A channel engineering earthwork monitoring management system based on big data analysis

Abstract

The application relates to the technical field of channel dredging monitoring, and discloses a channel engineering earthwork monitoring management system based on big data analysis, which comprises a data acquisition layer, which collects the flow and density parameters, the draft depth parameters and the tidal water level of a sludge discharge pipeline; a processing server, which disperses a water area to be excavated into three-dimensional space voxels, calculates a time bias compensation factor to align data time stamps, calculates actual received wet weight loads based on the draft depth, calculates actual total dry matter mass by using the aligned parameters to update a dynamic conversion coefficient, combines a consumption model to calculate residual soil mass to generate a bottom mud viscosity label and write into the voxels, uses a neural network to predict an optimal loading wet weight threshold value based on the dynamic conversion coefficient and the bottom mud viscosity label to execute stop control, and uses a genetic algorithm to generate a dispatching and scheduling sequence. The application eliminates fluid transmission errors, avoids single-ship overflow or idle transportation capacity, and reduces the energy consumption of multi-ship collaborative scheduling.

Description

Technical Field

[0001] This invention relates to the field of waterway dredging monitoring technology, specifically a waterway engineering earthwork monitoring and management system based on big data analysis. Background Technology

[0002] In waterway dredging and land reclamation projects, dredgers are used in conjunction with multiple transport barges to excavate and dump mud. Accurately monitoring the amount of excavated earth and rationally scheduling multiple barges are crucial for controlling construction energy consumption and progress. Currently, the monitoring of earth volume mainly relies on geological survey data from the early stages of construction and the static draft of the transport barges before and after loading for estimation.

[0003] In actual construction operations, slurry is transported to the dredger via a slurry discharge pipeline. The flow and density sensors on the pipeline are installed on the dredger body or near the pump body, at a physical distance from the final discharge port. Existing technology directly extracts the sensor values ​​and the dredger's current loading time simultaneously when recording data, without considering the fluid movement time of the slurry during its transmission within the pipeline. This direct recording method causes a misalignment between the flow data measured by the sensors and the actual time the dredger receives the slurry, resulting in numerical calculation errors when calculating the total dry matter mass of a single load.

[0004] Furthermore, the physical properties of the underwater geological environment are constantly changing, and the volume expansion characteristics of soil in different areas after excavation vary significantly. The bottom sediment in some waters has strong viscosity, and when the barge arrives at the dumping area for unloading, some soil adheres to the bottom of the cargo hold and cannot be completely emptied. Existing monitoring systems use a uniform fixed volume conversion factor for earthwork conversion across the entire area, and assume that the barge is completely empty each time when calculating the carrying capacity of a single vessel, ignoring the impact of residual soil at the bottom of the hold on the actual usable volume and carrying capacity of the barge. Since the calculation parameters cannot be dynamically updated based on the actual expansion rate and adhesion characteristics of the bottom sediment at the working coordinates, construction personnel often have to rely on experience to set fixed loading times or fixed drafts to execute shutdown commands. This fixed threshold control method often results in mud overflow caused by excessive soil expansion during single-vessel operations, or premature shutdown before reaching the safe load limit.

[0005] Regarding multi-vessel collaborative scheduling, existing mud barge scheduling schemes are mostly based on fixed theoretical speeds and experienced loading times for cyclical dispatch. Because they fail to take into account the extended mud dumping time caused by local geological viscosity and the fluctuations in single operation time caused by mismatched loading thresholds, in actual construction, dredgers are often forced to stop and wait due to the lack of available barges, or multiple mud barges arrive at the vicinity of the dredger at the same time and queue up to be loaded. This disconnect between scheduling instructions and actual physical conditions results in high idling fuel consumption of the vessel formation and a reduction in overall construction efficiency. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a waterway engineering earthwork monitoring and management system based on big data analysis. This system solves the problems of existing waterway dredging monitoring failing to compensate for pipeline fluid transmission time offsets and ignoring the dynamic expansion and adhesion characteristics of bottom sediment, resulting in large errors in loading calculations and idle capacity of multiple vessels during scheduling.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solution: This invention provides a waterway engineering earthwork monitoring and management system based on big data analysis, comprising: The data acquisition layer collects instantaneous volumetric flow rate parameters, mud density parameters, draft parameters, and tidal water level parameters of the sludge discharge pipeline; The processing server, connected to the data acquisition layer, is configured to: discretize the water area to be excavated into three-dimensional spatial voxels; calculate the hydrodynamic time offset compensation factor to align the data timestamps; calculate the actual received wet weight load based on the draft depth parameter, and calculate the actual total dry matter mass using the aligned instantaneous volumetric flow rate parameter and mud density parameter to update the dynamic conversion factor; calculate the mass parameter of the residual mud at the bottom of the tank using the consumption model to generate a mud viscosity label and write it into the three-dimensional spatial voxels; and use a neural network to predict the optimal loading wet weight threshold based on the dynamic conversion factor and the mud viscosity label to execute shutdown control, and use a genetic algorithm to generate a scheduling sequence.

[0008] By employing the above technical solution, this invention extracts the parameters of the sludge discharge pipeline and the hydrostatic parameters of the vessel, and introduces time offset compensation to eliminate fluid transmission hysteresis errors caused by pipeline length, thus aligning the sensor readings with the actual loading status of the vessel on the timeline. Simultaneously, by utilizing the mass conservation relationship before and after loading and the consumption model during the vessel's empty return voyage, the physical volume conversion rate and sediment adhesion characteristics of the current three-dimensional spatial voxels are derived, generating dynamic geological tags. Using these geological tags as input features for the feedforward neural network and scheduling algorithm, the system can calculate the single-vessel shutdown load and the multi-vessel entry / exit sequence based on the actual geological conditions of the construction area, resolving loading errors and scheduling conflicts in multi-vessel collaborative operations.

[0009] Furthermore, discretizing the water area to be excavated into three-dimensional spatial voxels specifically includes: The design boundary and target excavation depth parameters of the waterway engineering are obtained by extracting the three-dimensional design model data. The three-dimensional mesh is generated by dividing the three-dimensional space into a set of three-dimensional spatial voxels using the preset physical spatial resolution parameters. The spatial coordinates of the geological borehole sampling points and their corresponding undisturbed soil dry density physical test values ​​are extracted by analyzing the geological survey data. The undisturbed dry density benchmark values ​​are assigned to each three-dimensional spatial voxel by spatial mapping fitting.

[0010] Through the above steps, the underwater construction area is divided into discrete three-dimensional grids, and the test values ​​of borehole sampling are assigned to corresponding three-dimensional spatial voxels, providing basic data support for the spatial positioning of subsequent excavation parameters and the dynamic updating of geological properties.

[0011] Furthermore, the computational fluid dynamics time bias compensation factor aligns the data timestamps, specifically including: When the instantaneous volumetric flow rate parameter is greater than the preset lower limit threshold, the instantaneous fluid velocity of the slurry inside the slurry discharge pipeline is calculated by combining the physical cross-sectional area parameter inside the slurry discharge pipeline; the fluid dynamics time offset compensation factor is generated by dividing the physical length parameter of the slurry discharge pipeline by the instantaneous fluid velocity; the fluid dynamics time offset compensation factor is subtracted from the actual loading start time and the actual loading end time of the slurry to estimate the historical measurement time interval when the corresponding slurry entity flows through the sensor detection point.

[0012] Through the above steps, the time required for the slurry to flow from the measurement point to the discharge port is calculated using pipeline geometric parameters and instantaneous flow velocity. This corrects the time reference of the sensor measurement sequence and eliminates measurement lag errors caused by the sensor installation position.

[0013] Furthermore, the step of calculating the actual received wet weight load based on the draft depth parameter, and calculating the actual total dry matter mass using the aligned instantaneous volumetric flow rate parameter and mud density parameter, specifically includes: Effective smooth data segments are extracted from the draft depth parameters, and the hydrostatic curve function is called to convert them into standard total displacement parameters. After density compensation calibration, the difference between the standard total displacement parameters before and after loading is calculated to generate the actual received wet weight load. The volume concentration is calculated by dividing the difference between the mud density parameter and the ambient water density by the difference between the reference density parameter and the ambient water density. This concentration is then multiplied by the instantaneous volumetric flow rate parameter and the reference density parameter, and a definite integral operation is performed along the historical measurement time interval to calculate the actual total dry matter mass.

[0014] Through the above steps, the total weight of the mixture actually carried by the ship is calculated using the draft, the instantaneous flow rate of solid matter is extracted by combining the density ratio, and the total weight of dry matter in a single loading process is obtained by integral calculation along the time interval.

[0015] Furthermore, the updated dynamic conversion factor specifically includes: The equivalent underwater undisturbed volume of a single voyage is calculated by dividing the total actual dry mass of the material by the comprehensive undisturbed dry density parameter of the three-dimensional voxels at the excavation source. When the equivalent underwater undisturbed volume is determined to be greater than a preset effective volume threshold, the physical conversion ratio of a single voyage is calculated by dividing the actual received wet weight load by the equivalent underwater undisturbed volume. A time-series exponential smoothing weighted algorithm is introduced to fuse the physical conversion ratio of a single voyage with the historical value of the dynamic conversion coefficient of the corresponding voxel to generate and update the dynamic conversion coefficient.

[0016] Through the above steps, the expansion conversion ratio of the soil is calculated based on the actual excavation volume and the loading volume. The exponential smoothing algorithm is used to reduce the random error caused by a single measurement, so that the dynamic conversion coefficient can objectively reflect the actual volume expansion characteristics of the soil in the target water area.

[0017] Furthermore, the step of calculating the mass parameters of the residual mud at the bilge bottom using the consumption model to generate a mud viscosity label and writing it into the three-dimensional voxel specifically includes: The hydrostatic curve function is called to convert the unloaded steady-state draft into the total displacement parameter of the return voyage. The dynamic tare weight parameter calculated by the ship's fuel and fresh water consumption model is subtracted to solve the mass parameter of the residual mud at the bottom of the hull. When it is determined that the actual received wet weight load is greater than the preset zero threshold, the mass parameter of the residual mud at the bottom of the hull is divided by the actual received wet weight load to generate the mud viscosity label. The time series exponential smoothing weighted algorithm is introduced to fuse and iterate with the historical value of the mud viscosity label, and the data inside the three-dimensional space voxel of the excavation source is overwritten in reverse.

[0018] By taking the above steps and combining the ship's fuel and fresh water consumption, the weight of the residual mud at the bottom of the cargo hold that has not been unloaded is calculated. This generates a viscous label that reflects the mud adhesion characteristics of the local area, providing data reference for multi-ship scheduling and mud dumping cycle assessment.

[0019] Furthermore, the step of using a neural network to predict the optimal loading wet weight threshold based on the dynamic conversion factor and the sediment viscosity label to perform shutdown control specifically includes: The dynamic conversion coefficient and bottom mud viscosity label of the three-dimensional space voxels of the initial target excavation are extracted, and combined with the dynamic tare weight parameters of the mud barge to be loaded and transported, the maximum physical volume parameters of the cargo hold, the maximum safe total displacement parameters, the mass parameters of the residual mud at the bottom of the previous voyage, and the average volumetric flow rate characteristic parameters of the mud discharge pipeline to construct a real-time input feature vector; the real-time input feature vector is input into a multi-layer feedforward neural network model to output the optimal loading wet weight threshold, and a full-load shutdown control command is issued when the predicted final loading wet weight reaches the optimal loading wet weight threshold.

[0020] Through the above steps, the neural network is used to process geological features, ship structural parameters and pipeline operating parameters to predict the optimal loading weight threshold that meets the current construction conditions, thus avoiding the waste of transport capacity due to soil expansion causing overflow or insufficient loading.

[0021] Furthermore, the processing server is also configured as follows: The three-dimensional spatial voxel set of the excavation source is extracted and the node values ​​are overwritten in the global three-dimensional geological model database. The known sample nodes updated by physical excavation are used as the data source. The spatial inverse distance weight interpolation algorithm is called to perform the deduction calculation and numerical filling of the dynamic conversion coefficient and the mud viscosity label on the unknown adjacent voxels within the surrounding preset radius. The multi-dimensional parameter matrix of the global three-dimensional geological model database is extracted to generate a digital twin stratigraphic color mapping image for visualization rendering.

[0022] Through the above steps, the geological feature data obtained from single-point excavation is spatially interpolated to surrounding unconstructed voxels to fill the data gaps caused by the limited number of sampling points, ensuring the integrity of the three-dimensional geological model database and the visualization effect.

[0023] Furthermore, before generating the scheduling sequence using the genetic algorithm, the processing server also performs the following: The theoretical loading time for a single voyage is calculated by dividing the optimal loading wet weight threshold by the product of the historical average mud density parameter and the historical average instantaneous volumetric flow rate parameter of the mud discharge pipeline; the corrected average mud dumping time is calculated by introducing the bottom mud viscosity label; and the theoretical loading time, the round-trip average voyage time, and the corrected average mud dumping time are linearly summed to generate the theoretical turnover cycle parameters for each transport mud barge for a single voyage.

[0024] Through the above steps, the theoretical loading time is calculated based on the predicted loading threshold and pipeline flow rate. The sediment viscosity label is then introduced to correct the mud dumping time, thereby obtaining more accurate single-voyage turnover cycle parameters for each transport mud barge.

[0025] Furthermore, the generation of scheduling sequences using genetic algorithms specifically includes: Based on the theoretical turnover cycle parameters of each transport barge for a single voyage, a collaborative scheduling cost function is constructed by combining the estimated idle waiting time cost of the dredger with the estimated idling fuel consumption cost of each transport barge waiting in line. A genetic algorithm is then used to encode the individual genes of each transport barge using the scheduling sequence of each transport barge. In the solution space, a crossover and mutation iterative search operation is performed based on the collaborative scheduling cost function to extract the global optimal solution. The global optimal solution is then issued as the scheduling sequence for physical traffic flow control.

[0026] Through the above steps, a cost function is constructed with the goal of reducing the idle waiting time of dredgers and the idling energy consumption of transport barges. A genetic algorithm is then used to search for the scheduling sequence with the lowest overall cost, thereby achieving traffic control under multi-vessel collaborative operation.

[0027] This invention provides a waterway engineering earthwork monitoring and management system based on big data analysis. It has the following beneficial effects: 1. This invention aligns the timestamps of instantaneous volumetric flow rate and density data of mud by calculating the time offset compensation factor of hydrodynamics, and calculates the actual received wet weight load and total dry mass by combining the ship's draft. This eliminates the fluid transmission hysteresis error caused by the physical length of the mud discharge pipeline, solves the problem of time mismatch between the sensor measurement sequence and the actual loading status of the transport mud barge, and improves the accuracy of the physical conversion ratio of a single ship and the calculation of the actual excavated earthwork volume.

[0028] 2. This invention calculates the mass of residual mud in the bilge using a ship consumption model during the empty return phase, generates a mud viscosity label, and writes it in reverse along with a dynamic conversion factor into the corresponding three-dimensional space voxel. Then, it uses a neural network to predict the optimal loading wet weight threshold under the current operating coordinates, so that the shutdown control command can be dynamically adjusted according to the actual expansion rate and adhesion characteristics of the mud in the construction water area, avoiding the waste of transport capacity caused by mud overflow or failure to reach the upper limit of load due to the use of fixed empirical thresholds in single-ship operations.

[0029] 3. This invention incorporates the predicted loading wet weight threshold and the mud viscosity label into the calculation of the theoretical turnover cycle of a single voyage of the transport mud barge. It also constructs a collaborative scheduling cost function based on the idle waiting time of the dredger and the idling fuel consumption of the transport mud barge while waiting in line. The invention uses a genetic algorithm to iteratively extract the global optimal solution as the scheduling sequence, which overcomes the limitation of traditional multi-ship scheduling that relies on fixed time intervals for dispatching ships and reduces the overall equipment energy consumption and idle rate under collaborative operation conditions. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the hardware architecture and multi-terminal data interaction topology of the system of the present invention; Figure 2 This is a schematic diagram of the overall implementation process of the present invention; Figure 3 This is a schematic diagram of the topology of the dynamic conversion factor and spatial volume derivation solution logic of the present invention; Figure 4 This is a schematic diagram of the logic topology for dynamic tare weight tracking and residual mass overwriting in this invention; Figure 5 This is a schematic diagram illustrating the evolution of sliding filtering and dynamic tare weight step compensation for no-load draft data in this invention. Figure 6This is a schematic diagram illustrating the distribution of discrete space voxel inverse distance weight interpolation derivation according to the present invention; Figure 7 This is a schematic diagram of the digital twin surface mapping of three-dimensional geological properties and sediment viscosity according to the present invention; Figure 8 This is a schematic diagram illustrating the temporal correlation between the instantaneous flow rate and the cumulative load of the sludge discharge pipeline of the present invention; Figure 9 This is a schematic diagram of the convergence curve of the training error of the multilayer feedforward neural network model of the present invention; Figure 10 This is a schematic diagram of the time sequence of the multi-ship cooperative scheduling state based on the genetic algorithm of the present invention.

[0031] Among them, 10 is the data acquisition layer; 11 is the dredger terminal; 12 is the mud barge terminal; 13 is the environmental access node; and 20 is the processing server. Detailed Implementation

[0032] 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.

[0033] See attached document Figure 1 The present invention provides a waterway engineering earthwork monitoring and management system based on big data analysis, which includes a data acquisition layer 10 and a processing server 20.

[0034] The data acquisition layer 10 includes a dredger terminal 11, a mud barge terminal 12, and an environmental access node 13.

[0035] The dredger terminal 11 is deployed on the construction dredger and establishes a communication connection with the shipboard programmable logic controller. The dredger terminal 11 collects the instantaneous volumetric flow rate parameters and mud density parameters of the mud discharge pipeline through the electromagnetic flow meter and Coriolis density meter installed on the mud discharge pipeline, and reads the three-dimensional spatial coordinates of the cutter head output by the differential global positioning system receiver.

[0036] The mud barge terminal 12 is deployed in each transport mud barge and is connected to the ship's onboard hydrostatic draft sensor array, automatic identification system, inertial navigation module, shipboard compartment level sensor and engine control unit. The mud barge terminal 12 collects the ship's draft depth parameters, speed parameters, acceleration parameters, fuel and fresh water level parameters, cargo hold mud level parameters and engine operating condition parameters.

[0037] Environmental access node 13 connects to the port hydrological forecasting server via a network interface to extract time-to-tidal water level sequence mapping data for the target operating area.

[0038] The processing server 20 establishes a two-way data transmission connection with the dredger terminal 11, the mud barge terminal 12, and the environmental access node 13 through a wireless communication network, stores the basic characteristic functions of the associated physical hardware, and runs control algorithm instructions.

[0039] See attached document Figure 2 This invention provides a method for monitoring and managing earthwork in waterway engineering based on big data analysis, comprising the following steps: S100 analyzes the three-dimensional design model and geological survey data of the water area to be excavated, discretizes the water area to be excavated into multiple three-dimensional spatial voxels in the data layer, generates the data structure of each three-dimensional spatial voxel, and records the factory empty tare weight parameters and hydrostatic curve function of the transport mud barge, the environmental water density parameters of the construction water area and the reference density parameters of the mud and sand solid particles. S200, obtains flow and density parameters through dredger terminal 11, introduces a fluid dynamic time offset compensation factor based on the physical length of the sludge discharge pipeline, and aligns the data timestamp from the sensor position to the sludge discharge port. S300, combined with the effective draft depth parameters within the relatively static steady-state window extracted by the mud barge terminal 12 and the instantaneous volumetric flow rate parameters and mud density parameters within the effective fluid data subset, calculates the actual total dry matter mass discharged into the cargo hold of the transport mud barge for a single voyage and the actual received wet weight load. S400 calculates the equivalent underwater original volume based on the actual total dry matter mass and the original dry density parameters of the current three-dimensional space voxels, and updates the dynamic conversion factor within the current three-dimensional space voxels. S500, during the mud dumping and return phase of the transport barge, extracts the empty draft data obtained by the barge terminal 12, calls the ship's fuel and fresh water consumption model to calculate the mass parameters of the residual mud at the bottom of the hull, and writes the mass parameters of the residual mud at the bottom of the hull back into the three-dimensional space voxel at the excavation source to generate the mud viscosity label. S600 predicts the optimal loading wet weight threshold based on a multi-layer feedforward neural network model and the current three-dimensional space voxel parameters of the excavation, and issues a full-load shutdown control command by comparing the actual received wet weight load of a single voyage in real time during the loading process. S700 extracts the three-dimensional spatial voxel parameters of the current target excavation, performs node numerical overwriting and spatial interpolation on the global three-dimensional geological model database, and generates a digital twin stratigraphic color mapping image which is pushed to the terminal for visualization and persistent storage. S800 calculates the theoretical turnover cycle parameters of each transport barge for a single voyage by combining the current target excavation three-dimensional space voxel parameters, and generates the target scheduling sequence based on the collaborative scheduling cost function using a genetic algorithm and sends it to each terminal to implement physical traffic flow control.

[0040] In this embodiment, step S100 provided by the present invention may include the following steps in specific implementation: S110, the processing server 20 parses the three-dimensional design model and geological survey data of the water area to be excavated, discretizes the water area to be excavated into multiple three-dimensional spatial voxels in the data layer, extracts the three-dimensional design model data of the waterway project to obtain the design boundary and target excavation depth parameters of the water area to be excavated, and performs three-dimensional mesh segmentation on the target excavation water area with preset physical spatial resolution parameters to generate a set of three-dimensional spatial voxels. The preset physical spatial resolution parameters are determined by the engineering design accuracy and can be selected as a spatial scale with a length and width of 10 meters and a height of 1 meter.

[0041] The mathematical expression for a three-dimensional set of voxels is: ; In the formula, A dataset of voxels in three-dimensional space; Each of the corresponding independent three-dimensional spatial elements in the set; To determine the total number of three-dimensional spatial voxels involved in the calculation, the processing server 20 assigns a unique data code to each three-dimensional spatial voxel unit and establishes the corresponding coordinates of the center point of the three-dimensional Cartesian coordinate system. Through the coordinate system, spatial mapping and addressing of the physical construction location to the underlying data structure can be realized.

[0042] S120, the processing server 20, based on the data structure of each generated three-dimensional spatial voxel, parses the geological survey data from the early stage of the project to extract the spatial coordinates of the geological borehole sampling points and their corresponding physical test values ​​of the dry density of the undisturbed soil samples. It then spatially maps and fits the geological borehole sampling data to the three-dimensional spatial voxel grid to assign a baseline value of the dry density of the undisturbed soil to each three-dimensional spatial voxel. During the initialization of the data structure, it establishes two dynamically updated variable fields within the data structure of the three-dimensional spatial voxels: a dynamic conversion coefficient to characterize the conversion ratio of the equivalent underwater undisturbed volume to the physical wet weight, and a sediment viscosity label to characterize the viscosity characteristics of the excavated soil. It assigns static empirical initial values ​​based on the soil type to the dynamic conversion coefficient and initializes the value of the sediment viscosity label to 0. The static empirical initial values ​​are determined according to the empirical wet density conversion parameters of the corresponding soil type in the dredging engineering specifications, with a value range of 1.4 to 2.0 tons per cubic meter.

[0043] In this embodiment, the distribution pattern of geological parameters in the three-dimensional grid can be established by using the inverse distance weighting algorithm or the Kriging spatial interpolation algorithm. The analytical mechanism of the three-dimensional design model and the spatial interpolation processing of geological survey data are well-known technologies in the field.

[0044] S130, the processing server 20 establishes an independent data mapping table for each mud transport barge registered in the system in the underlying database and enters the factory empty tare weight parameters and hydrostatic curve functions. It stores the standard factory empty tare weight parameters marked on the corresponding mud transport barge ship drawings into the data mapping table, imports the hydrostatic characteristic data provided by the ship designer to establish a mapping relationship from draft to total displacement and generates a hydrostatic curve function. This hydrostatic curve function takes the ship's draft as the independent variable and outputs the standard total weight parameters of the water displaced by the ship under the corresponding draft state. The processing server 20 binds the factory-delivered empty tare weight parameters and hydrostatic curve functions of each transport barge to the unique identification code of that transport barge and stores them in the memory database. It establishes a basic conversion benchmark between the physical signals of the draft sensor and the actual weight variables of the ship, and initializes the bilge residual soil mass parameter field of each transport barge from the previous voyage to zero. It also pre-enters the environmental water density parameters of the construction area and the benchmark density parameters of the sediment solid particles in the underlying database as global configuration variables.

[0045] In this embodiment, the hydrostatic curve of the ship can be obtained by using conventional mathematical fitting software according to the specifications of the ship hydrostatic manual, and a polynomial fitting function for discrete data points can be generated. The process of generating the polynomial function is a well-known technology in this field.

[0046] In this embodiment, step S200 provided by the present invention may include the following steps in specific implementation: S210, the processing server 20 establishes communication with the onboard programmable logic controller on the dredging vessel through the dredging vessel terminal 11 to obtain the fluid state parameters during the dredging operation, analyzes and extracts the continuous time series signals collected by the Coriolis density meter and electromagnetic flow meter installed on the dredging pipeline to obtain the instantaneous volumetric flow rate parameters of the slurry inside the dredging pipeline and the slurry density parameters at the corresponding time. In physical engineering, there is an objective time difference in the flow of slurry from the sensor detection point to the dredging outlet, so the delay needs to be calculated. After the processing server 20 determines that the instantaneous volumetric flow rate parameter is greater than the preset lower limit threshold, it calculates the instantaneous fluid velocity of the slurry inside the slurry discharge pipeline by combining the pre-input physical cross-sectional area parameter inside the slurry discharge pipeline. In this embodiment, the value range of the pre-input physical cross-sectional area parameter inside the slurry discharge pipeline is 0.2 to 0.8 square meters. The physical length parameter of the slurry discharge pipeline between the actual installation position of the sensor and the slurry discharge port is extracted and combined with the instantaneous fluid velocity to calculate the physical time of slurry flow inside the slurry discharge pipeline to generate a fluid dynamics time bias compensation factor. If the instantaneous volumetric flow rate parameter is determined to be no greater than the preset lower flow rate threshold, the mud in the pipeline is determined to be in a stagnant state. The system starts the residence timer to accumulate the stagnation time of the fluid in that section, and after the flow rate recovers to be greater than the preset lower flow rate threshold, the accumulated stagnation time is added to the hydrodynamic time offset compensation factor of the corresponding mud entity.

[0047] The mathematical expressions for instantaneous fluid velocity and fluid dynamics time bias compensation factor are as follows: ; ; In the formula, for The flow rate of mud inside the mud discharge pipeline at all times; It is a time variable; for Instantaneous volumetric flow rate measured at the cross-section of the sludge discharge pipeline at any given time; This refers to the physical cross-sectional area inside the sludge discharge pipeline. for Time delay compensation parameters for the transmission of mud within the mud discharge pipeline; The total physical pipeline length from the cutter head to the sludge discharge port; before performing the calculation, the processing server 20 uniformly converts the numerical dimensions of each physical quantity parameter to the standard dimension system under the International System of Units.

[0048] S220, the processing server 20 performs time-series reconstruction of the original sensor recording sequence based on the fluid dynamics time offset compensation factor, performs data timestamp alignment processing from the sensor position to the mud discharge port, obtains the actual loading start time when the transport mud barge docks at the mud discharge port and begins receiving mud, and the actual loading end time when it leaves the mud discharge port after loading. The fluid dynamics time offset compensation factor at the corresponding moment is subtracted from the absolute loading time interval formed by the actual loading start time and the actual loading end time to calculate the historical measurement time interval when the mud substance falling into the cargo hold of the mud barge flows through the sensor detection point. The instantaneous volumetric flow rate parameter and mud density parameter within the historical measurement time interval are extracted to form an effective fluid data subset. Timestamp alignment ensures that the recorded fluid parameters are consistent with the actual substance falling into the cargo hold, eliminating deviations caused by pipeline residues.

[0049] In this embodiment, the serial binary numerical data stream output by the hardware register node of the underlying fluid sensor can be directly read by combining the industrial field communication bus protocol parsing module inside the shipborne programmable logic controller. The original high-frequency electrical signal output by the underlying sensor is digitally filtered and smoothed using a low-pass digital filter algorithm to generate floating-point time-series data features that participate in the top-level algebraic operation logic. By setting a cyclic sampling period with fixed frequency parameters inside the underlying embedded system, the discretization resampling and data persistence operation of the continuous fluid state signal feature sequence is realized. The communication protocol parsing and extraction method and digital filtering and cleaning logic for the underlying hardware communication signal of industrial instruments are well known technologies in the field.

[0050] In this embodiment, step S300 provided by the present invention may include the following steps in specific implementation: S310, the processing server 20 obtains the real-time motion status data of the transport barge through the barge terminal 12, extracts the continuous acceleration parameters and speed parameters output by the inertial navigation module and the automatic identification system, and extracts the continuous draft data output by the shipborne hydrostatic draft sensor array. In the water construction environment, environmental waves and ship dynamic displacement have objective interference to the hydrostatic sensor signal. The processing server 20 establishes a gating judgment logic based on kinematic threshold to filter the steady-state window when the ship is in a relatively static state. In the adjacent time period before the start point of the absolute loading time interval and the adjacent time period after the end point, the effective smoothed data segments that meet the gating judgment logic are extracted from the continuous draft data. The effective smoothed data segments are processed by the moving average filtering algorithm to eliminate random high-frequency water wave noise, and the spatial mean is calculated for the multi-point smoothed data of the sensor array. The effective pre-loading average draft of the transport barge and the effective post-loading average draft of the transport barge are obtained after the loading operation.

[0051] S320, the processing server 20 calls the system's preset hydrostatic curve function to convert the average draft after effective loading and the average draft before effective loading into the corresponding standard total displacement parameters. Combining the ratio of the pre-entered environmental water density parameters of the construction water area to the standard calibration water density parameters corresponding to the hydrostatic curve function, the standard total displacement parameters are calibrated by density compensation. The difference between the two is calculated to generate the actual received wet weight load for a single voyage. The processing server 20 extracts the instantaneous volumetric flow rate parameters and mud density parameters within the effective fluid data subset. Combining the environmental water density parameters and the baseline density parameters of sediment particles pre-entered during system initialization, the volume concentration of sediment inside the mud is calculated by dividing the difference between the instantaneous mud density and the environmental water density by the difference between the baseline density parameters and the environmental water density. This is then multiplied by the instantaneous volumetric flow rate parameters and the baseline density parameters to generate the instantaneous dry matter mass flow rate. Finally, a definite integral operation is performed along the historical measurement time interval to calculate the actual total dry matter mass discharged into the cargo hold of the transport mud barge for a single voyage. The physical properties of the mixed mud are then mathematically converted to the physical quantity level of pure dry solids.

[0052] The mathematical expressions for the gating criteria and load calculation are as follows: ; ; ; In the formula, The instantaneous acceleration value acquired by the sensor; In this embodiment, the preset high-frequency noise filtering threshold for acceleration is set to a value ranging from 0.05 to 0.2 meters per square second. The instantaneous velocity value acquired by the sensor; The preset high-frequency noise filtering threshold for speed is, in this embodiment, a value ranging from 0.1 to 0.5 meters per second. For the actual wet weight load received in a single voyage; For the pre-calibrated hydrostatic curve function of the ship; This refers to the draft of the transport barge when it is fully loaded. This refers to the draft of the mud barge when it is unloaded. This refers to the total actual dry mass of a single voyage; This is the start time of the actual loading for the current voyage; This is the end time of the current voyage's excavation operations; It is a time variable; The instantaneous volumetric flow rate of the mud after time delay compensation; for Time delay compensation parameters for the transmission of mud within the mud discharge pipeline; The instantaneous mud density after time delay compensation; The current water density in the construction area; The theoretical solid reference density of the bottom sand particles; For time differential variables.

[0053] In this embodiment, a low-pass filtering algorithm combined with a sliding time window can be used to calculate the interval arithmetic mean to obtain smoothed static draft elevation data. The ship kinematic state analysis logic based on inertial navigation data and the basic moving average filtering denoising algorithm are well-known technologies in the field.

[0054] See attached document Figure 3 In specific implementations, step S400 provided by the present invention may include the following steps: S410, the processing server 20 obtains the continuous three-dimensional spatial coordinate sequence of the working cutter head during the historical excavation time corresponding to the time after the physical time compensation of the entire pipeline during the absolute loading time interval through the dredger terminal 11. The physical time of the entire pipeline is calculated by dividing the total length of the dredging pipeline from the working cutter head to the discharge port by the instantaneous fluid velocity. At the same time, it extracts the tidal water level sequence mapping data of the corresponding time obtained by the environmental access node 13, and maps and transforms the continuous three-dimensional spatial coordinate sequence from the absolute geodetic coordinate system to the engineering design reference surface based on the local tidal water level to obtain the absolute elevation coordinates under the unified engineering reference surface. The calibrated coordinate sequence is input into the underlying data structure to perform spatial location mapping matching to locate the three-dimensional spatial voxel set of the excavation source corresponding to the voyage. The original dry density baseline values ​​stored in each voxel in the three-dimensional space voxel set at the excavation source are extracted and the arithmetic mean is calculated to generate the comprehensive original dry density parameter for this voyage. During the underwater excavation stage of the project, when the sediment is stripped from its compacted original state and transformed into a mixed fluid state inside the cargo hold, an objective physical volume expansion phenomenon will occur. Based on the law of conservation of dry matter mass, the processing server 20 divides the actual total mass of dry matter by the comprehensive original dry density parameter to estimate the equivalent underwater original volume of the actual excavation for this voyage.

[0055] The mathematical expression for the equivalent underwater undisturbed volume is: ; In the formula, This represents the equivalent underwater original volume actually excavated during this voyage; This refers to the total actual dry mass of a single voyage; The original dry density reference value is used to bind the three-dimensional voxels of the current operation.

[0056] S420, after determining that the equivalent underwater undisturbed volume is greater than the preset effective volume threshold, the processing server 20 divides the actual received wet weight load of a single voyage by the equivalent underwater undisturbed volume to calculate the single physical conversion ratio measured in the current voyage. It extracts the historical values ​​of the dynamic conversion coefficients currently stored in each voxel in the three-dimensional space voxel set at the excavation source, and introduces a time-series exponential smoothing weighted algorithm to fuse and iterate the single physical conversion ratio with the historical values ​​of the dynamic conversion coefficients of each voxel to generate updated dynamic conversion coefficients that reflect the changes in the current excavation geological and physical state. The newly calculated iteration results are then used to overwrite the original dynamic conversion coefficients within the data structure of each voxel in the three-dimensional space voxel set at the excavation source. If the equivalent underwater undisturbed volume is determined to be less than the preset effective volume threshold, the current fusion iteration operation is skipped and the original dynamic conversion coefficients remain unchanged.

[0057] The mathematical expression for the dynamic conversion factor update is: ; In the formula, This is the updated dynamic conversion factor; In order to control the exponential smoothing weight coefficient of the iteration step size, in this embodiment, the value range of the exponential smoothing weight coefficient of the iteration step size is 0.1 to 0.3. The values ​​are historical values ​​of the dynamic conversion factor before the update; For the actual wet weight load received in a single voyage; This represents the equivalent underwater undisturbed volume actually excavated during this voyage.

[0058] In this embodiment, a spatial octree index structure can be used to construct a fast traversal query mapping logic from real-time three-dimensional Cartesian spatial coordinates to the underlying three-dimensional mesh voxels. The spatial coordinate hash addressing algorithm and the basic time series exponential smoothing weighting logic are well-known technologies in the field.

[0059] See attached document Figure 4 In specific implementations, step S500 provided by the present invention may include the following steps: S510, when the transport barge completes the unloading operation and enters the mud dumping and return stage and arrives at the loading berth, the processing server 20 obtains the real-time empty draft data of the ship after the mud dumping is completed through the barge terminal 12. Combined with the extracted ship acceleration parameters and speed parameters, a kinematic gating judgment logic is established to filter the relatively static state. The moving average filtering algorithm is applied to extract the effective smooth data segment of the steady-state window period when the ship is relatively static after the return arrival to generate the empty steady-state draft. During the operation of the ship, the main engine and auxiliary engine will continuously consume fuel and fresh water, which will cause the ship's basic tare weight to undergo objective dynamic drift. The processing server 20 calls the system's preset ship fuel and fresh water consumption model to calculate the single voyage consumption mass parameter between the previous voyage return measurement node and the current return measurement node. The updated dynamic tare weight parameter of the vessel at the current voyage return node is calculated by subtracting the mass consumption parameter of a single voyage from the dynamic tare weight parameter of the previous voyage recorded in the system for the corresponding transport barge. After determining that the vessel is in a non-navigating berthing state by extracting the vessel's acceleration and speed parameters and continuously monitoring the positive increase of the fuel and fresh water level parameters obtained by the barge terminal 12 that exceeds the preset step change threshold, and determining that the vessel is to perform fuel and fresh water replenishment operation, the updated dynamic tare weight parameter is positively compensated and reset based on the step increase of the fuel and fresh water level parameters obtained by the barge terminal 12 and the refueling mass obtained by converting the pre-entered vessel compartment capacity curve function in the system. When the system initializes the first voyage, the value of the dynamic tare weight parameter of the previous voyage is the sum of the factory empty tare weight parameter and the initial fuel and fresh water mass measured at the initial time of the vessel's arrival. The starting time for calculating the single voyage integral is the initial measurement time of the vessel's arrival operation.

[0060] The processing server 20 calls the hydrostatic curve function to convert the obtained unloaded steady-state draft into the total return displacement parameter. It then combines the ratio of the environmental water density parameter of the construction area pre-entered by the system to the standard calibration water density parameter corresponding to the hydrostatic curve function to perform density compensation calibration on the total return displacement parameter. Finally, it subtracts the dynamic tare weight parameter of the ship at the current voyage return node from the total return displacement parameter to calculate the mass parameter of the residual mud adhering to the bottom of the cargo hold.

[0061] The mathematical expression for the consumption model and residual mass calculation is as follows: ; ; In the formula, The dynamic tare weight parameter of the transported mud barge at the current voyage return point; The empty tare weight parameter for transporting mud barges at the factory; This is the start time of the actual loading for the current voyage; This refers to the absolute moment when the transport barge is at the return measurement node; It is a time variable; This refers to the fuel consumption rate of ship engines. Freshwater consumption rate of ship equipment; For time differential variables; The mass parameters of the residual soil adhering to the bottom of the cargo hold; For the pre-calibrated hydrostatic curve function of the ship; Real-time draft of the mud barge during the empty return phase; This refers to the dynamic tare weight parameters of the transported mud barge at the current voyage return point.

[0062] S520, the processing server 20 extracts the historical spatial coordinate sequence of the cutter head associated with the current transport barge during the loading operation phase and performs reverse spatial indexing to locate the three-dimensional spatial voxel set of the excavation source; In engineering practice, the amount of residual soil at the bottom of the cargo hold objectively reflects the physical viscosity characteristics of the geological components of the excavated underwater. After determining that the actual received wet weight load is greater than the preset zero threshold, the processing server 20 divides the mass parameter of the residual soil at the bottom of the hold by the actual received wet weight load of the voyage to calculate and generate a bottom mud viscosity label characterizing the soil adhesion characteristics. In this embodiment, the preset zero threshold ranges from 5 to 10 tons. The historical values ​​of the bottom mud viscosity labels currently stored in each voxel in the three-dimensional space voxel set of the excavation source are extracted. A time series exponential smoothing weighted algorithm is introduced to fuse and iterate the calculated bottom mud viscosity labels characterizing the soil adhesion characteristics with the historical values ​​of the bottom mud viscosity labels of each voxel. The newly calculated iteration results are written in reverse to the data structure of each voxel in the three-dimensional space voxel set of the excavation source, overwriting the initialization assignment or original value inside. If it is determined that the actual received wet weight load is not greater than the preset zero threshold, the generation and writing operation of the bottom mud viscosity tag is skipped, and the original value inside the data structure of each voxel in the three-dimensional space voxel set of the excavation source remains unchanged.

[0063] The mathematical expression for generating sediment viscosity labels is as follows: ; In the formula, A mud viscosity label used to characterize soil adhesion properties; The mass parameters of the residual soil adhering to the bottom of the cargo hold; This refers to the actual wet weight load received during a single voyage.

[0064] In this embodiment, a polynomial fitting curve can be constructed by combining the ship engine factory bench test log and liquid level sensor data to calibrate the fuel consumption rate function at different speeds. The energy consumption modeling method based on engine operating conditions and the backtracking retrieval logic of the three-dimensional spatial database are well-known technologies in the field.

[0065] See attached document Figure 5 In specific implementations, step S600 provided by the present invention may include the following steps: S610, the processing server 20 builds and deploys a multi-layer feedforward neural network model inside the underlying system. The multi-layer feedforward neural network model consists of a fully connected input layer, a double hidden layer containing 64 and 32 neurons, and a single-node output layer in the network hierarchy. The processing server 20 extracts historical loading records from the historical construction database as a sample dataset for model training. Normalization preprocessing is performed on the sample dataset to map the feature values ​​of each dimension to a dimensionless range of 0 to 1. The input data dimensions of this multi-layer feedforward neural network model cover the dynamic conversion coefficient and mud viscosity label inside the three-dimensional space voxel of the initial target excavation of the corresponding historical voyage, as well as the original dry density benchmark value. The dynamic tare weight parameters of the mud barge to be loaded and transported, the pre-entered maximum physical volume parameters of the cargo hold, the maximum safe total displacement parameters converted based on the hydrostatic curve, the mass parameters of the residual mud at the bottom of the hold of the previous voyage, and the average volumetric flow rate characteristic parameters of the mud discharge pipeline are input simultaneously. The output result is defined as the optimal loading wet weight threshold under the corresponding working condition.

[0066] The processing server 20 combines the time information of the ship's onboard compartment liquid level sensor reaching the preset full-load high-level threshold in the historical records, extracts the actual received wet weight load when the critical state of mud overflow was triggered in the historical voyage, and deducts the preset system delay safety margin as the supervised learning label of the corresponding training sample. The loss function of the multilayer feedforward neural network model is set as the mean square error function. The gradient descent optimization algorithm based on backpropagation is used to iteratively update the weight matrix and bias vector inside the multilayer feedforward neural network model. The loss value calculated by the mean square error function is monitored during the training process until the loss value converges to below the preset error tolerance or reaches the preset maximum number of iterations. The preset error tolerance value range is set to 0.001 to 0.01. The iteration is stopped and the network parameters are frozen to generate the multilayer feedforward neural network model that has completed training.

[0067] S620, in actual construction operations, the composition and physical state of the mud directly affect the loading capacity of the ship's cargo hold. When entering the loading operation stage of a new voyage, the processing server 20 obtains the current starting target excavation three-dimensional space voxel coordinates sent by the dredger terminal 11. Extract the dynamic conversion coefficient, bottom mud viscosity label, and original dry density benchmark value that are updated in real time within the three-dimensional space voxel data structure of the current starting target excavation. Combine these with the dynamic tare weight parameters of the current transport barge registered and scheduled by the system, the maximum physical volume parameters of the cargo hold pre-entered by the barge, the maximum safe total displacement parameters of the barge, the mass parameters of the residual mud at the bottom of the hold updated by the barge in the previous voyage, and the average volumetric flow rate parameters of the sludge discharge pipeline within the preset sliding time window to construct a real-time input feature vector. When the average volumetric flow rate parameter within the preset sliding time window is missing or not greater than the preset lower limit threshold, the rated sludge discharge flow rate parameter configured internally by the system is called to replace it, generating the average volumetric flow rate characteristic parameter of the sludge discharge pipeline and participating in the construction of the feature vector. The real-time input feature vector is normalized according to the normalization standard in the model training stage and then sent to the trained multi-layer feedforward neural network model for forward propagation calculation. The model output result is then denormalized to restore the physical dimensions, thereby outputting the optimal loading wet weight threshold under the current actual excavation geology and equipment matching conditions.

[0068] During the loading process, the processing server 20 extracts the real-time dynamic draft obtained by the hydrostatic draft sensor array through the mud barge terminal 12, performs high-frequency fluctuation elimination processing on the real-time dynamic draft using a sliding low-pass filter algorithm, and converts it into the current dynamic total displacement by combining the environmental water density parameters of the construction area. The initial total displacement obtained by subtracting the effective average draft before loading that was intercepted before the official start of the current voyage loading operation and similarly performing density compensation is then used to continuously calculate the dynamic cumulative loading wet weight of the current voyage. The physical cross-sectional area parameter inside the sludge discharge pipeline is extracted, and the total length parameter of the sludge discharge pipeline from the cutter head to the sludge discharge port and the sludge density parameter at the current moment are multiplied to calculate the mass of sludge retained inside the pipeline. The dynamic cumulative loaded wet weight is summed with the mass of sludge retained inside the pipeline to generate the predicted final loaded wet weight, and this is compared with the output optimal loaded wet weight threshold in real time. The current dynamic total drainage volume is added to the mass of sludge retained inside the pipeline to generate the predicted final total drainage volume. At the same time, the predicted final total drainage volume is compared with the maximum safe total drainage volume calculated by combining the maximum safe draft of the transport mud barge in real time. When it is determined that the predicted final wet weight of the load reaches or exceeds the optimal wet weight threshold, or the predicted final total displacement reaches the maximum safe total displacement of the transport barge, the processing server 20 generates a full-load shutdown control command and sends the command to the dredger terminal 11 via a wireless communication link. The dredger terminal 11 then forwards the command to the onboard programmable logic controller on the dredger via the shipboard industrial fieldbus. The controller controls the cutter head to stop digging and keeps the mud pump running to discharge all the mud slurry retained in the discharge pipeline into the transport barge. Then, the controller drives the three-way reversing valve of the discharge pipeline to perform a mud bypass action or controls the mud pump to reduce its speed to idle, so as to safely terminate the mud loading operation process of the current barge.

[0069] The mathematical expression for the loss function and forward propagation operation is as follows: ; ; In the formula, The loss value is calculated using the mean squared error function; This represents the total number of sample batches input to the feedforward neural network model in a single iteration. This serves as the index variable for the sample sequence during the iterative calculation process. For input number The supervised learning label corresponding to each sample; For the model targeting the first The optimal loading wet weight threshold prediction result for each sample output; This is the optimal wet weight threshold prediction result for a single forward propagation. For non-linear activation functions between neural network layers; These are the weight matrix parameters from the hidden layer to the output layer in the model; These are the parameters of the weight matrix from the input layer to the first hidden layer in the model; It is a real-time input feature vector composed of geological parameters and ship status parameters; The first bias vector parameter of the hidden layer; The second bias vector parameter of the output layer.

[0070] In this embodiment, the backpropagation algorithm can be used to calculate the partial derivatives of the loss function with respect to the parameters of each layer, and the adaptive moment estimation optimizer can be used to perform parameter update actions. The normalization processing logic for multidimensional data and the forward and backward propagation operation rules of the basic feedforward neural network are well known technologies in the art.

[0071] See attached document Figure 6 In specific implementations, step S700 provided by the present invention may include the following steps: S710, after the transport barge corresponding to the current voyage has completed unloading and returned, and after updating the dynamic conversion coefficient and bottom mud viscosity label, the processing server 20 extracts the coordinates of each voxel in the three-dimensional space voxel set of the excavation source and the corresponding bound dynamic conversion coefficient and bottom mud viscosity label, and writes the above discrete voxel feature parameters into the global three-dimensional geological model database to execute node numerical overwrite. In actual engineering, the sampling points formed by the excavation trajectory are scattered and the surrounding unexcavated areas lack direct physical observation data. The processing server 20 extracts unknown adjacent voxels within a preset radius around the three-dimensional space voxel set of the excavation source that have not yet been physically excavated and updated. The voxels that have been numerically overwritten within the three-dimensional space voxel set of the excavation source and other historical known sample nodes within the preset radius that have been physically excavated and updated are used together as the data source of known sample nodes. In this embodiment, the preset radius range is 20 to 50 meters. After determining that the number of unknown adjacent voxels that have not yet been physically excavated and updated is greater than zero, a zero-deviation constant is introduced into the distance attenuation calculation to avoid algebraic singularities caused by spatial node overlap. The spatial inverse distance weighted interpolation algorithm is then called to perform dynamic conversion coefficients and sediment viscosity labels on the unknown adjacent voxels for deduction calculation and numerical filling. If the number of unknown adjacent voxels that have not yet been physically excavated and updated is equal to zero, the deduction calculation and numerical filling operation of the spatial inverse distance weighted interpolation algorithm are skipped.

[0072] The mathematical expression for the derivation of geological attribute parameters is as follows: ; ; In the formula, The predicted geological attribute values ​​at unknown spatial nodes are derived through interpolation. The total number of known sample nodes participating in the spatial interpolation calculation; This is the index variable used during the traversal and calculation of known sample nodes; For unknown space nodes up to the first The three-dimensional geometric spatial distance between known sample nodes; To avoid algebraic singularities, the value range of the zero-determination deviation constant is set to 0.0001 to 0.001 meters; The power-law parameter used to control the rate of spatial distance decay; For the first The actual geological attribute values ​​recorded within each known sample node. For unknown space nodes up to the first The three-dimensional geometric spatial distance between known sample nodes; These are the three spatial coordinate components of an unknown spatial node in a three-dimensional Cartesian coordinate system. For the first The three spatial coordinate components of a known sample node in a three-dimensional Cartesian coordinate system.

[0073] In the S720 system application, it is necessary to synchronously feed back the updated background data to the construction operation front end with objective physical characteristics. The processing server 20 extracts the multi-dimensional parameter matrix from the updated global three-dimensional geological model database to generate digital twin stratigraphic color mapping images based on different geological attribute dimensions. The digital twin stratigraphic color mapping image, the actual total dry matter mass of the current voyage, and the actual received wet weight load are packaged and encapsulated into standard message protocol data. The standard message protocol data is synchronously distributed and pushed to the display modules of the dredger terminal 11 and the mud barge terminal 12 through the wireless communication link to perform visualization rendering. The processing server 20 drives the underlying database engine to perform persistent storage of the complete construction feature data and model parameters of the current voyage.

[0074] In this embodiment, a color lookup table algorithm can be used to convert discrete geological attribute floating-point values ​​into a continuous red, green, and blue color space pixel matrix to perform image rendering. The data persistence disk operation logic based on relational database and the basic color mapping rendering algorithm mechanism are well-known technologies in the field.

[0075] In this embodiment, step S800 provided by the present invention may include the following steps in specific implementation: After completing the node value overwrite in the global three-dimensional geological model database, the S810 processing server 20 starts the multi-vessel collaborative scheduling logic to extract the historical voyage status data of all registered transport barges in the current construction sea area. It analyzes the average round-trip voyage time and average mud dumping time of each transport barge in different geological attribute intervals. If it is determined that the transport barge lacks corresponding historical voyage status data, it calls the default design time parameters preset in the system to replace the average round-trip voyage time and average mud dumping time. In the water construction, there are significant physical differences in the unloading efficiency of transport barges with different cabin structures for high-viscosity bottom mud. Combined with the bottom mud viscosity label bound to the three-dimensional space voxel of the current target excavation and the original dry density benchmark value, the subsequent time compensation calculation is performed. The processing server 20 determines whether the historical average instantaneous volumetric flow rate of the sludge discharge pipeline is greater than the preset flow zero threshold. In this embodiment, the preset flow zero threshold ranges from 0.1 to 0.5 cubic meters per second. If it is not greater than the preset flow zero threshold, the system calls the rated sludge discharge flow rate parameter configured in the system to replace the historical average instantaneous volumetric flow rate to generate the processed historical average instantaneous volumetric flow rate parameter of the sludge discharge pipeline. If it is greater than the preset flow rate, the historical average instantaneous volumetric flow rate of the sludge discharge pipeline is directly used as the processed historical average instantaneous volumetric flow rate parameter of the sludge discharge pipeline. The processing server 20 calls the multi-layer feedforward neural network model trained in step S600. For each transport barge participating in the scheduling, the dynamic conversion coefficient and bottom mud viscosity label bound inside the three-dimensional space voxel of the current working area corresponding to the starting target excavation, as well as the original dry density benchmark value, the current dynamic tare weight parameter of each transport barge, the maximum physical volume parameter of the cargo hold corresponding to each barge, the maximum safe total displacement parameter, the mass parameter of the residual soil at the bottom of the hold corresponding to each barge of the previous voyage, and the historical average instantaneous volumetric flow rate parameter of the treated sludge discharge pipeline are spliced ​​together to form a feature vector. After normalizing the feature vector according to the normalization standard in the model training stage, forward propagation calculation is performed, and the calculation result is inversely normalized to output the optimal loading wet weight threshold prediction result for each transport barge. In conjunction with the historical average mud density of the corresponding excavation area stored in the system and the historical average instantaneous volumetric flow rate of the processed mud discharge pipeline, when it is determined that the historical average mud density of the corresponding excavation area is missing, the system calls the preset empirical mud density parameter to replace the historical average mud density to generate the processed historical average mud density parameter. The optimal loading wet weight threshold prediction result is divided by the product of the processed historical average mud density parameter and the processed historical average instantaneous volumetric flow rate of the processed mud discharge pipeline to calculate the theoretical loading time of each transport mud barge for the current target excavation three-dimensional space voxel. Subsequently, the processing server 20 introduces a sediment viscosity label to perform a delay-weighted correction on the average mud dumping time. Specifically, the average mud dumping time is multiplied by a sum of the sediment viscosity label value to obtain the corrected average mud dumping time. The calculated theoretical loading time, the average round-trip voyage time, and the corrected average mud dumping time are linearly summed to generate the theoretical turnover cycle parameters for each transport mud barge under the geological conditions.

[0076] The mathematical expression for the theoretical turnaround time calculation for a single voyage is as follows: ; In the formula, For the first Theoretical turnaround cycle parameters for a single voyage of a transport barge; The index number in the set of transport barges participating in the scheduling; For the multilayer feedforward neural network model targeting the first Prediction results of the optimal loading wet weight threshold for the mud barge; This represents the historical average instantaneous volumetric flow rate of the sludge discharge pipeline within the current construction section. The historical average mud density of the material transported within the mud discharge pipeline; For the first Average sailing time for a mud barge transporting mud in different geological ranges; For the first The average time for a transport barge to dump mud in the dumping area; A mud viscosity label used to characterize soil adhesion properties.

[0077] S820, the processing server 20 constructs a collaborative scheduling cost function based on the theoretical turnaround cycle parameters of each transport barge for a single voyage, extracts the current physical location coordinates and real-time operating status of each transport barge, and calculates the absolute time for each transport barge to arrive at the dredger loading berth by combining the current operating status and the corresponding time consumption items within the theoretical turnaround cycle parameters for a single voyage. The estimated idle waiting time cost of the dredger and the estimated idling fuel consumption cost of each transport barge while queuing are input into the collaborative scheduling cost function as cost components. The initial population size of the genetic algorithm is set to 50 to 200 individuals and the crossover probability parameter is set in the range of 0.6 to 0.9. Using a genetic algorithm, the scheduling sequence of each transport barge is used as the gene encoding of the population. Within a preset solution space, multiple rounds of crossover and mutation iterative search operations are performed, combining mutation probability parameters in the range of 0.01 to 0.1, until the preset maximum number of generations is reached or the population fitness meets the convergence condition. The iteration stops when this condition is met. The processing server 20 extracts the global optimal solution output by the genetic algorithm as the target scheduling sequence. The target scheduling sequence is encapsulated into control signals containing berth allocation information and arrival time nodes and sent to the dredger terminal 11 and each barge terminal 12 via wireless communication links to perform physical traffic flow control at the construction site.

[0078] The mathematical expression of the cost function for coordinated scheduling is as follows: ; In the formula, The total cost value output by the coordinated scheduling cost function during iteration; Mathematical operators for extracting the minimum objective within the solution space; This represents the total number of mud barges participating in the dispatch within the current construction sea area. The index number in the set of transport barges participating in the scheduling; To measure the time cost weighting coefficient for the cost of dredging vessel idleness and downtime, in this embodiment, the value range of the time cost weighting coefficient for measuring the cost of dredging vessel idleness and downtime is 0.4 to 0.6; A mathematical operator for extracting the largest numerical element within parentheses; For the first The estimated absolute time for the transport barge to arrive at the dredger loading berth; The absolute time during which a dredger completes the loading of the preceding transport barge according to the shift schedule and enters an idle state; In this embodiment, the fuel cost weighting coefficient for measuring idling energy consumption cost ranges from 0.4 to 0.6. For the first Fuel consumption rate of transport barges during idling waiting period; For the first The estimated waiting time for the transport barges to enter the berth is as follows: queuing outside the berth.

[0079] In this embodiment, a roulette wheel selection strategy combined with a real number encoding mechanism can be used to construct the underlying population evolution rules of the genetic algorithm. The optimization solution logic based on the objective function and the signaling distribution mechanism of the basic wireless communication network are well-known technologies in the field.

[0080] Application Examples: To better understand the technical solution of the present invention, the following uses the application scenario of the second phase expansion project of the main channel of a large port as an example. This embodiment selects the second phase expansion project of the main channel of a large port as the background. The construction equipment includes a large cutter suction dredger and three mud barges of the same type. The project area includes a mixture of cohesive soil and silt.

[0081] The processing server 20 accesses the three-dimensional design map of the waterway and divides the target excavation area into three-dimensional spatial voxels with a length of 10 meters, a width of 10 meters, and a height of 1 meter. Based on the previous geological drilling data, each voxel is assigned a baseline value of 1 ton per cubic meter of original dry density and the viscosity label of the bottom mud is initialized to zero.

[0082] The dredger performs processing operations, causing the slurry to flow in a 120-meter-long discharge pipe. The processing server 20 reads data from the electromagnetic flowmeter and density meter and calculates that there is a 30-second physical delay from the detection point to the discharge port when the flow rate is 4 meters per second. The processing server 20 shifts the sensor timestamp backward by 30 seconds and aligns it with the actual time when the slurry falls into the cargo hold of the first mud barge.

[0083] After the first mud barge is loaded, the processing server 20 intercepts the draft in a stable state and converts it into the total displacement. The processing server 20 subtracts the dynamic tare weight of the ship to obtain the actual wet weight load of 2,200 tons received in a single run. The processing server 20 calculates the equivalent underwater original volume of the mud based on the total dry mass of the aligned flow meter and updates the dynamic conversion factor of the corresponding excavation volume element.

[0084] After the first mud barge sails to the mud dumping area and dumps mud, it returns to port. The processing server 20 reads the empty steady draft data during the return journey. The processing server 20 deducts the weight of fuel and fresh water consumed during the round trip through the fuel and fresh water level gauges and calculates the mass parameters of the 18 tons of residual mud at the bottom of the cargo hold. The processing server 20 divides the mass parameters of the residual mud at the bottom of the cargo hold by the actual received wet weight load and writes them back into the corresponding voxel to generate a mud viscosity tag.

[0085] When the dredger is preparing to load the second mud barge, the processing server 20 extracts the forward voxel data. The multi-layer feedforward neural network model combines the average volumetric flow rate parameter and the maximum safe total displacement parameter to output the optimal loading wet weight threshold of 2,100 tons. When the processing server 20 predicts that the final loading wet weight will reach 2,100 tons, it issues a full-load shutdown control command to control the dredger to perform mud bypass action and reduce the mud pump speed to idle.

[0086] The processing server 20 uses a spatial interpolation algorithm to extrapolate and calculate the dynamic conversion coefficient and the bottom mud viscosity label to the unknown neighboring voxels that have not been physically excavated and updated, and renders the stratum color mapping image on the terminal. The processing server 20 combines the bottom mud viscosity label to increase the corrected average mud dumping time of the first mud barge, and regenerates the target scheduling sequence through a genetic algorithm and sends it to the third mud barge to control it to enter the standby state.

[0087] A 30-day comparative experiment was conducted during the waterway expansion project. The first 15 days used the traditional mode as the control group, and the last 15 days were connected to the system of this invention as the experimental group.

[0088] The test results are shown in the table below.

[0089] Comparison table of application effects between traditional mode and the system of this invention: The experimental data in the table above proves that the system provided by this invention solves the data distortion problem caused by pipeline delays, changes in ship fuel consumption, and sediment adhesion. The neural network anti-overflow mechanism and the genetic algorithm scheduling model improve the loading efficiency of a single ship and the collaborative operation efficiency of multiple ships, while reducing the overall construction energy consumption and cost.

[0090] Appendix Figure 7 It presents a three-dimensional gridded block map, with the horizontal and vertical axes representing three-dimensional geodetic coordinates. The seabed model is cut into a regular grid matrix. Excavated areas are marked as outlines, and unexcavated areas are characterized by warm and cool color gradations representing the magnitude of the sediment viscosity labels stored inside the spatial voxels. Red areas correspond to high-viscosity geology, and blue areas correspond to soft geology.

[0091] Appendix Figure 8This is a two-dimensional time series curve with two vertical axes. The horizontal axis represents absolute time. The first curve represents the instantaneous flow fluctuation characteristics of the electromagnetic flowmeter in its original reading. The second curve represents the total weight change characteristics of the draft sensor after smoothing. The third curve represents the instantaneous flow characteristics after being reconstructed by shifting the time offset compensation factor. The fluctuation nodes of the reconstructed flow curve are vertically aligned with the starting node of the increase in the weight of the mud barge on the time axis.

[0092] Appendix Figure 9 This is a graph showing the error decay trend during the supervised learning process of the model. The horizontal axis represents the number of training generations of the feedforward neural network model, and the vertical axis represents the mean square error between the predicted output and the actual loading limit. In the graph, the broken line is at a high position in the initial stage and is accompanied by oscillations. As the horizontal axis extends to the right, the broken line slides down and tends to flatten in the middle and later stages.

[0093] Appendix Figure 10 The horizontal bar timeline chart shows the construction timeline on the horizontal axis and the dredger and transport barge numbers on the vertical axis. Green bars represent dredging towards the discharge port, orange bars represent loading operations, purple bars represent heavy-load navigation and dumping, and gray bars represent empty return. The dredger working status bars are closely connected, and the arrival nodes of each transport barge are connected end to end.

[0094] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A waterway engineering earthwork monitoring and management system based on big data analysis, characterized in that, include: The data acquisition layer collects instantaneous volumetric flow rate parameters, mud density parameters, draft parameters, and tidal water level parameters of the sludge discharge pipeline; The processing server, connected to the data acquisition layer, is configured to discretize the water area to be excavated into three-dimensional spatial voxels. Computational fluid dynamics time bias compensation factor aligns data timestamps; The actual received wet weight load is calculated based on the draft depth parameter, and the actual total dry matter mass is calculated using the aligned instantaneous volumetric flow rate parameter and mud density parameter to update the dynamic conversion factor. The mass parameters of residual mud at the bottom of the hull are calculated by combining the consumption model to generate a mud viscosity label and write it into the three-dimensional space voxel; based on the dynamic conversion coefficient and the mud viscosity label, the optimal loading wet weight threshold is predicted by a neural network to execute shutdown control, and a scheduling sequence is generated by a genetic algorithm.

2. The waterway engineering earthwork monitoring and management system based on big data analysis according to claim 1, characterized in that, Discretizing the water area to be excavated into three-dimensional spatial voxels specifically includes: Extract the three-dimensional design model data of the waterway project to obtain the design boundary and target excavation depth parameters, and perform three-dimensional mesh segmentation with preset physical space resolution parameters to generate a three-dimensional spatial voxel set; The spatial coordinates of geological borehole sampling points and their corresponding undisturbed soil dry density physical test values ​​are extracted from the geological survey data. The undisturbed dry density benchmark values ​​are then assigned to each three-dimensional spatial voxel through spatial mapping fitting.

3. The waterway engineering earthwork monitoring and management system based on big data analysis according to claim 1, characterized in that, The computational fluid dynamics time bias compensation factor aligns the data timestamps, specifically including: When the instantaneous volumetric flow rate parameter is greater than the preset lower limit threshold, the instantaneous fluid velocity of the slurry inside the slurry discharge pipeline is calculated by combining the physical cross-sectional area parameter inside the slurry discharge pipeline. The fluid dynamics time offset compensation factor is generated by dividing the physical length parameter of the sludge discharge pipeline by the instantaneous fluid velocity. By subtracting the fluid dynamics time bias compensation factor from the actual start time and actual end time of mud loading, the historical measurement time interval of the corresponding mud entity flowing through the sensor detection point can be calculated.

4. The waterway engineering earthwork monitoring and management system based on big data analysis according to claim 1, characterized in that, The calculation of the actual received wet weight load based on the draft depth parameter, and the calculation of the actual total dry matter mass using the aligned instantaneous volumetric flow rate parameter and mud density parameter, specifically includes: Effective smooth data segments are extracted from the draft depth parameters, and the hydrostatic curve function is called to convert them into standard total displacement parameters. After performing density compensation calibration, the difference between the standard total displacement parameters before and after loading is calculated to generate the actual received wet weight load. The volume concentration is calculated by dividing the difference between the mud density parameter and the ambient water density by the difference between the reference density parameter and the ambient water density. This concentration is then multiplied by the instantaneous volumetric flow rate parameter and the reference density parameter, and a definite integral is performed along the historical measurement time interval to calculate the actual total dry matter mass.

5. The waterway engineering earthwork monitoring and management system based on big data analysis according to claim 4, characterized in that, The updated dynamic conversion factor specifically includes: The equivalent underwater undisturbed volume of a single voyage is calculated by dividing the total actual dry mass of the material by the comprehensive undisturbed dry density parameter of the three-dimensional space voxels at the excavation source. When it is determined that the equivalent underwater undisturbed volume is greater than the preset effective volume threshold, the actual received wet weight load is divided by the equivalent underwater undisturbed volume to calculate the single physical conversion ratio. A time-series exponential smoothing weighted algorithm is introduced to fuse the single physical conversion ratio with the historical value of the dynamic conversion coefficient of the corresponding voxel to generate and update the dynamic conversion coefficient.

6. The waterway engineering earthwork monitoring and management system based on big data analysis according to claim 5, characterized in that, The process of calculating the mass parameters of the residual mud at the bilge bottom using the consumption model, generating a mud viscosity label, and writing it into the three-dimensional voxel specifically includes: The hydrostatic curve function is called to convert the unloaded steady-state draft into the total displacement parameter for return voyage. The dynamic tare weight parameter calculated by the ship's fuel and fresh water consumption model is subtracted to solve the mass parameter of the residual mud in the bilge. When it is determined that the actual received wet weight load is greater than the preset zero threshold, the mass parameter of the residual soil at the bottom of the hull is divided by the actual received wet weight load to generate a mud viscosity label. A time series exponential smoothing weighted algorithm is introduced to fuse and iterate with the historical value of the mud viscosity label, and the data inside the three-dimensional space voxel of the excavation source is overwritten in reverse.

7. The waterway engineering earthwork monitoring and management system based on big data analysis according to claim 1, characterized in that, The step of using a neural network to predict the optimal loading wet weight threshold based on the dynamic conversion factor and the sediment viscosity label to perform shutdown control specifically includes: The dynamic conversion coefficient and bottom mud viscosity label of the three-dimensional space voxel of the initial target excavation are extracted, and combined with the dynamic tare weight parameter of the mud barge to be loaded and transported, the maximum physical volume parameter of the cargo hold, the maximum safe total displacement parameter, the mass parameter of the residual mud at the bottom of the hold of the previous voyage, and the average volumetric flow rate characteristic parameter of the mud discharge pipeline to construct a real-time input feature vector. The real-time input feature vector is input into a multi-layer feedforward neural network model to output the optimal loading wet weight threshold, and a full-load shutdown control command is issued when the predicted final loading wet weight reaches the optimal loading wet weight threshold.

8. The waterway engineering earthwork monitoring and management system based on big data analysis according to claim 1, characterized in that, The processing server is also configured to: The three-dimensional spatial voxel set of the excavation source is extracted and the node values ​​are overwritten in the global three-dimensional geological model database. The known sample nodes updated by physical excavation are used as the data source. The spatial inverse distance weight interpolation algorithm is called to perform the deduction calculation and numerical filling of the dynamic conversion coefficient and the mud viscosity label on the unknown adjacent voxels within the surrounding preset radius. The multi-dimensional parameter matrix of the global three-dimensional geological model database is extracted to generate a digital twin stratigraphic color mapping image for visualization rendering.

9. The waterway engineering earthwork monitoring and management system based on big data analysis according to claim 1, characterized in that, Before generating the scheduling sequence using a genetic algorithm, the processing server also performs the following: The theoretical loading time for a single voyage is calculated by dividing the optimal loading wet weight threshold by the product of the historical average mud density parameter and the historical average instantaneous volumetric flow rate parameter of the sludge discharge pipeline. The corrected average mud dumping time is calculated by introducing the bottom mud viscosity label, and the theoretical loading time, the round-trip average voyage time and the corrected average mud dumping time are linearly summed to generate the single-voyage theoretical turnover cycle parameters for each transport mud barge.

10. The waterway engineering earthwork monitoring and management system based on big data analysis according to claim 9, characterized in that, The generation of scheduling sequences using a genetic algorithm specifically includes: Based on the theoretical turnover cycle parameters of each transport barge for a single voyage, a collaborative scheduling cost function is constructed by combining the estimated idle waiting time cost of the dredger with the estimated idling fuel consumption cost of each transport barge waiting in line. Using a genetic algorithm, the scheduling sequence of each transport barge is used as the gene encoding of the individual population. In the solution space, a crossover and mutation iterative search operation is performed based on the cooperative scheduling cost function to extract the global optimal solution. The global optimal solution is then issued as the scheduling sequence to implement physical traffic flow control.

Images