Hydraulic synchronous intelligent control system for segmental overturning of overweight steel structure

Abstract

The present application relates to the technical field of heavy equipment intelligent control, in particular to a hydraulic synchronous intelligent control system for super-heavy steel structure segmented roll-over, comprising: the present application discloses a kind of hydraulic synchronous intelligent control system for super-heavy steel structure segmented roll-over.The system acquires structure surface full-field strain data by deploying flexible mesh sensor array, constructs deformation gradient topology to identify instability risk area.Establish the dynamic fluid network model of hydraulic system, fluid-structure coupling simulation is carried out in combination with deformation characteristics, and the transmission path of pressure pulsation under different control instructions is predicted.The system dynamically optimizes data sampling strategy and network model parameters according to the predicted path, generates hydraulic cylinder coordinated action instruction verified by pressure balance, and finally drives the actuator to complete the roll-over operation.The scheme realizes the full-field real-time perception of structure deformation and the coupling prediction control of hydraulic fluid dynamics, and improves the safety and control accuracy of super-heavy component roll-over process.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for heavy equipment, and in particular to a hydraulic synchronous intelligent control system for the segmented turning of ultra-heavy steel structures. Background Technology

[0002] The safety of segmented overturning operations of ultra-heavy steel structures highly depends on real-time monitoring of the structure's deformation state. Current technologies generally employ monitoring methods that involve installing discrete strain or displacement sensors at pre-set points on the structure's surface. This method can only acquire local data from a limited number of points and cannot continuously and completely describe the overall deformation field of the structure throughout the entire operation. Due to the lack of a global understanding of the spatial distribution gradient of deformation and its evolution, existing methods struggle to identify high-risk areas of concentrated deformation or potential abrupt changes from a physical mechanism perspective before macroscopic instability occurs, resulting in insufficient accuracy and timeliness of safety warnings.

[0003] At the hydraulic synchronization control level, existing technologies typically treat each hydraulic actuator as an independent control unit, with the control system primarily relying on position or velocity feedback from each unit for following and correction. This control strategy fails to incorporate the inherent hydrodynamic characteristics of the hydraulic drive system into the control model, particularly neglecting the modeling and analysis of the transmission and superposition paths of pressure pulsations generated during the coordinated action of multiple actuators in complex pipeline networks. The action commands issued by the control system are decoupled from the impact of the resulting pressure fluctuations within the hydraulic system on the already supported structure. This makes the control process itself a potential source of disturbance, inducing unintended structural vibrations or localized stress concentrations, thus limiting further improvements in control accuracy and system dynamic stability.

[0004] Existing technologies suffer from two major drawbacks: incomplete perception of structural deformation and lack of coupled analysis of hydraulic control and structural dynamic response. There is a need for an intelligent control system capable of real-time perception of deformation across the entire field and coupled prediction and collaborative optimization of hydraulic system fluid dynamics and structural solid deformation. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a hydraulic synchronous intelligent control system for the segmented turning of ultra-heavy steel structures.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a hydraulic synchronous intelligent control system for segmented turning of ultra-heavy steel structures, comprising:

[0007] The deformation sensing module acquires strain field distribution images and displacement vectors of key hinge points on the structural surface through a flexible mesh sensor array deployed on the segmented surface, and inputs them into the embedded edge computing node to generate a gradient topology map of structural deformation.

[0008] The risk prediction module, based on the historical evolution sequence of the gradient topology graph, calculates the potential unstable areas of the structure during the overturning process, marks them as high-risk nodes, and extracts multiple sets of feature vectors from them;

[0009] The network modeling module establishes a dynamic fluid network model, abstracting the pipelines and actuators of the hydraulic synchronous intelligent control system into network nodes and edges. Based on the dynamic fluid network model, fluid-structure interaction simulation is performed on the feature vectors to predict the pressure pulsation transmission path under different control strategies.

[0010] The sampling optimization module reversely corrects the data sampling frequency of the embedded edge computing node based on the convergence point position of the pressure pulsation transmission path. Combined with the corrected data sampling frequency, it re-collects the real-time data of the high-risk node, optimizes the topology connection weight of the dynamic fluid network model online, and generates a set of coordinated action commands for hydraulic cylinders.

[0011] The instruction execution module, after verifying the pressure balance of the coordinated action instruction through the dynamic fluid network model, sends it to the corresponding hydraulic servo valve group to drive the ultra-heavy steel structure to complete the overturning action in segments.

[0012] Preferably, the generation of the gradient topology map of structural deformation includes:

[0013] The raw sensing data collected by the flexible grid sensor array is spatially aligned and time-stamped to form a spatiotemporally unified data cube;

[0014] On each slice of the data cube, an anisotropic diffusion algorithm is used to smooth noise while preserving the edge transition features caused by structural deformation;

[0015] The gradient magnitude and direction of the smoothed data in three-dimensional space are calculated, and the gradient magnitude is binarized and segmented according to the preset deformation threshold to identify the main deformation region and the secondary deformation region.

[0016] The center point of the main deformation region is set as the key vertex, and the center point of the secondary deformation region is set as the auxiliary vertex. Based on the continuity of the gradient direction between vertices, a gradient topology graph representing the deformation transmission relationship is constructed. The edge weight of the gradient topology graph is proportional to the gradient magnitude.

[0017] Preferably, the potential instability region of the calculated structure during the turning process includes:

[0018] Load the historical evolution sequence of the gradient topology graph, which is formed by stacking gradient topology graphs of multiple consecutive turning cycles in chronological order;

[0019] Calculate the variance of the deformation trajectory at each spatial location point in the historical evolution sequence, and mark the points whose variance exceeds the dynamic stability threshold as the initial instability points;

[0020] Centered on the initial instability point, and combined with the connection strength of the edges in the gradient topology graph of the structural deformation, a region growth is performed, and adjacent points with connection strength higher than the connectivity threshold are included in the same potential instability region.

[0021] For each of the potential instability regions, the release rate of its deformation energy is calculated, and the region with the fastest release rate is identified as a high-risk node.

[0022] Preferably, establishing the dynamic fluid network model includes:

[0023] The network parameters of the dynamic fluid network model are initialized based on the gradient topology graph;

[0024] Identify the physical locations and connections of all hydraulic cylinders, valve blocks, pipelines, and accumulators in the hydraulic synchronous intelligent control system, and map them as network nodes of the dynamic fluid network model.

[0025] Based on the length and diameter of the pipeline and the viscosity parameters of the fluid, the flow resistance coefficient and inertia coefficient of each edge in the dynamic fluid network model are calculated.

[0026] The location information of high-risk nodes marked in the gradient topology graph is spatially correlated with the location of the nearest hydraulic actuator, and this correlation is used as an additional constraint condition for the pressure monitoring points in the dynamic fluid network model.

[0027] Using the aforementioned flow resistance coefficient, inertia coefficient, and additional constraints, the state equations of a dynamic fluid network model with pressure and flow rate as state variables are constructed.

[0028] Preferably, performing fluid-structure interaction simulation on the feature vector includes:

[0029] The feature vector includes the strain accumulation rate and the displacement offset angle;

[0030] The strain accumulation rate in the eigenvector is converted into a reduction factor for the local stiffness of the structure, and the displacement offset angle is converted into a disturbance vector for the load direction of the hydraulic cylinder.

[0031] In the state equation of the dynamic fluid network model, the reduction coefficient and the disturbance vector are introduced as time-varying parameters to simulate the influence of structural deformation on the pressure distribution of the fluid network.

[0032] Multiple candidate valve group opening control strategies are set, each control strategy is executed sequentially in the dynamic fluid network model, and the state equations are solved to obtain the transient pressure distribution cloud map of the entire network.

[0033] Nodes whose pressure surges exceed the safety threshold are extracted from the transient pressure distribution cloud map, and the transmission links of the pressure surges in the dynamic fluid network model are traced. These transmission links are the pressure pulsation transmission paths.

[0034] Preferably, the reverse correction of the data sampling frequency of the embedded edge computing node includes:

[0035] Analyze the convergence point of the pressure pulsation transmission path to determine the critical pipeline area with the most severe pressure fluctuations;

[0036] Map the key pipeline area back to the physical space to find the corresponding structural monitoring area, which contains one or more high-risk nodes.

[0037] Based on the characteristic frequency of pressure change in the pressure pulsation transmission path, the minimum sampling frequency required to capture the characteristic frequency is calculated, and a preset safety redundancy is added to generate a new target sampling frequency.

[0038] Send a command to the embedded edge computing node responsible for the structure monitoring area to dynamically switch its data sampling frequency to the target sampling frequency.

[0039] Preferably, the topology connection weights of the dynamic fluid network model are optimized online, including:

[0040] Using real-time data acquired at the corrected sampling frequency, the deviation between the actual deformation rate and the expected deformation rate of each high-risk node is calculated.

[0041] The deviation is used as the input to the loss function, and the variable of the loss function is the connection weight of the key edge in the dynamic fluid network model;

[0042] An iterative shrinkage algorithm is used to calculate the gradient of the loss function with respect to the connection weights in each iteration, and to update the connection weights with an adaptive step size in the opposite direction of the gradient.

[0043] After each update, the new connection weights are substituted into the dynamic fluid network model, and the pressure balance of the entire network is simulated and calculated. The process stops when the pressure balance reaches the preset optimal range or the number of iterations is exhausted, and the final optimized set of connection weights is output.

[0044] Preferably, the generation of coordinated action commands for a set of hydraulic cylinders includes:

[0045] Based on the final optimized set of connection weights, the theoretical output force required by each hydraulic cylinder node in the dynamic fluid network model to balance the pressure of the entire network is recalculated.

[0046] The theoretical output force is converted into a target displacement curve for the corresponding hydraulic cylinder. The target displacement curve consists of multiple discrete time points and their corresponding displacement values.

[0047] The target displacement curve is kinematically smoothed to ensure that displacement, velocity, and acceleration are continuous without abrupt changes, and preliminary coordinated action commands are generated.

[0048] The initial cooperative action command is input again into the dynamic fluid network model to perform a forward simulation to verify whether the pressure of the entire system remains within the allowable range during command execution.

[0049] Preferably, the coordinated action command verifies its pressure equalization via the dynamic fluid network model, including:

[0050] The cooperative action command to be issued is used as the input boundary condition of the dynamic fluid network model;

[0051] Run the dynamic fluid network model to simulate the complete time history from the start to the end of the command, and output the pressure change curves of all nodes in the model throughout the entire time history;

[0052] Calculate the peak pressure difference and trough pressure difference for all pressure change curves, and count the number of nodes where the pressure change rate exceeds the critical value.

[0053] If both the peak pressure difference and the valley pressure difference are less than the synchronization tolerance, and the number of nodes with excessive pressure change rate is zero, then the pressure balance verification of the coordinated action command is deemed to have passed.

[0054] Preferably, driving the heavy steel structure to complete the turning action in segments includes:

[0055] The coordinated action command that has passed the pressure equalization verification is broken down into real-time control sub-commands for each hydraulic servo valve group according to the time sequence.

[0056] The real-time control sub-instructions are distributed to the local controllers of the corresponding hydraulic servo valve groups via a real-time industrial network.

[0057] The local controller drives the servo valve core to move according to the received real-time control sub-instructions, thereby controlling the flow and direction of the hydraulic cylinder.

[0058] All hydraulic cylinders operate synchronously under the coordination of the real-time control sub-command, pushing the heavy steel structure segments to move along the preset overturning trajectory until they reach the target posture.

[0059] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0060] A flexible grid sensor array deployed on segmented surfaces replaces traditional discrete point sensors, enabling the acquisition of strain field distribution images and displacement vectors covering the entire working area. This technical solution represents a leap in monitoring capabilities from discrete points to continuous fields, generating a gradient topology map characterizing the spatial rate of deformation change. As a result, the system gains real-time, panoramic perception of the overall structural deformation morphology, intuitively revealing areas of concentrated deformation and their evolution direction. Based on historical sequence analysis of the gradient topology, singularities or discontinuous regions in the deformation field can be identified. This allows for early detection of high-risk nodes based on spatial patterns before physical instability manifests, providing spatially correlated early warning information that far exceeds local threshold alarms for control decisions.

[0061] This approach abstracts the complex physical hydraulic system into a dynamic fluid network model, mapping pumps, valves, cylinders, and pipelines as network nodes and edges. Based on this, structural deformation characteristics are input for fluid-structure interaction simulation. This technical solution enables the system to calculate and visualize the transmission path and convergence point of pressure pulsations in the pipeline network under different control commands. This achieves feedforward prediction of the interaction between the internal dynamics of the hydraulic system and structural deformation. Based on the predicted pressure pulsation path, the system adjusts the data sampling strategy for high-risk areas in reverse and uses real-time data to optimize network model parameters online, forming a closed loop of perception-prediction-verification. The final generated hydraulic cylinder coordinated action command, after pressure equalization verification, can proactively avoid structural risks caused by harmful pressure wave superposition, realizing a paradigm shift from passive feedback correction to proactive predictive coordination in control. Attached Figure Description

[0062] Figure 1 This is a timing diagram of the hydraulic synchronous intelligent control system for segmented turning of ultra-heavy steel structures as described in this invention.

[0063] Figure 2 A flowchart for generating a gradient topology graph of structural deformation;

[0064] Figure 3 A flowchart for estimating the potential instability zone of the structure during the overturning process;

[0065] Figure 4 A bar chart comparing the theoretical output force of hydraulic cylinders at each stage during the segmented overturning process of an ultra-heavy steel structure;

[0066] Figure 5 Radar chart showing the performance of nodes in a hydraulic dynamic fluid network. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0068] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0069] See Figure 1 The deformation sensing module utilizes a flexible mesh sensor array deployed on the segmented surface to collect strain field distribution images of the structural surface and displacement vectors of key hinge points. This data is input into embedded edge computing nodes to generate a gradient topology map of structural deformation. The risk prediction module, based on the historical evolution sequence of the gradient topology map, calculates potential instability areas during the overturning process and marks them as high-risk nodes, while extracting multiple sets of feature vectors. The network modeling module establishes a dynamic fluid network model, abstracting the hydraulic system's pipelines and actuators as network nodes and edges. Based on this model, fluid-structure interaction simulation is performed on the extracted feature vectors to predict the pressure pulsation transmission path under different control strategies. The sampling optimization module, based on the convergence point location of the predicted pressure pulsation transmission path, reverse-corrects the data sampling frequency of the embedded edge computing nodes. Combined with the corrected frequency, real-time data from high-risk nodes is re-collected, and the topology connection weights of the dynamic fluid network model are optimized online, ultimately generating a set of coordinated action commands for the hydraulic cylinders. The command execution module verifies the pressure balance of this coordinated action command through the dynamic fluid network model before issuing it to the corresponding hydraulic servo valve group, driving the heavy-duty steel structure segments to complete the overturning action.

[0070] In one embodiment of the present invention, see [reference] Figure 2The raw sensor data collected by the flexible mesh sensor array is spatially aligned and time-stamped to form a spatiotemporally unified data cube. On each slice of the data cube, an anisotropic diffusion algorithm is used to smooth noise while preserving edge transition features caused by structural deformation. The gradient magnitude and direction of the smoothed data in three-dimensional space are calculated, and the gradient magnitude is binarized and segmented according to a preset deformation threshold to identify primary and secondary deformation regions. The center point of the primary deformation region is set as the key vertex, and the center point of the secondary deformation region is set as the auxiliary vertex. A gradient topology graph representing the deformation transmission relationship is constructed based on the continuity of gradient directions between vertices. The edge weights of this gradient topology graph are proportional to the gradient magnitude.

[0071] The deformation sensing module of the hydraulic synchronous intelligent control system for the segmented overturning of ultra-heavy steel structures acquires strain field distribution images of the structural surface and displacement vectors of key hinge points through a flexible mesh sensor array deployed on the segment surface. The raw sensing data acquired by the flexible mesh sensor array is input into an embedded edge computing node to generate a gradient topology map of structural deformation. In some embodiments, generating the gradient topology map of structural deformation involves processing the raw sensing data acquired by the flexible mesh sensor array. The embedded edge computing node spatially aligns the raw sensing data from the flexible mesh sensor array at different spatial locations, ensuring that all data points are in a unified three-dimensional coordinate system, and adds millisecond-accurate timestamps to all data points to achieve timestamp synchronization, thereby forming a spatiotemporally unified data cube containing both spatial and temporal dimensions. In a specific implementation, an anisotropic diffusion algorithm is used to smooth the data at each time slice of the spatiotemporally unified data cube. The anisotropic diffusion algorithm can effectively smooth random fluctuations caused by environmental disturbances or electronic noise, while preserving the edge transition characteristics that are spatially steep changes caused by the actual deformation of the structure. For example, in the scenario of a super-large ship section turning over, the boundary of the strain concentration area generated near the section support point in the initial stage of the turning action remains clear even after processing by the anisotropic diffusion algorithm.

[0072] In some embodiments, the gradient magnitude and direction of the smoothed data in three-dimensional space are calculated. For the strain value at each point in the data cube, its partial derivatives in the X, Y, and Z spatial directions are calculated to obtain the gradient vector at that point. The magnitude of the gradient vector is the gradient magnitude, characterizing the degree of deformation change at that point; the direction of the gradient vector indicates the direction in which deformation increases most rapidly in space. The calculated gradient magnitude is binarized and segmented according to a preset deformation threshold. Regions with gradient magnitudes exceeding the deformation threshold are identified as primary deformation regions, and regions with gradient magnitudes below the deformation threshold but above the background noise threshold are identified as secondary deformation regions.

[0073] Optionally, a method for calculating the midpoint in three-dimensional space. gradient magnitude The formula is expressed as:

[0074]

[0075] in: This represents the strain measurement value at that point. , , These represent the spatial rate of change of strain in the X, Y, and Z directions, respectively. In a turning operation example, the deformation threshold can be set to 150 microstrains, and all points with gradient amplitudes exceeding 150 microstrains constitute the main deformation region.

[0076] In practice, the center points of the identified primary deformation regions are designated as key vertices of the gradient topology graph, while the center points of secondary deformation regions are designated as auxiliary vertices. A gradient topology graph representing the deformation propagation relationship is constructed based on the continuity of gradient directions between vertices. If the direction of the line connecting two vertices is spatially continuous with the gradient directions at those vertices, then an edge connects these two vertices in the gradient topology graph. The edge weights of the gradient topology graph are proportional to the average gradient magnitude of the regions containing the two connected vertices. For example, the weight of an edge connecting two regions with an average gradient magnitude of 200 microstrain is twice the weight of an edge connecting two regions with an average gradient magnitude of 100 microstrain.

[0077] The resulting gradient topology graph is a network structure composed of vertices and weighted edges. Vertices represent the centers of regions where significant or slight deformation occurs in the structure, and weighted edges represent the spatial transmission paths and intensities of deformation. As an abstract mathematical description of the overall deformation pattern of the structure, the gradient topology graph is output to the risk prediction module for subsequent analysis.

[0078] In one embodiment of the present invention, see [reference] Figure 3The historical evolution sequence of the gradient topology graph is loaded, which is formed by stacking gradient topology graphs of multiple consecutive turning cycles in chronological order. The variance of the deformation trajectory at each spatial location point in the historical evolution sequence is calculated, and points with variances exceeding the dynamic stability threshold are marked as initial instability points. Using the initial instability point as the center, region growth is performed based on the connection strength of edges in the gradient topology graph of structural deformation, incorporating adjacent points with connection strengths higher than the connectivity threshold into the same potential instability region. The release rate of deformation energy is calculated for each potential instability region, and the region with the fastest release rate is identified as a high-risk node. The network parameters of the dynamic fluid network model are initialized based on the gradient topology graph. The physical locations and connections of all hydraulic cylinders, valve blocks, pipelines, and accumulators in the hydraulic synchronous intelligent control system are identified and mapped to network nodes of the dynamic fluid network model. Based on the pipeline length, diameter, and fluid viscosity parameters, the flow resistance coefficient and inertia coefficient of each edge in the dynamic fluid network model are calculated. The locations of high-risk nodes marked in the gradient topology graph are spatially correlated with the locations of the nearest hydraulic actuators. This correlation is then used as an additional constraint on the pressure monitoring points in the dynamic fluid network model. The state equations of the dynamic fluid network model, with pressure and flow rate as state variables, are constructed using the flow resistance coefficient, inertia coefficient, and additional constraints.

[0079] The risk prediction module loads the historical evolution sequence of the gradient topology graph, which is formed by stacking gradient topology graphs from multiple consecutive turning cycles in chronological order. In the example scenario of turning over sections of a super-large ship hull, the gradient topology graphs generated from the five most recent complete 180-degree turning operations can be selected and arranged in chronological order to form a historical evolution sequence containing five time layers. The gradient topology graph of each time layer contains the same spatial vertices and edge structures. In some embodiments, estimating the potential instability region of the structure during the turning process requires calculating the variance of the deformation trajectory at each spatial location point in the historical evolution sequence of the gradient topology graph. For each fixed spatial vertex in the historical evolution sequence, the gradient magnitude corresponding to that vertex in all five time layer gradient topology graphs is extracted to form a time series containing five data points. The variance of this time series is calculated, and the variance value characterizes the degree of fluctuation in the deformation response of that vertex during multiple turning operations. A dynamic stability threshold is set, and vertices with variances exceeding the dynamic stability threshold are marked as initial instability points.

[0080] Optional, deformation trajectory variance The calculation formula is as follows:

[0081]

[0082] in: Represents the gradient topology graph of the first A spatial vertex, Represents the number of gradient topological graphs contained in the historical evolution sequence (in the example). ), Representing the Vertices in a historical gradient topology graph gradient magnitude, Representing the vertex In all The average gradient magnitude in the historical gradient topology graph.

[0083] In practice, region growth is performed centered on the marked initial instability point, taking into account the edge connectivity strength in the gradient topology graph of structural deformation. The edge connectivity strength in the gradient topology graph is directly reflected by the edge weight. Starting from an initial instability point, all adjacent vertices directly connected to it via edges are examined. If the edge weight connecting this point is higher than a preset connectivity threshold, this adjacent vertex is included in the currently growing potential instability region. Subsequently, the above checking and inclusion process is repeated, starting with the newly included vertex, until no new adjacent vertices that meet the conditions can be included. This process aggregates adjacent points with connectivity strength higher than the connectivity threshold into the same potential instability region. For each potential instability region determined by region growth, its deformation energy release rate is calculated. The deformation energy release rate is characterized by calculating the average of the squared differences of the gradient magnitudes of all vertices in the region over the most recent two time layers. All potential instability regions are sorted from high to low according to their deformation energy release rates, and the region with the fastest release rate is identified as a high-risk node.

[0084] It is understandable that high-risk nodes represent localized structural regions that have exhibited unstable trends and rapid energy release during historical reversal processes. The network modeling module establishes a dynamic fluid network model, and the network parameters of the dynamic fluid network model are initialized based on the gradient topology graph. The initialization process includes using the spatial distribution information of high-risk nodes in the gradient topology graph as an input condition when constructing the dynamic fluid network model.

[0085] In some embodiments, establishing a dynamic fluid network model requires identifying the physical locations and connections of all hydraulic cylinders, valve blocks, pipelines, and accumulators in the hydraulic synchronous intelligent control system, and mapping these physical entities as network nodes in the dynamic fluid network model. Each independent hydraulic cylinder, valve block, or accumulator is abstracted as a network node, and each pipeline connecting two physical entities is abstracted as an edge connecting two network nodes. Based on the pipeline length, diameter, and fluid viscosity parameters, the flow resistance coefficient and inertia coefficient of each edge in the dynamic fluid network model are calculated. The flow resistance coefficient reflects the pressure loss characteristics caused by friction when the fluid flows through the pipeline, and the inertia coefficient reflects the inertial effect exhibited by the fluid when accelerating or decelerating in the pipeline.

[0086] In practical implementation, the location information of high-risk nodes marked in the gradient topology graph is spatially correlated with the location of the nearest hydraulic actuator. For example, if the gradient topology graph indicates that a certain area in the midships of the port side of the segmented structure is a high-risk node, the hydraulic cylinder closest to the geometric center of this area is located in the 3D model through coordinate mapping. The node corresponding to this hydraulic cylinder in the dynamic fluid network model is marked as a pressure monitoring point requiring key attention, and this correlation is used as an additional constraint condition for the pressure monitoring points in the dynamic fluid network model. Using the calculated flow resistance coefficient and inertia coefficient of each edge, as well as the above additional constraint conditions, the state equation of the dynamic fluid network model is constructed, with the pressure values ​​of all nodes and the flow rates of all edges in the system as state variables. The state equation is a set of differential equations describing the dynamic relationship between pressure and flow rate in the fluid network.

[0087] In one embodiment of the invention, the eigenvector includes the strain accumulation rate and the displacement offset angle. The strain accumulation rate in the eigenvector is converted into a reduction factor for the local stiffness of the structure, and the displacement offset angle is converted into a disturbance vector for the load direction of the hydraulic cylinder. The reduction factor and the disturbance vector are introduced as time-varying parameters into the state equation of the dynamic fluid network model to simulate the influence of structural deformation on the pressure distribution of the fluid network. Multiple candidate valve group opening control strategies are set, and each control strategy is executed sequentially in the dynamic fluid network model to solve the state equation, thereby obtaining the transient pressure distribution cloud map of the entire network. Nodes where the pressure mutation exceeds the safety threshold are extracted from the transient pressure distribution cloud map, and the transmission link of the pressure mutation in the dynamic fluid network model is traced. This transmission link is the pressure pulsation transmission path. The convergence point of the pressure pulsation transmission path is analyzed to determine the critical pipeline area with the most severe pressure fluctuation. The critical pipeline area is mapped back to physical space to find the corresponding structural monitoring area, which contains one or more high-risk nodes. Based on the characteristic frequency of pressure change in the pressure pulsation transmission path, the minimum sampling frequency required to capture the characteristic frequency is calculated, and a new target sampling frequency is generated by superimposing a preset safety redundancy. Send instructions to the embedded edge computing node responsible for the structural monitoring area to dynamically switch its data sampling frequency to the target sampling frequency.

[0088] In practical implementation, the feature vector includes the strain accumulation rate and displacement offset angle. Multiple sets of feature vectors are extracted from high-risk nodes output by the risk prediction module. For example, when a section of a large container ship is overturned to a 60-degree angle, the strain accumulation rate of a certain high-risk node is measured to be 0.018 per second, and the displacement offset angle is measured to be 3.2 degrees. The strain accumulation rate in the feature vector is converted into a reduction factor for the local stiffness of the structure. The conversion relationship is linear reduction; the higher the strain accumulation rate, the smaller the reduction factor, indicating a greater decrease in local stiffness. The displacement offset angle is converted into a disturbance vector in the load direction of the hydraulic cylinder. The displacement offset angle is used to calculate a deflection amount added to the original load direction of the hydraulic cylinder.

[0089] In some embodiments, a reduction factor and a disturbance vector are introduced as time-varying parameters into the state equation of the dynamic fluid network model to simulate the influence of structural deformation on the pressure distribution of the fluid network. The reduction factor is directly multiplied into the equivalent stiffness parameter of the branch where the hydraulic actuator associated with the high-risk node is located, while the disturbance vector is added as an additional force vector to the load force boundary condition of the corresponding hydraulic cylinder node. Multiple candidate valve group opening control strategies are set, such as strategy A for all valve groups to open synchronously at a uniform speed, strategy B for opening the corresponding valve group in advance according to the location of the high-risk node, and strategy C for using a stepped opening curve. Each control strategy is executed sequentially in the dynamic fluid network model, and the state equation is solved to obtain the transient pressure distribution cloud map of the entire network.

[0090] Optionally, nodes where pressure surges exceed a safety threshold can be extracted from the transient pressure distribution cloud map. A pressure surge is defined as an absolute pressure change greater than 2.5 MPa within two consecutive simulation time steps. The transmission path of the pressure surge in the dynamic fluid network model is traced. The transmission path is the path from the starting node of the pressure surge to other nodes along the network edges; this transmission path is the pressure pulsation transmission path. In one simulation case, when strategy A is used, the pressure surge starts from the pump station node numbered P7, and is transmitted to the four hydraulic cylinder nodes via three main pipelines, forming a clear tree-like transmission path.

[0091] It is understandable that the pressure pulsation transmission path reveals the propagation law of pressure fluctuations within a system under a specific control strategy. Analyzing the convergence point of the pressure pulsation transmission path—a node in the dynamic fluid network model to which multiple transmission links converge—identifies the critical pipeline region where pressure fluctuations are most severe. This critical pipeline region includes the network portion covered by the convergence point and its directly connected edges. Mapping this critical pipeline region back to physical space reveals the corresponding structural monitoring area. For example, the critical pipeline region corresponds to the valve group cluster numbered VALVE_GROUP_3 in the dynamic fluid network model, which is located in the lower left rear region of the segmented structure in physical space. The structural monitoring area includes one or more high-risk nodes defined by the deformation sensing module near this physical location.

[0092] In practical implementation, the minimum sampling frequency required to capture the characteristic frequency is calculated based on the characteristic frequency of pressure changes in the pressure pulsation transmission path. The characteristic frequency of pressure changes is obtained by performing spectral analysis on the pressure time series of key nodes in the transmission path to identify the dominant frequency component with the highest amplitude. The calculation of the minimum sampling frequency must satisfy the Nyquist sampling theorem, and a preset safety redundancy is added to generate a new target sampling frequency. Commands are sent to the embedded edge computing nodes responsible for the structural monitoring area to dynamically switch their data sampling frequency to the target sampling frequency.

[0093] Optional, target sampling frequency The calculation formula is as follows:

[0094]

[0095] in: This represents the highest characteristic frequency extracted from the pressure pulsation transmission path analysis, measured in Hertz. It is a sampling coefficient greater than 2, used to satisfy the basic requirements of the Nyquist sampling theorem; This is a preset safety redundancy frequency value, measured in Hertz, used to address frequency estimation errors or higher-order harmonics. It refers to the highest characteristic frequency extracted from data from a specific turning point. Set the sampling coefficient to 8 Hz. The safety redundancy frequency value is 2.5. The target sampling frequency is calculated to be 10 Hz. It is 30 Hz.

[0096] In some embodiments, upon receiving an instruction, the embedded edge computing node dynamically switches its data sampling frequency for the flexible mesh sensor array from the default 20 Hz to a calculated 30 Hz to ensure complete capture of the dynamic deformation characteristics of high-risk nodes associated with pressure pulsations. The sampling frequency correction is a closed-loop process; if the characteristic frequency of the pressure pulsation transmission path changes during subsequent turning actions, the sampling optimization module recalculates and readjusts the sampling frequency. It can be understood that through this sampling frequency reverse correction mechanism based on pressure pulsation transmission path analysis, the data acquisition resources of the embedded edge computing node are preferentially concentrated on the structural monitoring areas that have the most significant impact on the system's fluid dynamics, thereby providing more timely and relevant input data for the online optimization of the dynamic fluid network model.

[0097] In one embodiment of the invention, real-time data collected using a modified sampling frequency is used to calculate the deviation between the actual deformation rate and the expected deformation rate of each high-risk node. This deviation is used as input to a loss function, the variables of which are the connection weights of key edges in the dynamic fluid network model. An iterative shrinkage algorithm is employed, calculating the gradient of the loss function with respect to the connection weights in each iteration, and updating the connection weights with an adaptive step size in the opposite direction of the gradient. After each update, the new connection weights are substituted into the dynamic fluid network model, and the overall network pressure balance is simulated. The simulation stops when the pressure balance reaches a preset optimal range or the number of iterations is exhausted, and the final optimized set of connection weights is output. Based on the final optimized set of connection weights, the theoretical output force required by each hydraulic cylinder node in the dynamic fluid network model to balance the overall network pressure is recalculated. The theoretical output force is converted into the target displacement curve of the corresponding hydraulic cylinder, which consists of multiple discrete time points and corresponding displacement values. The target displacement curve is kinematically smoothed to ensure that displacement, velocity, and acceleration are continuous without abrupt changes, generating preliminary coordinated action commands. The initial coordinated action commands were input again into the dynamic fluid network model for a forward simulation to verify whether the pressure of the entire system remained within the allowable range during command execution.

[0098] In practical implementation, the sampling optimization module uses real-time data acquired at a corrected sampling frequency to calculate the deviation between the actual and expected deformation rates of each high-risk node. The actual deformation rate is calculated in real-time by the embedded edge computing node from the flexible mesh sensor array at the updated frequency, while the expected deformation rate comes from the simulation prediction value of the dynamic fluid network model under the current control strategy. Taking a high-risk node as an example, its actual deformation rate at the 120th second of the turning action is 0.25 degrees per second, while the expected deformation rate predicted by the dynamic fluid network model is 0.22 degrees per second, with a deviation of 0.03 degrees per second.

[0099] In some embodiments, the deviation between the actual and expected deformation rates of each high-risk node is used as the input to the loss function, where the variables are the connection weights of key edges in the dynamic fluid network model. Key edges refer to the edges connecting the nodes containing the hydraulic actuators of high-risk nodes in the dynamic fluid network model, as well as the main transmission edges identified on the pressure pulsation transmission path. The loss function aims to minimize the sum of squares of the deviations of all high-risk nodes. An iterative shrinkage algorithm is used to optimize the connection weights online. In each iteration, the gradient of the loss function with respect to the connection weights is calculated; the gradient indicates the direction and extent of the influence of changes in connection weights on the overall deviation.

[0100] Optionally, an iterative shrinking algorithm for updating connection weights can be expressed as follows:

[0101]

[0102] in: Representing the The set vector of key edge connection weights in the dynamic fluid network model at each iteration; Represents the loss function exist The gradient of the set of connection weight vectors; It is the first The adaptive step size of the next iteration is adjusted according to the ratio of the gradient magnitude of the current iteration to the gradient magnitude of the previous iteration; This represents the updated set vector of connection weights.

[0103] In practice, the connection weights are updated with an adaptive step size along the opposite direction of the gradient. After each update, the new set of connection weights is substituted into the dynamic fluid network model, and a simplified transient simulation is run to calculate the overall network pressure balance. The overall network pressure balance is defined as the average of the standard deviations of the pressures of all hydraulic cylinder nodes within the simulation time window. The optimization process stops when the pressure balance reaches the preset optimal range or the number of iterations is exhausted, and the final optimized set of connection weights is output. Refer to Table 1 for an example of the changes in the connection weights of critical edges during an optimization process.

[0104] Table 1: Online Optimization Process of Key Edge Connection Weights

[0105]

[0106] In some embodiments, the theoretical output force required by each hydraulic cylinder node in the dynamic fluid network model to balance the overall network pressure is recalculated based on the final optimized set of connection weights. The state equations of the optimized dynamic fluid network model are recalculated to solve for the output force of each hydraulic cylinder node that minimizes the overall network pressure difference under the target overturning posture. The theoretical output force is converted into the target displacement curve of the corresponding hydraulic cylinder. The target displacement curve consists of multiple discrete time points and their corresponding displacement values. The conversion process is based on the force-displacement characteristic curve of the hydraulic cylinder and the static relationship of the segmented structure.

[0107] It is understandable that kinematic smoothing is applied to the target displacement curve to ensure that displacement, velocity, and acceleration are continuous without abrupt changes, generating preliminary coordinated action commands. The smoothing process employs a fifth-order polynomial interpolation method to construct a smooth displacement-time function between adjacent discrete time points. The preliminary coordinated action commands are then input back into the dynamic fluid network model to perform a forward simulation of the complete time history, verifying whether the overall system pressure remains within acceptable limits during command execution. For example, verifying whether the pressure at all nodes remains within the system's safe operating pressure range of 5 MPa to 28 MPa.

[0108] See Figure 4This is a bar chart comparing the theoretical output force of hydraulic cylinders at different stages during the segmented overturning process of a super-heavy steel structure, clearly showing the load variation of each hydraulic cylinder at different overturning stages. The load peak occurs at the 45° overturning stage, where the theoretical output force of all four hydraulic cylinders reaches its maximum. This is because the horizontal distance between the segment's center of gravity and the support point is the greatest at this stage, causing each hydraulic cylinder to bear the greatest torque, making it the most critical stress stage in the entire overturning process. The output force of all hydraulic cylinders exhibits a symmetrical variation pattern of "initial stage → rising → reaching peak at 45° overturning → falling → positioning stage," indicating that the hydraulic synchronous control system performs well in coordinating the actions of each cylinder. In all stages, the theoretical output force of hydraulic cylinder 4 is the largest among the four cylinders, which is closely related to its support position and stress characteristics in the structure. The state of this cylinder needs to be closely monitored in actual engineering projects.

[0109] In one embodiment of the present invention, the coordinated action command to be issued is used as the input boundary condition of the dynamic fluid network model. The dynamic fluid network model is run to simulate the complete time history from the start to the end of the command, and the pressure change curves of all nodes in the model during the entire time history are output. The peak pressure difference and valley pressure difference of all pressure change curves are calculated, and the number of nodes whose pressure change rate exceeds the critical value is counted. If both the peak pressure difference and valley pressure difference are less than the synchronization tolerance, and the number of nodes whose pressure change rate exceeds the limit is zero, then the pressure balance verification of the coordinated action command is deemed to have passed. The coordinated action command that has passed the pressure balance verification is decomposed into real-time control sub-commands for each hydraulic servo valve group according to the time sequence. The real-time control sub-commands are distributed to the local controller of the corresponding hydraulic servo valve group through the real-time industrial network. The local controller drives the servo valve core to move according to the received real-time control sub-commands, thereby controlling the flow and direction of the hydraulic cylinder. All hydraulic cylinders move synchronously under the coordination of the real-time control sub-commands, pushing the heavy steel structure segments to move along the preset overturning trajectory until they reach the target posture.

[0110] In practical implementation, the pressure uniformity of the coordinated action command is verified through a dynamic fluid network model. The verification process uses the coordinated action command to be issued as the input boundary condition of the dynamic fluid network model; that is, the target displacement curve of each hydraulic cylinder in the coordinated action command is converted into the displacement input condition of the corresponding node in the dynamic fluid network model. The dynamic fluid network model is run to simulate the complete time history from the start to the end of the command, with a simulation time step set to 0.01 seconds, simulating a 120-second turning motion, and outputting the pressure change curves of all nodes in the dynamic fluid network model throughout the entire time history.

[0111] In some embodiments, the peak pressure difference and valley pressure difference of all pressure change curves are calculated. The peak pressure difference is defined as the maximum value of the difference between the highest and lowest pressures at all hydraulic cylinder nodes at the same time over the entire time history, and the valley pressure difference is defined as the minimum value of the difference between the highest and lowest pressures at all hydraulic cylinder nodes at the same time over the entire time history. The number of nodes whose pressure change rate exceeds a critical value is counted. The pressure change rate is calculated by dividing the difference between the pressure change curves at any two adjacent simulation time points by the time step, and the critical value is set to 15 MPa per second. If both the peak pressure difference and the valley pressure difference are less than the synchronization tolerance, and the number of nodes with pressure change rates exceeding the limit is zero, then the pressure balance verification of the coordinated action command is considered successful. In one example, the simulation results show that the peak pressure difference is 1.8 MPa, the valley pressure difference is 0.2 MPa, the set synchronization tolerance is 2.0 MPa, and no node has a pressure change rate exceeding 15 MPa per second, therefore the verification is successful.

[0112] Optional, peak pressure difference The calculation formula is as follows:

[0113]

[0114] in: Represents the simulation time. Represents the total simulation duration. This represents the set of all hydraulic cylinder actuator nodes in the dynamic fluid network model. Represents time Time node The pressure value. Indicates at time The highest pressure value among all hydraulic cylinder nodes. Indicates at time The lowest pressure value among all hydraulic cylinder nodes. That is, the entire time course. The maximum value of this pressure difference.

[0115] In practical implementation, the coordinated action command that has passed pressure equalization verification is broken down into real-time control sub-commands for each hydraulic servo valve group according to time sequence. These real-time control sub-commands include the target opening degree of the valve group, the rate of change of the target opening degree, and the effective time window. Through a real-time industrial network, these real-time control sub-commands are distributed to the local controllers of the corresponding hydraulic servo valve groups. The real-time industrial network uses a deterministic Ethernet protocol based on Ethernet to ensure the timeliness and determinism of command transmission. It can be understood that the local controller of the hydraulic servo valve group drives the servo valve core to move according to the received real-time control sub-commands, thereby controlling the flow and direction entering the hydraulic cylinder. The local controller has a built-in high-response servo amplifier and position feedback sensor, which can accurately track the target valve core position given by the real-time control sub-commands. All hydraulic cylinders move synchronously under the coordination of the real-time control sub-commands, pushing the heavy-duty steel structure segments along a preset overturning trajectory until they reach the target posture.

[0116] See Figure 5 This is a radar chart showing the performance of nodes in a hydraulic dynamic fluid network. It provides a direct comparison of the performance of the hydraulic pump, accumulator, and five hydraulic cylinders from two dimensions: pressure stability and flow efficiency. The hydraulic pump performs best in both pressure stability and flow efficiency (approaching 100%). As the system's power source, its high stability provides a solid guarantee for the entire hydraulic system. The red "Pressure Stability" curve consistently surrounds the blue "Flow Efficiency" curve, indicating that pressure stability is generally higher than flow efficiency across all monitored nodes, reflecting a more mature design in pressure control. The accumulator exhibits high levels of both pressure stability and flow efficiency, demonstrating its crucial role in absorbing pressure fluctuations and stabilizing system flow.

[0117] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A hydraulic synchronous intelligent control system for turning over a super-heavy steel structure section, characterized in that, The method comprises the following steps: a deformation perception module acquires a strain field distribution image of a structure surface and a displacement vector of a key hinge point through a flexible mesh sensor array deployed on a segmented surface, and inputs the image and the vector into an embedded edge computing node to generate a gradient topology graph of structure deformation; a risk prediction module calculates potential instability regions of the structure during a turning-over process according to a historical evolution sequence of the gradient topology graph, marks the regions as high-risk nodes, and extracts a plurality of feature vectors from the regions; a network modeling module abstracts pipelines and actuators of a hydraulic synchronous intelligent control system as network nodes and edges based on a dynamic fluid network model, and performs fluid-structure coupling simulation on the feature vectors based on the dynamic fluid network model to predict pressure pulsation transmission paths under different control strategies; a sampling optimization module reversely corrects a data sampling frequency of the embedded edge computing node according to positions of converging points of the pressure pulsation transmission paths, reacquires real-time data of the high-risk nodes in combination with the corrected data sampling frequency, optimizes topology connection weights of the dynamic fluid network model online, and generates a group of coordinated action instructions of hydraulic cylinders; an instruction execution module issues the coordinated action instructions to corresponding hydraulic servo valve groups after verifying pressure balance of the instructions via the dynamic fluid network model, and drives the segmented super-heavy steel structure to complete a turning-over action.

2. The hydraulic synchronous intelligent control system for the super-heavy steel structure section to roll over, according to claim 1, characterized in that, The gradient topology graph of structure deformation comprises the following steps: spatial alignment and time stamp synchronization are performed on original sensing data acquired by the flexible mesh sensor array to form a spatiotemporal unified data cube; an anisotropic diffusion algorithm is used to smooth noise on each slice of the data cube while retaining edge jump characteristics caused by structure deformation; gradient amplitudes and directions in a three-dimensional space are calculated for the smoothed data, and the gradient amplitudes are binarized and segmented according to a preset deformation threshold to identify major deformation regions and minor deformation regions; center points of the major deformation regions are set as key vertices, center points of the minor deformation regions are set as auxiliary vertices, and a gradient topology graph representing deformation transmission relationships is constructed according to gradient direction continuity between the vertices, and edge weights of the gradient topology graph are proportional to gradient amplitudes.

3. The hydraulic synchronous intelligent control system for the super-heavy steel structure section to roll over, according to claim 1, characterized in that, The potential instability regions of the structure during the turning-over process comprise the following steps: a historical evolution sequence of the gradient topology graph is loaded, the historical evolution sequence is formed by stacking gradient topology graphs of a plurality of consecutive turning-over cycles in chronological order; deformation trajectory variances of each spatial position point in the historical evolution sequence are calculated, and points with variances exceeding a dynamic stability threshold are marked as initial instability points; region growing is performed with the initial instability points as centers and in combination with connection strengths of edges in the gradient topology graph of structure deformation, and adjacent points with connection strengths higher than a connectivity threshold are included in the same potential instability region; a release rate of deformation energy is calculated for each potential instability region, and a region with the fastest release rate is determined as a high-risk node.

4. The hydraulic synchronous intelligent control system for the super-heavy steel structure to be rolled over by sections according to claim 1, characterized in that, The dynamic fluid network model is established by the following steps: network parameters of the dynamic fluid network model are initialized according to the gradient topology graph. Identify the physical location and connection relationship of all hydraulic cylinders, valve blocks, pipelines and accumulators in the hydraulic synchronous intelligent control system, and map them to the network nodes of the dynamic fluid network model; According to the length, diameter of the pipeline and the viscosity parameters of the fluid, the flow resistance coefficient and the inertia coefficient of each edge in the dynamic fluid network model are calculated; The high-risk node position information marked in the gradient topology diagram is spatially associated with the position of the nearest hydraulic actuator, and this association is taken as an additional constraint condition for the pressure monitoring points in the dynamic fluid network model; Using the flow resistance coefficient, inertia coefficient and additional constraint condition, the state equation of the dynamic fluid network model with pressure and flow as state variables is constructed.

5. The hydraulic synchronous intelligent control system for the super-heavy steel structure section to roll over, according to claim 4, characterized in that, Fluid-structure coupling simulation is performed on the feature vector, including: The feature vector contains strain accumulation rate and displacement offset angle; The strain accumulation rate in the feature vector is converted into a reduction coefficient for the local stiffness of the structure, and the displacement offset angle is converted into a disturbance vector for the load direction of the hydraulic cylinder; In the state equation of the dynamic fluid network model, the reduction coefficient and disturbance vector are introduced as time-varying parameters to simulate the influence of structure deformation on fluid network pressure distribution; A plurality of candidate valve group opening control strategies are set, each control strategy is executed in the dynamic fluid network model in turn, and the state equation is solved to obtain the transient pressure distribution cloud diagram of the whole network; From the transient pressure distribution cloud diagram, nodes with pressure mutation exceeding the safety threshold are extracted, and the transmission link of pressure mutation in the dynamic fluid network model is tracked, which is the pressure pulsation transmission path.

6. The hydraulic synchronous intelligent control system for the super-heavy steel structure to segment roll over according to claim 1, characterized in that, The data sampling frequency of the embedded edge computing node is corrected in reverse, including: Analyze the convergence point position of the pressure pulsation transmission path to determine the critical pipe network area with the most severe pressure fluctuation; Map the critical pipe network area back to the physical space to find the corresponding structure monitoring area, which contains one or more high-risk nodes; According to the characteristic frequency of pressure change in the pressure pulsation transmission path, calculate the minimum sampling frequency required to capture the characteristic frequency, and superimpose a preset safety redundancy to generate a new target sampling frequency; Send instructions to the embedded edge computing node responsible for the structure monitoring area to dynamically switch its data sampling frequency to the target sampling frequency.

7. The hydraulic synchronous intelligent control system for the super-heavy steel structure to segment roll over according to claim 1, characterized in that, Online optimization of the topological connection weight of the dynamic fluid network model, including: Using real-time data collected with the corrected sampling frequency, calculate the deviation between the actual deformation rate and the expected deformation rate of each high-risk node; Input the deviation as a loss function, and the variable of the loss function is the connection weight of the key edge in the dynamic fluid network model; Using the iterative shrinkage algorithm, in each iteration, calculate the gradient of the loss function with respect to the connection weight, and update the connection weight in the opposite direction of the gradient with an adaptive step size; After each update, substitute the new connection weight into the dynamic fluid network model, and simulate the pressure equalization degree of the whole network. When the pressure equalization degree reaches the preset optimal range or the number of iterations is exhausted, stop and output the final optimized connection weight set.

8. The hydraulic synchronous intelligent control system for the super-heavy steel structure section to roll over, according to claim 7, characterized in that, The generating a set of hydraulic cylinder cooperative action instructions comprises: According to the final optimized connection weight set, the theoretical output force required by each hydraulic cylinder node in the dynamic fluid network model to balance the whole network pressure is recalculated; The theoretical output force is converted into the target displacement curve of the corresponding hydraulic cylinder, and the target displacement curve is composed of a plurality of discrete time points and corresponding displacement values; Kinematic smoothing processing is performed on the target displacement curve to ensure continuous and non-jump displacement, velocity and acceleration, and a preliminary cooperative action instruction is generated; The preliminary cooperative action instruction is input into the dynamic fluid network model again for forward simulation to verify whether the whole system pressure is always within the allowable range during instruction execution.

9. The hydraulic synchronous intelligent control system for the super-heavy steel structure to segment roll over according to claim 1, characterized in that, The pressure balance of the cooperative action instruction is verified by the dynamic fluid network model, comprising: The cooperative action instruction to be issued is taken as the input boundary condition of the dynamic fluid network model; The dynamic fluid network model is run to simulate the complete time history from the start to the end of the instruction, and output the pressure change curve of all nodes in the model during the whole time history; The peak pressure difference and the valley pressure difference of all pressure change curves are calculated, and the number of nodes with pressure change rate exceeding the critical value is counted; If the peak pressure difference and the valley pressure difference are both less than the synchronization tolerance, and the number of nodes with pressure change rate exceeding the critical value is zero, it is determined that the pressure balance verification of the cooperative action instruction is passed.

10. The hydraulic synchronous intelligent control system for the super-heavy steel structure section to roll over, according to claim 9, characterized in that, Driving the super-heavy steel structure segment to complete the turning-over action comprises: The cooperative action instruction whose pressure balance verification is passed is disassembled into immediate control sub-instructions for each hydraulic servo valve group according to the time sequence; The immediate control sub-instructions are distributed to the local controllers of the corresponding hydraulic servo valve groups through the real-time industrial network; The local controllers drive the servo valve core to move according to the received immediate control sub-instructions, thereby controlling the flow and direction of the hydraulic cylinder; All hydraulic cylinders act synchronously under the coordination of the immediate control sub-instructions to push the super-heavy steel structure segment to move along the preset turning-over trajectory until reaching the target posture.

Images