtain the real-time dynamic deviation. Determine whether the absolute value of the real-time dynamic deviation exceeds the set load verification deviation tolerance, and start an internal timer to monitor the duration of the over-limit state if it exceeds the tolerance. When the duration continuously exceeds the preset filtering time window, it is confirmed that the current actual stress state substantially deviates from the predicted environmental benchmark, and the deviation is determined to be excessive.

8. The intelligent optimization method for the hoisting path of a hinged recessed unit according to claim 7, characterized in that, The step of truncating the current instruction and triggering resampling to replan the remaining time slice when the deviation exceeds the limit specifically includes: When the deviation exceeds the limit, an interrupt command is immediately sent to the programmable logic controller to truncate the current instruction and trigger the follow-up force relief mechanism; The control hoisting mechanism releases the brake lock-up state through the follow-up force relief mechanism, allowing the drive motor to reverse and retract under the fluid impact force exceeding the safety tension threshold, and relocks the hoisting mechanism brake when the accumulated reversal and retraction distance reaches the maximum retraction stroke limit to complete the emergency avoidance action; After completing the emergency avoidance action, a high-frequency wake-up signal is sent to the underwater sensing system to trigger the acoustic Doppler current profiler to resample in order to collect the latest environmental data of the current water area. The three-dimensional predictive environment model is reconstructed using the latest environmental data as the initial input, and the remaining time slices are replanned using the new environmental feature matrix.

Images