A control method and system for a cutter suction dredger

Abstract

The application discloses a cutter suction dredger control method and system, belonging to the technical field of dredger control. The method comprises the following steps: dividing the control system into a master control system and multiple sub-control units to form mutually independent sub-control areas, and arranging working data sampling points on the sludge discharge pipes in adjacent sub-control areas; training a regional control model and a global control model for each sub-control area by using a deep learning algorithm combined with working condition constraints; taking the sampling points as regulation and control nodes, generating basic operation parameters according to regulation and control parameters and fusing the global and regional control models; predicting the output data change trend of the regulation and control nodes by a model predictive control algorithm, and adjusting the output data of the sub-control areas in advance in a reverse direction according to optimization constraints. By adopting a hierarchical and partitioned control mode, the master control system only regulates and controls the partitioned input and output without directly intervening in the operation of the underlying components, thereby reducing frequent regulation and control of equipment, realizing autonomous adaptation and blocking of fault conduction in the partitioned area, and improving the control accuracy and stability of dredging operations.

Description

Technical Field

[0001] This application belongs to the field of dredger control technology, and relates to a control method and system for a cutter suction dredger. Background Technology

[0002] A cutter suction dredger is a highly efficient dredging device that integrates dredging, conveying, discharging, and treating sludge, enabling continuous operation on water. The working process of a cutter suction dredger is a continuous cycle combining mechanical cutting and hydraulic conveying, mainly consisting of the following three steps: A. Cutting and loosening: A specially designed rotating cutter head is lowered under the bridge at the front of the hull to cut, break, and loosen the mud, sand, rocks, or sediments on the bottom of the water, so that they are mixed with water to form a high-concentration mud slurry; B. Suction and Lifting: The high-power centrifugal mud pump in the cabin starts, generating a strong vacuum negative pressure suction force to suck the loosened mud into the pump through the mud suction pipe; C. Transportation and Discharge: After being pressurized by a mud pump, the mud is continuously transported through a fully enclosed mud discharge pipeline (floating pipe on water or land pipeline) to a designated reclamation area or stockpile several kilometers or even tens of kilometers away.

[0003] As can be seen from the above steps, the core working components of a cutter suction dredger include: a cutter head and cutter cutter cutter blades that are responsible for directly crushing and cutting underwater soil; a mud pump that generates suction to suck in mud and provides sufficient pressure to discharge it over long distances through pipelines; a mud discharge pipeline that is responsible for transporting the mud to the final disposal site; and a control system that enables unmanned and precise operation. In practical applications, for some large dredgers, a set of dredging pipelines usually has multiple sub-dredging pipes connected in series, and multiple mud pumps are installed on them to provide sufficient pumping pressure to overcome the huge resistance brought by long-distance pipelines. At the same time, in specific construction scenarios, dual-pump parallel connection (setting up two deck pumps to deliver fluid to the same pressure pipeline at the same time, mainly to significantly increase the flow rate of mud under short discharge distance) and multi-pipe parallel connection (multiple dredging pipes are set up in parallel and the opening and closing of each pipeline is controlled by electro-hydraulic gate valves) mode have also emerged. This not only improves the adaptability of dredgers to different working conditions, but also allows the dredger to continue to operate when one pipeline or mud pump fails, thus improving the continuity and reliability of operations.

[0004] To ensure that dredgers can adapt to different operating environments, such as underwater soil type, water depth and topography, and water flow speed, current dredgers are usually equipped with intelligent control systems. These systems collect a large amount of parameter data in real time through a network of high-precision sensors distributed throughout the vessel. Then, through machine learning algorithms, neural networks, and pre-established digital twin models, they perform high-speed calculations and deductions to autonomously calculate the most efficient operating plan suitable for the current working conditions and automatically issue instructions to various actuators, thus achieving closed-loop control.

[0005] Current intelligent control systems for cutter suction dredgers typically use real-time data from all working points (or flow nodes) on the dredger as training material, which is then trained via a neural network. This control model can accurately determine the correlations between various parameters at each working point, such as the relationship between flow rate, concentration, and mud pump operating parameters, and the relationship between pipeline resistance and discharge pressure curves. The intelligent control system built upon this model offers significant advantages in practice. For example, it can quickly respond to changes in the underwater soil conditions at the cutter head, automatically adjusting the cutter's cutting thickness and rotation speed, while simultaneously coordinating the output power of each downstream mud pump. Furthermore, when a malfunction occurs at any point on the dredger, the control model can respond rapidly, adjusting the control parameters of other working points to ensure stable system operation.

[0006] However, in practice, it has been found that using the control model to generate a control scheme for the entire dredger control system can lead to a situation where an anomaly at one working point can affect numerous other working points. For example, when the pressure in a certain sludge pipe exceeds a set value, the control model will coordinately adjust the power of the sludge pumps or change the cutter head speed at multiple points, essentially extending the impact of the anomaly to the entire system. This control model has at least two drawbacks: First, it cannot suppress the impact of anomalies at a single point within a certain range, requiring real-time adjustments to operating parameters at certain working points. Frequent operation of components such as hydraulic valves or sludge pumps shortens their lifespan and increases equipment failure rates. Second, because the training data for the control model is global data, some key but scarce anomaly data is treated as noise and smoothed out when training the model using a neural network. This results in a weaker perception of the working state of local areas in the later-generated control model, reducing the reliability of the control to some extent.

[0007] Ensuring that the impact of abnormal locations on the cutter suction dredger does not extend to the entire control system, while simultaneously ensuring the overall stable operation of all core components and maintaining operational efficiency, is a pressing issue that needs to be addressed. Summary of the Invention

[0008] To address the issue that in practical applications, the impact of anomalies at a single working point of a cutter suction dredger can easily spread to the entire dredger control system, leading to excessively frequent responses at some working points, reduced control system stability, and increased probability of failure, this application provides a control method for a cutter suction dredger. This method divides the entire dredger control system into multiple distinct control regions, configuring each region with its own independent control model and parameters. Then, a global control model is used to coordinate the control of each region. This effectively suppresses the impact of anomalies at a single working point, maintaining overall system stability, while also accurately detecting anomalies in localized areas, reducing the probability of failure. To implement this control method, this application also proposes a control system for a cutter suction dredger, the specific scheme of which is as follows: A control method for a cutter suction dredger includes: Based on the working steps of the dredger, the control system is divided into a main control system and multiple sub-control units connected to it. Each sub-control unit corresponds to a sub-control area, and at least one working data sampling point is set in the sludge discharge pipe between two adjacent sub-control areas. Obtain the working components and their operating parameters contained in each sub-control area, as well as the input data of the sub-control area and / or the output data of the sub-control area under the above operating parameter conditions; Based on the operating parameters, input data, and output data of each working component in the sub-control area, and taking the output data as the target control data, the correlation between the combination of each input data and operating parameter and the target control data is obtained through deep learning algorithm analysis, and a regional control model is generated based on the working condition constraints. Based on the data connection relationship between each sub-control area and the input and output data corresponding to each sub-control area, one or more set parameter data in one or more working data sampling points are used as target optimization data. The correlation relationship between the input data, output data and target optimization data of each sub-control area is obtained through deep learning algorithm analysis, and a global control model is generated according to the working condition constraints. Using the working data sampling point where the target optimization data is located as the control node, the control parameters for the control node input by the user are obtained, and the basic operating parameters required by each working component are generated based on the global control model and the regional control model. Based on the aforementioned required basic operating parameters, control commands are generated and issued to the corresponding working components. During the operation of each working component, the global control model acquires the output data of each working data sampling point, predicts the output data of the control node based on the model predictive control algorithm, and outputs the working operation parameters in advance based on the prediction results and optimization constraints to reverse the output data of the corresponding sub-control area. The input and output data include: flow rate data, concentration data, and pressure data; The operating conditions constraints include: pipeline pressure threshold range, mud-water concentration threshold range, cutter speed, and mud pump power range. The optimization constraints include minimizing the number of sub-control areas that need to be adjusted in reverse based on the prediction results.

[0009] Unlike the global control in existing technologies, the above-described technical solution divides the entire control system into multiple relatively independent sub-control areas. During control, the global control model in the main control system only controls the inputs and outputs of the sub-control areas and does not involve direct regulation of the underlying working components. Furthermore, during the control process, the global control model predicts the output data of the control nodes in advance based on the model predictive control algorithm, and only regulates a small number of sub-control areas, reducing frequent intervention in the underlying working components. If a working component in a sub-control area fails, the area control model corresponding to the sub-control area can coordinate the working parameters of the internal working components to reduce the impact of the failure and prevent the failure from being transmitted to related sub-control areas, thus ensuring the stable operation of the entire control system.

[0010] Optionally, a regional control model is generated based on the operating condition constraints, including: Acquire and store the historical operating parameters of each working component in the sub-control area, as well as the input and output data of the sub-control area; The above data are normalized and noise data is removed to generate the first training dataset. The region control model is generated by training a neural network based on the first training dataset. The area control model includes the correlation between the operating parameters of each working component and is equipped with an abnormal alarm unit: when the correlation between the operating parameters of the working components does not match the setting, an alarm message is output. The working components and operating parameters include: the auger and its rotation speed, the suction pump and its suction power, and the delivery pump and its output power; The operating conditions constraints include: the cutter speed range, the suction pump operating power range, and the delivery pump operating power range.

[0011] The above technical solution can accurately construct the area control model corresponding to the sub-control area. The model can monitor the working status of each working component in the sub-control area and output an alarm in time when the operating parameters of the working component are abnormal.

[0012] Optionally, the control method further includes: After obtaining the operating parameters output by the global control model, the response speed of each corresponding sub-control area is obtained, and the operating parameters, sub-control areas, and response speeds are associated and stored to form optimization constraint reference data. Based on the above optimization constraint reference data, the current optimization constraints are adjusted as follows: After normalizing the response speed and number of sub-control areas, a new optimization constraint is formed by weighting and fusing the response speed of each sub-control area and the number of sub-control areas that need to be responded to. Based on the output data of the current control node, and in conjunction with the new optimization constraints, new operating parameters are generated and output.

[0013] The above technical solution can find the optimal balance between the number of controls and the response speed in the sub-control areas, taking into account both the control response speed and the number of controls, and maintaining the efficient and stable operation of the control system.

[0014] Optionally, the control method further includes: Acquire and analyze the fluctuation range of the operating parameters of each working component under various operating parameters based on the historical operating data of the control system, and set the control priority for each working component according to the magnitude of the fluctuation range; The current optimization constraints are adjusted according to the aforementioned control priority, and the adjustment method is as follows: After normalizing the priority and number of sub-control areas, a new optimization constraint is formed by weighting and fusing the priority of each sub-control area and the number of sub-control areas that need to be responded to. Based on the output data of the current control node, and in conjunction with the new optimization constraints, new operating parameters are generated and output.

[0015] The above technical solution not only considers the number of sub-control areas to be adjusted when determining the sub-control areas to be adjusted, but also takes into account the fluctuation range required by the working components in each sub-control area, so as to avoid large fluctuations in the working parameters of the working components in a certain sub-control area.

[0016] Optionally, the control method further includes: Analyze whether there are functionally interchangeable working components in each sub-control area, and mark the above working components as optional components; After each optional component is connected to the sub-control area, the operating parameters, input data, and output data of each working component are obtained. The output data is used as the target control data to generate multiple area control models and store them in association with each optional component.

[0017] Through the above technical solution, when there are multiple replaceable working components in the sub-control area, different area control models will be configured accordingly to ensure the stable operation of each sub-control area.

[0018] Optionally, a global control model is generated based on the operating condition constraints, including: Select one or more set parameter data from one or more working data sampling points, perform normalization processing on them, set a weighted fusion formula, and generate the target control data based on the weighted fusion formula; The input and output data corresponding to each sub-control area, as well as the data of external influencing factors, are obtained as the second training dataset; The global control model is generated by training a neural network based on the second training dataset. The external influencing factors data include: dredging depth, sludge discharge distance, and ship tilt and heel attitude data; The operating condition constraints also include: dredging depth threshold, sludge discharge distance threshold, and hull attitude threshold.

[0019] The above technical solution integrates the control targets of multiple working data sampling points into one, which facilitates model generation. At the same time, external influencing factors such as ship hull status and dredging depth are also introduced as training materials in the generation of the global control model to ensure the accuracy of the global control model and its adaptability to dredging conditions in different waters and construction scenarios.

[0020] Optionally, the operating parameters of the working component are directly fed back and / or collected and output via a specific sensor; The main control system and each sub-control unit are configured with a dual-link data transmission architecture, including a wired bus transmission link and a wireless redundant transmission link.

[0021] The above technical solution can synchronize the operating parameters, input data, output data, and working data sampling points of each sub-control area in real time, ensuring the real-time and complete data interaction between the overall control model and the regional control model.

[0022] To implement the above-mentioned cutter suction dredger control method, this application also proposes a cutter suction dredger control system, which includes a main control system and multiple sub-control units that are data-connected to the main control system. Each of the multiple sub-control units is provided with a sub-control area, and at least one working data sampling point is provided on the sludge discharge pipe between two adjacent sub-control areas. The working data sampling point is used to collect flow data, concentration data and pressure data. Each of the sub-control units is equipped with a region modeling module. The region modeling module is used to obtain the operating parameters of the working components in the corresponding sub-control region, the input data and output data of the sub-control region, and use the output data as the target control data. It uses deep learning algorithms to mine the correlation between the input data, the combination of operating parameters and the target control data, and generates a region control model in combination with preset working condition constraints. The main control system is equipped with a global modeling module, a parameter generation module, and a prediction and control module. The global modeling module is used to generate a global control model based on the data connection relationship of each sub-control unit, the input data and output data of each sub-control area, the set parameter data of the working data sampling point as the target optimization data, the deep learning algorithm to mine the correlation relationship between the input and output data of each sub-control area and the target optimization data, and the working condition constraints. The parameter generation module is used to obtain the control parameters of the control nodes input by the user, using the working data sampling points corresponding to the target optimization data as control nodes, and to generate the basic operating parameters required by each working component by combining the global control model and the control models of each region. Based on the basic operating parameters, control commands are generated and sent to the corresponding working components. The prediction and control module is used to acquire the output data of each working data sampling point in real time, predict the change trend of the output data of the control node through the model prediction control algorithm, and adjust the output data of the corresponding sub-control area in advance based on the prediction result and the preset optimization constraints. The operating constraints include the range of pipeline pressure threshold, the range of slurry concentration threshold, the range of cutter head speed, and the range of slurry pump power; the optimization constraint is to minimize the number of sub-control areas with reverse adjustment; the main control system only performs global control on the input and output data of each sub-control area, without directly interfering with the operation of the underlying working components, and each sub-control unit independently coordinates the operating parameters of its internal working components through the regional control model.

[0023] Optionally, the region modeling module includes a data preprocessing unit and a model training unit; The data preprocessing unit is used to collect and store the historical operating parameters of each working component in the sub-control area, the input data and output data of the sub-control area, normalize the collected data and remove noise, and generate the first training dataset. The model training unit is used to generate a region control model by training a neural network based on the first training dataset. The area control model embeds an abnormal alarm unit, which is used to monitor the correlation between the operating parameters of each working component. When the actual parameter correlation does not match the preset correlation, an abnormal alarm message is output. The working components include a cutter, a suction pump, and a delivery pump, with corresponding operating parameters of cutter speed, suction pump suction power, and delivery pump output power, respectively. The operating condition constraints also include the operating power range of the suction pump and the operating power range of the delivery pump; The global modeling module includes a target fusion unit and a dataset construction unit; The target fusion unit is used to normalize the set parameter data of multiple working data sampling points and integrate them into unified target control data through a preset weighted fusion formula. The dataset construction unit is used to collect input and output data and external influencing factor data of each sub-control area as a second training dataset. The external influencing factor data includes dredging depth, mud discharge distance, and ship hull tilt and trim attitude data. The global modeling module generates a global control model based on the second training dataset through neural network training. The working condition constraints also include dredging depth threshold, mud discharge distance threshold, and hull attitude threshold.

[0024] Optionally, the main control system is also configured with a constraint adaptive optimization module; The constraint adaptive optimization module is used to obtain the response speed of each sub-control area after the global control model outputs the working parameters, associate and store the working parameters, sub-control area identifiers and response speeds, and construct optimization constraint reference data. The constraint adaptive optimization module is used to normalize and weight the response speed data and the number of sub-control areas, update and generate new optimization constraints, and iteratively generate working parameters that are adapted to the current working conditions based on the updated optimization constraints and the global control model. The constraint adaptive optimization module is also used to analyze the parameter fluctuation range of each working component under each working operating parameter based on the historical operating data of the control system, and set the control priority for each working component and the corresponding sub-control area according to the size of the fluctuation range. The constraint adaptive optimization module normalizes and weights the priority data and the number of sub-control areas, and then updates and optimizes the constraints in a second time.

[0025] This application includes at least one of the following beneficial effects: (1) Unlike traditional global direct control, the control method in this application divides the system into multiple independent sub-control zones. The global model only performs macroscopic control over the input and output of the zone, without directly intervening in the underlying components, thus achieving effective decoupling between the management layer and the execution layer; (2) The global model is based on the Model Predictive Control (MPC) algorithm, which predicts the output data of the control nodes in advance and only performs necessary control on a small number of sub-control areas. The forward-looking control strategy greatly reduces the frequent intervention of the underlying working components and improves the smoothness of control; (3) When a working component fails, the corresponding area control model can automatically coordinate the parameters of other internal components to compensate. The autonomous capability of the above-mentioned sub-control areas can quickly reduce the impact of single-point failures on local operations; (4) The independence of the sub-control area constitutes a natural fault "firewall". Single point faults are effectively limited to the current area, avoiding the transmission of fault effects to related areas, thereby ensuring the stable operation of the entire control system under abnormal conditions. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the dredger control method of this application; Figure 2 A schematic diagram illustrating the method for constructing a global control model; Figure 3 This is a schematic diagram showing the connection of the functional modules of the control system of this application.

[0027] Figure reference numerals: 100, Main control system; 110, Global modeling module; 120, Parameter generation module; 130, Predictive control module; 140, Constrained adaptive optimization module; 150, Global control model; 200, Sub-control unit; 210, Regional modeling module; 220, Regional control model; 300, Working component. Detailed Implementation

[0028] The embodiments of this application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0029] In the description of this specification, the references to "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples" refer to specific features, structures, materials, or characteristics described in connection with the described embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0030] It should be noted that the cutter suction dredger control system and control method disclosed in this embodiment are applicable to various conventional and complex dredging conditions such as inland waterway dredging, port dredging, land reclamation, and underwater foundation pit excavation. They can adapt to different dredging depths, different discharge distances, and different hull attitudes in mud and water transport control scenarios, and solve the technical problems caused by the traditional dredger's unified control across the entire area, such as control lag, large parameter fluctuations, frequent start-stop adjustments of bottom components, easy spread of faults across the entire area, and poor adaptability to complex working conditions.

[0031] A control method for a cutter suction dredger, such as Figure 1 As shown, the main steps include the following: S100: Based on the dredging vessel's operating steps, the control system is divided into a main control system and multiple sub-control units connected to it for data transmission. Each sub-control unit corresponds to a sub-control area, and at least one working data sampling point is set in the sludge discharge pipe between two adjacent sub-control areas. The aforementioned working data sampling point is preferably located in the sludge discharge pipe and is used to collect flow rate data, concentration data, and pressure data from the pipe.

[0032] S200: Acquire the working components and their operating parameters contained in each sub-control zone, as well as the input data of the sub-control zone and / or the output data corresponding to the sub-control zone under the above operating parameter conditions. The input data and output data include flow data, concentration data, and pressure data. The working components include a cutter head, a suction pump, and a delivery pump, and the corresponding operating parameters are cutter head speed, suction pump suction power, and delivery pump output power, respectively.

[0033] S300, based on the operating parameters, input data, and output data of each working component in the sub-control area, uses the output data as the target control data. Through deep learning algorithms, it analyzes and obtains the correlation between the combination of input data and operating parameters and the target control data, generating a regional control model based on operating condition constraints. These constraints include: pipeline pressure threshold range, slurry concentration threshold range, cutterhead speed, and slurry pump power range. That is, when using the aforementioned regional control model to regulate the operating parameters of each working component, data such as pipeline pressure must not exceed the threshold ranges set in the operating condition constraints, thereby ensuring the safe operation of the equipment.

[0034] S400, based on the data connection relationships between each sub-control area and the corresponding input and output data of each sub-control area, uses one or more set parameter data from one or more working data sampling points as target optimization data. Through deep learning algorithms, it analyzes and obtains the correlation between the input and output data of each sub-control area and the target optimization data, and generates a global control model according to the operating condition constraints. Here, the operating condition constraints are also: pipeline pressure threshold range, slurry concentration threshold range, cutterhead rotation speed, and slurry pump power range.

[0035] S500: Using the working data sampling point where the target optimization data is located as the control node, the control parameters for the control node input by the user are obtained, and the basic operating parameters required by each working component are generated based on the global control model combined with the regional control model.

[0036] S600 generates control commands based on the aforementioned required basic operating parameters and sends them to the corresponding working components.

[0037] During the operation of each working component, the global control model acquires the output data of each working data sampling point, predicts the output data of the control node based on the Model Predictive Control (MPC) algorithm, and outputs the working operation parameters in advance based on the prediction results and optimization constraints to adjust the output data of the corresponding sub-control area in reverse. In this embodiment, the above-mentioned optimization constraints include minimizing the number of sub-control areas adjusted in reverse based on the prediction results.

[0038] In step S300 above, generating a regional control model based on the operating condition constraints specifically includes: S310: Acquire and store the historical operating parameters of each working component in the sub-control area, as well as the input and output data of the sub-control area; S320, Normalize the above data and remove noisy data to generate the first training dataset; S330, The region control model is generated by training a neural network based on the first training dataset.

[0039] In a specific implementation, the regional control model not only includes the correlation between the operating parameters of each working component, but also incorporates an anomaly alarm unit: when the correlation between the operating parameters of the working components does not conform to the settings, an alarm message is output. For example, when the relationship between the output power of the centrifugal pump and the pipeline pressure data is abnormal, it usually indicates that the pipeline has ruptured and leaked. At this time, the aforementioned regional control model outputs an alarm message to the designated monitoring terminal, such as the monitoring center server. In the prior art, local pipeline ruptures are usually difficult for the global control model to detect (a small amount of low-frequency fault data is easily smoothed out as noise during model training). By independently configuring the regional control model, it is possible to clearly know whether the status of each working component in each sub-control area is normal.

[0040] In this embodiment, the working components and operating parameters included in the training area control model are: the cutter head and its rotational speed, the suction pump and its suction power, and the delivery pump and its output power. The operating constraints include: the cutter head rotational speed range, the suction pump operating power range, and the delivery pump operating power range.

[0041] When generating the regional control model, the control method described in this application further includes: S340, analyze and determine whether there are functionally interchangeable working components in each sub-control area, and mark the aforementioned working components as optional components. Examples of such functionally interchangeable working components include: two mud pumps connected in parallel, with the mud pump connection status determined by an electrically controlled three-way valve.

[0042] S350 acquires the operating parameters, input data, and output data of each working component after each optional component is connected to the sub-control area, uses the output data as the target control data, generates multiple area control models, and stores them in association with each optional component.

[0043] When there are multiple alternative working components in a sub-control area, different area control models will be configured accordingly to ensure the stable operation of each sub-control area.

[0044] In step S400, a global control model is generated based on the operating condition constraints, such as... Figure 2 As shown, it specifically includes: S410: Select one or more set parameter data from one or more working data sampling points, perform normalization processing on them, set a weighted fusion formula, and generate the target control data based on the weighted fusion formula; S420, acquire the input and output data corresponding to each sub-control area, as well as the data of external influencing factors, as the second training dataset; S430, The global control model is generated by training a neural network based on the second training dataset.

[0045] In this embodiment, the aforementioned external influencing factor data includes: dredging depth, sludge discharge distance, and hull tilt and heel attitude data. The corresponding operating condition constraints also include: dredging depth threshold, sludge discharge distance threshold, and hull attitude threshold.

[0046] In step S410 above, the control target of multiple working data sampling points is integrated into one, which facilitates model generation. At the same time, in step S420, external influencing factors such as hull status and dredging depth are also introduced as training materials in generating the global control model to ensure the accuracy of the global control model and its adaptability to dredging conditions in different waters and construction scenarios.

[0047] To find the optimal balance between the number of controls to be adjusted in the sub-control area and the response speed, the control method further includes: S610: After obtaining the operating parameters output by the global control model, the response speed of each corresponding sub-control area is obtained, and the operating parameters, sub-control areas and response speeds are associated and stored to form optimization constraint reference data. S611, Adjust the current optimization constraints based on the above optimization constraint reference data. The adjustment method is as follows: After normalizing the response speed and number of sub-control areas, a new optimization constraint is formed by weighting and fusing the response speed of each sub-control area and the number of sub-control areas that need to be responded to. Based on the output data of the current control node, and in conjunction with the new optimization constraints, new operating parameters are generated and output.

[0048] The above scheme can balance the control response speed and the number of controls, thus maintaining the efficient and stable operation of the control system.

[0049] In practice, when determining the sub-control areas that need adjustment, not only the number of sub-control areas to be adjusted is considered, but also the required fluctuation range of the working components within each sub-control area is taken into account to avoid large fluctuations in the working parameters of the working components in a certain sub-control area. Further optimized by this, the control method also includes: S620: Acquire and analyze the fluctuation range of the operating parameters of each working component under various operating parameters based on the historical operating data of the control system, and set the control priority for each working component according to the magnitude of the fluctuation range; S621, adjust the current optimization constraints according to the aforementioned control priority, and adjust them in the following way: After normalizing the priority and number of sub-control areas, a new optimization constraint is formed by weighting and fusing the priority of each sub-control area and the number of sub-control areas that need to be responded to. Based on the output data of the current control node, and in conjunction with the new optimization constraints, new operating parameters are generated and output.

[0050] In order to synchronize the operating parameters, input data, output data and working data sampling points of each sub-control area in real time, and to ensure the real-time and complete data interaction between the global control model and the regional control model, the operating parameters of the working components are directly fed back and output by the working components and / or collected and output through specific sensors.

[0051] The main control system and each sub-control unit are configured with a dual-link data transmission architecture, including a wired bus transmission link and a wireless redundant transmission link.

[0052] To implement the above-mentioned cutter suction dredger control method, this application also discloses a cutter suction dredger control system, such as... Figure 3As shown, the system adopts a hierarchical control architecture, including a main control system 100 and multiple sub-control units 200 that establish real-time data communication with the main control system 100. Each sub-control unit 200 corresponds to an independent sub-control area, realizing segmented and zoned refined management and control. At least one working data sampling point is set on the sludge discharge pipe connecting two adjacent sub-control areas. Each sampling point can collect sludge flow rate data, sludge concentration data, and pipeline pressure data in real time during the dredging operation, providing real-time data support for model training, parameter adjustment, and trend prediction.

[0053] Each sub-control unit 200 is equipped with a dedicated area modeling module 210, responsible for constructing an area control model 220 adapted to the corresponding zone's operating conditions. Specifically, the area modeling module 210 continuously collects the operating parameters of core working components 300 such as the cutterhead, suction pump, and delivery pump within the corresponding sub-control area, while simultaneously acquiring the input and output data of that sub-control area. Using the sub-control area's output data as the target control benchmark, a deep learning algorithm is employed to analyze the inherent correlation between the acquired sub-control area's input data, component operating parameter combinations, and target control data. Combined with preset operating condition constraints, this data is used to train and generate an area control model 220 adapted to independent operation in a single area.

[0054] Specifically, the regional modeling module 210 described in this application includes a data preprocessing unit and a model training unit. The data preprocessing unit is responsible for collecting and storing historical operating parameters, regional input and output data of each working component 300 within the sub-control area, and performing purification processing on the raw collected data, such as normalizing to unify dimensions and removing abnormal noise data, to form a standardized first training dataset. The model training unit, based on the first training dataset, completes iterative model training through a neural network, ultimately outputting a stable and usable regional control model 220.

[0055] In this embodiment, the area control model 220 embeds an anomaly alarm unit, which can monitor the matching and correlation relationships of the operating parameters of each working component 300 within the sub-control area in real time. When the correlation logic of the actual operating parameters of the component does not match the preset standard correlation relationship, and problems such as parameter abnormalities or operational imbalances occur, an anomaly alarm is triggered and output, realizing real-time early warning of zonal operation faults. The core operating parameters of each working component 300 are: cutterhead speed, suction pump suction power, and delivery pump output power. The corresponding basic operating condition constraints include the pipeline pressure threshold range, mud-water concentration threshold range, cutterhead speed range, mud pump power range, suction pump operating power range, and delivery pump operating power range, ensuring that the operating parameters of a single sub-control area are within a safe and compliant range.

[0056] The main control system 100 is equipped with a global modeling module 110, a parameter generation module 120, and a predictive control module 130. The main control system 100 is responsible for the global optimization and control of the input and output data of each sub-control area, without directly interfering with the specific operation of the underlying working components 300. Each sub-control unit 200 independently coordinates the operating parameters of its internal working components 300 according to its own regional control model 220, taking into account both global synergy and zonal flexibility.

[0057] The global modeling module 110 is responsible for constructing a global control model 150 covering the entire operation link. During implementation, based on the data connection logic of each sub-control unit 200 and the input and output data of each sub-control area, and taking the set parameter data of the sludge discharge pipe working data sampling point as the target optimization benchmark, the module uses deep learning algorithms to mine the correlation between the input and output data of each sub-control area and the target optimization data, and trains and generates the global control model 150 in combination with the global working condition constraints.

[0058] Specifically, the global modeling module 110 includes a target fusion unit and a dataset construction unit. The target fusion unit normalizes the set parameter data of multiple working data sampling points, integrates all sampling point parameters through a preset weighted fusion formula, and generates unified global target control data, solving the problem of inconsistent parameter standards among multiple sampling points. The dataset construction unit collects the input and output data of each sub-control area, and also incorporates external influencing factor data such as dredging depth, sludge discharge distance, and ship attitude, integrating them to form a second training dataset. The global modeling module 110 completes neural network training based on the second training dataset, generating a global control model 150. The corresponding working condition constraints include dredging depth threshold, sludge discharge distance threshold, and ship attitude threshold, fully adapting to complex dredging operation environments.

[0059] The parameter generation module 120 is the core control unit for distributing system parameters. During implementation, it uses each working data sampling point as a control node, receives the target control parameters of the control node input by the user, and combines the global optimization logic of the global control model 150 with the partition adaptation logic of each area control model 220 to iteratively calculate the basic operating parameters required by each working component 300. Finally, it is transformed into standardized control instructions and distributed to the working components 300 in the corresponding sub-control areas to complete the precise configuration of operation parameters.

[0060] The predictive control module 130 enables advanced prediction and advance control. During operation, it collects output data from each working data sampling point in real time and dynamically predicts the changing trends of the output data at each control node through a model predictive control algorithm. Based on the prediction results and combined with the preset optimization constraint of "minimizing the number of sub-control areas for reverse adjustment", it adjusts the output data of the corresponding sub-control areas in advance, effectively avoiding operational anomalies such as sudden changes in pipeline pressure, imbalance of slurry concentration, and flow fluctuations, and ensuring the stability of dredging operations.

[0061] In a preferred embodiment, the main control system 100 is also equipped with a constraint adaptive optimization module 140 to achieve dynamic adaptation between constraints and operating conditions. The constraint adaptive optimization module 140 first acquires the operating parameters output by the global control model 150, and then the operating response speed of each sub-control area. It then associates and stores the operating parameters, sub-control area identifiers, and response speeds to construct an optimized constraint reference database. By normalizing and weighting the response speed data and the number of sub-control areas participating in the control, it updates and optimizes the constraints, and iteratively generates component operating parameters adapted to the current real-time operating conditions based on the global control model 150.

[0062] Further optimization involves the constraint adaptive optimization module 140 retrieving historical system operation data to analyze the parameter fluctuation amplitude of each working component 300 under different operating parameters. Based on the magnitude of the fluctuation range, the module prioritizes the control of each working component 300 and its corresponding sub-control area. The sub-control area priority data and the number of sub-control areas are then normalized and weighted and merged again. The constraints are updated and optimized in a second iteration, further optimizing the control logic. While ensuring the stability of the entire operation, the module minimizes the number of control zones, reduces system control energy consumption and operational fluctuations, and adapts to complex dredging scenarios with different water depths, sludge discharge distances, and vessel attitudes.

[0063] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A control method for a cutter suction dredger, characterized in that, include: Based on the working steps of the dredger, the control system is divided into a main control system and multiple sub-control units connected to it. Each sub-control unit corresponds to a sub-control area, and at least one working data sampling point is set in the sludge discharge pipe between two adjacent sub-control areas. Obtain the working components and their operating parameters contained in each sub-control area, as well as the input data of the sub-control area and / or the output data of the sub-control area under the above operating parameter conditions; Based on the operating parameters, input data, and output data of each working component in the sub-control area, and taking the output data as the target control data, the correlation between the combination of each input data and operating parameter and the target control data is obtained through deep learning algorithm analysis, and a regional control model is generated based on the working condition constraints. Based on the data connection relationship between each sub-control area and the input and output data corresponding to each sub-control area, one or more set parameter data in one or more working data sampling points are used as target optimization data. The correlation relationship between the input data, output data and target optimization data of each sub-control area is obtained through deep learning algorithm analysis, and a global control model is generated according to the working condition constraints. Using the working data sampling point where the target optimization data is located as the control node, the control parameters for the control node input by the user are obtained, and the basic operating parameters required by each working component are generated based on the global control model and the regional control model. Based on the aforementioned required basic operating parameters, control commands are generated and issued to the corresponding working components. During the operation of each working component, the global control model acquires the output data of each working data sampling point, predicts the output data of the control node based on the model predictive control algorithm, and outputs the working operation parameters in advance based on the prediction results and optimization constraints to reverse the output data of the corresponding sub-control area. The input and output data include: flow rate data, concentration data, and pressure data; The operating conditions constraints include: pipeline pressure threshold range, mud-water concentration threshold range, cutter speed, and mud pump power range. The optimization constraints include minimizing the number of sub-control areas that need to be adjusted in reverse based on the prediction results.

2. The control method for a cutter suction dredger according to claim 1, characterized in that, Generate a regional control model based on the operating condition constraints, including: Acquire and store the historical operating parameters of each working component in the sub-control area, as well as the input and output data of the sub-control area; The above data are normalized and noise data is removed to generate the first training dataset. The region control model is generated by training a neural network based on the first training dataset. The area control model includes the correlation between the operating parameters of each working component and is equipped with an abnormal alarm unit: when the correlation between the operating parameters of the working components does not match the setting, an alarm message is output. The working components and operating parameters include: the auger and its rotation speed, the suction pump and its suction power, and the delivery pump and its output power; The operating conditions constraints include: the cutter speed range, the suction pump operating power range, and the delivery pump operating power range.

3. The control method for a cutter suction dredger according to claim 2, characterized in that, The control method further includes: After obtaining the operating parameters output by the global control model, the response speed of each corresponding sub-control area is obtained, and the operating parameters, sub-control areas, and response speeds are associated and stored to form optimization constraint reference data. Based on the above optimization constraint reference data, the current optimization constraints are adjusted as follows: After normalizing the response speed and number of sub-control areas, a new optimization constraint is formed by weighting and fusing the response speed of each sub-control area and the number of sub-control areas that need to be responded to. Based on the output data of the current control node, and in conjunction with the new optimization constraints, new operating parameters are generated and output.

4. The control method for a cutter suction dredger according to claim 3, characterized in that, The control method further includes: Acquire and analyze the fluctuation range of the operating parameters of each working component under various operating parameters based on the historical operating data of the control system, and set the control priority for each working component according to the magnitude of the fluctuation range; The current optimization constraints are adjusted according to the aforementioned control priority, and the adjustment method is as follows: After normalizing the priority and number of sub-control areas, a new optimization constraint is formed by weighting and fusing the priority of each sub-control area and the number of sub-control areas that need to be responded to. Based on the output data of the current control node, and in conjunction with the new optimization constraints, new operating parameters are generated and output.

5. The control method for a cutter suction dredger according to claim 3, characterized in that, The control method further includes: Analyze whether there are functionally interchangeable working components in each sub-control area, and mark the above working components as optional components; After each optional component is connected to the sub-control area, the operating parameters, input data, and output data of each working component are obtained. The output data is used as the target control data to generate multiple area control models and store them in association with each optional component.

6. The control method for a cutter suction dredger according to claim 1, characterized in that, Generate a global control model based on operating condition constraints, including: Select one or more set parameter data from one or more working data sampling points, perform normalization processing on them, set a weighted fusion formula, and generate the target control data based on the weighted fusion formula; The input and output data corresponding to each sub-control area, as well as the data of external influencing factors, are obtained as the second training dataset; The global control model is generated by training a neural network based on the second training dataset. The external influencing factors data include: dredging depth, sludge discharge distance, and ship tilt and heel attitude data; The operating condition constraints also include: dredging depth threshold, sludge discharge distance threshold, and hull attitude threshold.

7. The control method for a cutter suction dredger according to claim 1, characterized in that, The operating parameters of the working component are directly fed back and output by the working component and / or collected and output via specific sensors; The main control system and each sub-control unit are configured with a dual-link data transmission architecture, including a wired bus transmission link and a wireless redundant transmission link.

8. A control system for a cutter suction dredger, characterized in that, The control system includes a main control system (100) and multiple sub-control units (200) that are data-connected to the main control system (100); Each of the multiple sub-control units (200) is provided with a sub-control area, and at least one working data sampling point is provided on the sludge discharge pipe between two adjacent sub-control areas. The working data sampling point is used to collect flow data, concentration data and pressure data. Each of the sub-control units (200) is equipped with a region modeling module (210). The region modeling module (210) is used to obtain the operating parameters of the working components (300) in the corresponding sub-control area, the input data and output data of the sub-control area, and use the output data as the target control data. It uses deep learning algorithms to mine the correlation between the input data, the combination of operating parameters and the target control data, and generates a region control model (220) in combination with preset working condition constraints. The main control system (100) is equipped with a global modeling module (110), a parameter generation module (120), and a prediction and control module (130); The global modeling module (110) is used to generate a global control model (150) based on the data connection relationship of each sub-control unit (200), the input data and output data of each sub-control area, the set parameter data of the working data sampling point as the target optimization data, the deep learning algorithm to mine the correlation between the input and output data of each sub-control area and the target optimization data, and the working condition constraints. The parameter generation module (120) is used to obtain the control parameters of the control node input by the user by taking the working data sampling point corresponding to the target optimization data as the control node, and generate the basic operating parameters required by each working component (300) by combining the global control model (150) and the regional control model (220), and generate control instructions based on the basic operating parameters and send them to the corresponding working component (300). The prediction and control module (130) is used to acquire the output data of each working data sampling point in real time, predict the change trend of the output data of the control node through the model prediction control algorithm, and adjust the output data of the corresponding sub-control area in advance according to the prediction result and the preset optimization constraints. The operating conditions constraints include the range of pipeline pressure threshold, the range of mud concentration threshold, the range of cutter speed, and the range of mud pump power; the optimization constraint is to minimize the number of sub-control areas for reverse adjustment; the main control system (100) only performs full-domain control on the input and output data of each sub-control area and does not directly interfere with the operation of the underlying working components (300); each sub-control unit (200) independently coordinates the operating parameters of the internal working components (300) through the regional control model (220).

9. The control system for a cutter suction dredger according to claim 8, characterized in that, The region modeling module (210) includes a data preprocessing unit and a model training unit; The data preprocessing unit is used to collect and store the historical operating parameters of each working component (300) in the sub-control area, the input data and output data of the sub-control area, and to perform normalization processing and noise removal on the collected data to generate the first training dataset. The model training unit is used to generate a region control model (220) by training a neural network based on the first training dataset; The area control model (220) has an embedded abnormal alarm unit. The abnormal alarm unit is used to monitor the correlation of the operating parameters of each working component (300). When the actual parameter correlation does not match the preset correlation, it outputs abnormal alarm information. The working component (300) includes a cutter, a suction pump and a delivery pump, with corresponding operating parameters of cutter speed, suction pump suction power and delivery pump output power, respectively. The operating condition constraints also include the operating power range of the suction pump and the operating power range of the delivery pump; The global modeling module (110) includes a target fusion unit and a dataset construction unit; The target fusion unit is used to normalize the set parameter data of multiple working data sampling points and integrate them into unified target control data through a preset weighted fusion formula. The dataset construction unit is used to collect input and output data and external influencing factor data of each sub-control area as a second training dataset. The external influencing factor data includes dredging depth, mud discharge distance, and ship hull tilt and trim attitude data. The global modeling module (110) generates a global control model (150) based on the second training dataset through neural network training. The working condition constraints also include dredging depth threshold, mud discharge distance threshold and hull attitude threshold.

10. The cutter suction dredger control system according to claim 9, characterized in that, The main control system (100) is also equipped with a constraint adaptive optimization module (140); The constraint adaptive optimization module (140) is used to obtain the response speed of each sub-control area after the output of the working operation parameters of the global control model (150), and to associate and store the working operation parameters, sub-control area identifiers and response speeds to construct optimization constraint reference data; The constraint adaptive optimization module (140) is used to normalize and weight the response speed data and the number of sub-control areas, update and generate new optimization constraints, and iteratively generate working parameters that are adapted to the current working conditions based on the updated optimization constraints and the global control model (150). The constraint adaptive optimization module (140) is also used to analyze the parameter fluctuation range of each working component (300) under each working operating parameter based on the historical operating data of the control system, and set the control priority for each working component (300) and the corresponding sub-control area according to the fluctuation range. The constraint adaptive optimization module (140) normalizes and weights the priority data and quantity data of the sub-control areas, and then updates and optimizes the constraint conditions for the second time.

Images