Deep-sea mining green factory system based on artificial intelligence

Abstract

The invention belongs to the field of deep-sea resource development and environmental protection, and discloses a deep-sea mining green factory system based on artificial intelligence, which comprises a marine environment sensing module, a deep-sea heavy-load mining vehicle intelligent control module, an ore pulp lifting and tail water treatment module, a seabed environment monitoring and repairing module and an AI cooperative control center, the environment sensing module collects submarine topography point cloud and flow field data, and a real-time three-dimensional environment reconstruction model is generated through fusion; the deep-sea heavy-load mining vehicle control module outputs power parameters of an advancing and collecting mechanism based on the model, and path planning and posture self-adaption are achieved; the monitoring and repairing module serves as an ecological protection unit, acquires plume concentration and diffusion data in real time, calculates and adjusts the throwing flow speed of a green flocculant through a feedback control algorithm, and achieves active suppression of operation disturbance; and the tail water treatment module dynamically adjusts the operation condition of the grading purification device according to the spectral analysis data of the backflow water body. According to the invention, collaborative protection of deep sea resource development and ecological environment is realized.

Description

Technical Field

[0001] This invention relates to the field of deep-sea resource development and environmental protection, and in particular to a green deep-sea mining factory system based on artificial intelligence. Background Technology

[0002] With the increasing depletion of terrestrial mineral resources and the transformation and upgrading of the energy structure, deep-sea mineral resources, due to their abundant reserves and diverse types, have become an important source of strategic new resources. Polymetallic nodules, cobalt-rich crusts, and hydrothermal sulfides, among other deep-sea minerals, contain large amounts of key metals, providing fundamental support for future energy transformation and high-end manufacturing. However, deep-sea mining environments are complex, extremely stressful, and communication is limited. Traditional mining technologies have significant shortcomings in high-precision control, environmental protection, and system coordination, hindering the large-scale and green development of deep-sea resources.

[0003] Currently, the deep-sea mining field mainly faces the following system-level bottlenecks: (1) The operation mode is "discrete" and lacks system-level integration. Existing mining systems are mostly composed of independent deep-sea heavy-duty mining vehicles, lifting pump pipes and surface mother ship platforms in a temporary network. The lack of deep information interconnection and working condition coupling between the equipment makes mining operations similar to "single machine operation" and cannot form a continuous, collaborative assembly line operation capability similar to land-based factories.

[0004] (2) Lack of independent operation capability. The perception and decision-making of the existing system are highly dependent on the remote control of the surface mother ship platform. Faced with the high latency and complex and ever-changing environment of the deep sea (such as sudden ocean currents and sudden changes in terrain), the system lacks localized autonomous decision-making and real-time response capabilities, making it difficult to achieve long-term independent and stable production.

[0005] (3) Production and environmental governance are separated. Traditional development models follow the logic of "pollute first, then treat" or "mining while monitoring passively". Environmental governance devices are often independent and external, not integrated into the control loop of mining production, resulting in delayed environmental remediation and failure to achieve truly green and sustainable operations.

[0006] Chinese patent application CN120725552A discloses a deep-sea mineral lifting unmanned transportation system. This system constructs a framework of "unmanned vehicle - intelligent execution - flexible scheduling," comprising four main modules: unmanned transport equipment, a communication network system, a scheduling and control center, and a support and maintenance system. The unmanned transport equipment achieves mineral transportation based on buoyancy; the communication network adopts a layered redundancy architecture of underwater acoustic OFDM and laser emergency links to ensure highly reliable communication and centimeter-level positioning; the scheduling and control center improves operational efficiency through intelligent task allocation and trajectory planning; and the support and maintenance system integrates energy management, digital twin monitoring, and emergency response mechanisms. The operational process adopts a "primarily unmanned autonomous operation, supplemented by remote monitoring" mode, covering ballast diving, acoustic-optical docking, buoyancy adjustment, and dynamic unloading from the mother ship. However, it is only a "deep-sea mineral lifting unmanned transportation system," and its core functional architecture (unmanned transport equipment, communication, scheduling, and support and maintenance) revolves around the logistics sub-link of "mineral transportation," thus limiting its system positioning and integration dimensions. Its scheduling and control center mainly realizes task allocation, trajectory planning and collision avoidance, which is a relatively independent transportation scheduling logic and lacks the linkage feedback logic between production and logistics; although its energy supply module supports multi-mode energy supply, its logic focuses on improving the single-unit range and charging efficiency, and energy management lacks system-level dynamic game; the environmental status monitoring of this patent application is mainly used to assist vehicles in avoiding turbulence or assessing risks, lacking active governance methods, and the environmental protection mechanism remains at the passive monitoring stage; the patent application focuses on the unidirectional enhancement of minerals, and the material cycle lacks an overall dynamic balance mechanism.

[0007] Chinese patent application CN118153411A discloses a method for inverting the concentration profile of plumes in deep-sea mining based on electrical monitoring using deep learning. This method comprises four parts: acquisition and normalization of in-situ electrical monitoring data; construction of a model for inverting suspended particle concentration from spontaneous potential measurement results; model verification and optimization; and inversion of the plume particle concentration profile. This patent application solves the problem of "quantitative sensing" of plume concentration in deep-sea environments by establishing an inversion model between spontaneous potential and suspended particle concentration. However, from an industrial closed-loop perspective, this patent application only achieves "passive monitoring" of environmental impact, representing an isolated detection tool. Its biggest drawback is the lack of a "decision-making and execution" link after data sensing, failing to transform the monitored "pollution data" into "governance actions," and thus failing to construct the industrial full life-cycle closed loop of "disturbance generation—real-time sensing—active suppression—ecological restoration" emphasized by this technology. Summary of the Invention

[0008] The purpose of this invention is to solve the problems in existing deep-sea mining technologies, such as "discrete equipment, remote control, and lack of closed-loop operation," which prevent deep-sea mining operations from forming a self-contained system. The invention provides a green factory system for deep-sea mining based on artificial intelligence. By constructing a system structure that integrates "intelligent mining, high-efficiency improvement, green governance, and intelligent operation and maintenance," the invention achieves system-level intelligent, green, and sustainable operation of deep-sea mining operations.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: An AI-based green factory system for deep-sea mining includes: The marine environment three-dimensional perception module is used to reconstruct the three-dimensional topography of the work area and generate ore body abundance distribution maps, providing spatiotemporal benchmarks and resource data for factory production; The intelligent control module of the deep-sea heavy-duty mining vehicle, as the core operating terminal of the system, is based on multi-source heterogeneous data from the operating area and solves the problem through a multi-objective optimization algorithm. This multi-objective optimization algorithm integrates multi-source heterogeneous data perceived from the operating area (including seabed topography slope, ore body abundance, flow field vector, and local track slip state, etc.) as input for the calculation. During the calculation process, the system constructs a dynamic operation cost function with maximizing the ore output rate per unit time as the core optimization objective, and simultaneously couples minimizing ineffective energy consumption and minimizing basement disturbance as evaluation constraints. The algorithm kernel adjusts the weight coefficients of each objective function in real time according to different working condition commands issued by the mining system integration and intelligent operation and maintenance module (such as steady-state production mode or environmental / energy efficiency constraint mode), dynamically calculates and outputs the expected travel speed, driving torque, aspect ratio geometric parameters of the variable track teeth, and jet propagation parameters such as the jet pressure and incident angle of the hydraulic jet of the deep-sea heavy-duty mining vehicle. Through this multi-objective optimization process, the deep-sea heavy-duty mining vehicle can complete the optimal trajectory planning and adaptive adjustment of the acquisition posture in complex terrain while ensuring the ecological red line. It also outputs the travel parameters and collection operation parameters of the deep-sea heavy-duty mining vehicle, enabling efficient stripping and collection of ore; The ultra-deepwater resource enhancement and tailwater treatment module serves as the system's vertical logistics transport channel, used to establish a multiphase flow transport link for slurry from the seabed to the water surface and a tailwater return link, and to maintain the dynamic material balance between the upward solid-liquid flux and the downward liquid phase flux. The seabed environmental disturbance monitoring and remediation module, as the system's environmental constraint and protection unit, is used to quantify plume diffusion flux in real time during production operations and proactively perform ecological restoration. The mining system integration and intelligent operation and maintenance module, as the central collaborative control hub of the system, is connected to the above modules through a heterogeneous network. Among them, each module achieves information sharing and dynamic optimization through the "artificial intelligence decision kernel" inside the mining system integration and intelligent operation and maintenance module, forming a closed-loop system of "perception-decision-execution-feedback-repair", thereby achieving efficient and low-disturbance green mining operations in complex deep-sea environments.

[0010] The mining system integration and intelligent operation and maintenance module incorporates an artificial intelligence decision-making kernel. This kernel includes a model inference unit, a strategy generation unit, and a parameter distribution interface. It receives multi-source state parameters from environmental perception, mining execution, logistics transportation, and environmental remediation modules, and generates a global control strategy based on a pre-trained model and online learning mechanism. The model inference unit, as the kernel's data entry point, receives and aggregates sensor data in real time from four modules: marine environmental three-dimensional perception, intelligent control of deep-sea heavy-duty mining vehicles, resource enhancement and tailwater treatment, and environmental monitoring and remediation. This data includes key state parameters such as seabed topography, slurry concentration, riser pressure drop, and plume turbidity. Using a built-in pre-trained model, it extracts and fuses features from this heterogeneous data to construct state parameters reflecting the plant's current operating conditions and assesses in real time whether the system faces potential risks such as overload, congestion, or environmental exceedances. The system state assessment results generated by the inference unit are transmitted to the strategy generation unit. This unit, based on a hierarchical scheduling logic of "production priority, environmental constraints," performs multi-objective optimization calculations within the red lines of total power quota and material flow balance. The generated logical strategy is transmitted to the parameter distribution interface through an internal interface. This interface is responsible for converting the "strategy" into physical parameters that can be recognized by each work module. By establishing a mapping relationship between the strategy and hardware frequency and torque, the final execution instructions (such as adjusting the frequency of the inverter of the deep-sea heavy-duty mining vehicle, the height of the ore head above the ground, or the output power of the pump group) are sent to the execution mechanism of each work module, thereby realizing the adaptive and self-stabilizing production operation of the entire plant system.

[0011] The AI ​​decision kernel is deployed in a collaborative manner between edge computing nodes and surface computing nodes. Millisecond-level response decision logic is completed on the underwater edge nodes, while long-cycle optimization and model updates are completed on the surface nodes.

[0012] The mining system integration and intelligent operation and maintenance module (i.e., the central collaborative control hub) establishes a global knowledge graph model, integrating status monitoring units and autonomous scheduling control terminals. This module comprehensively analyzes the mining operation status, energy consumption distribution, and environmental feedback of the entire system, and supports dynamic interaction of cross-module information flow. Through artificial intelligence learning mechanisms, it realizes equipment health management, fault prediction, and task scheduling optimization, supports unified decision-making for mining condition identification, energy consumption optimization, and equipment maintenance, and is the brain that enables "factory-like" independent operation.

[0013] The three-dimensional marine environment perception module includes acoustic detection devices, optical imaging units, flow field monitors, chemical sensors, and edge computing nodes. The module adopts an integrated detection platform architecture, with edge computing nodes embedded in the titanium alloy pressure-resistant sealed chambers of the deep-sea heavy-duty mining vehicle, intermediate pumping station, and submersible. All hardware is uniformly connected to the internal edge computing nodes for synchronous data acquisition and preprocessing. In terms of location and coordination, acoustic detection devices (such as multi-beam sonar) are responsible for long-range, large-scale terrain reconstruction; the optical imaging unit, in conjunction with a constant illumination system, performs close-range, fine-grained identification; the flow field monitor (ADCP) captures environmental flow vectors in real time; and the high-frequency scanning lidar is used for roughness calculation of micro-terrain. All hardware components work together through an "acoustic-optical linkage" verification logic to pinpoint the distribution of ore bodies. In terms of data interaction, the raw physical signals acquired by each sensor (such as terrain depth, sound wave scattering intensity, optical pixel matrix, and flow velocity) are aggregated to the edge computing node. After fusion processing, a high-resolution 3D reconstruction model of the seabed environment and an operational cost map are generated and distributed to the mining system integration and intelligent operation and maintenance module and the control terminal of the deep-sea heavy-duty mining vehicle, providing dynamic data support for the path planning and parameter matching of the entire system. The system uses artificial intelligence algorithms to fuse data from multiple sensor sources, enabling real-time perception and anomaly identification of seabed topography, ore body distribution, flow field structure, and environmental parameters. The perceived data is analyzed by AI to generate a high-resolution 3D reconstruction model of the seabed environment, providing dynamic data support for the path planning and operational optimization of the deep-sea heavy-duty mining vehicle.

[0014] The intelligent control module of the deep-sea heavy-duty mining vehicle uses deep learning and reinforcement learning algorithms to achieve terrain recognition, attitude adaptive adjustment, path optimization and conflict avoidance. The operating parameters of the acquisition device, the travel device and the jet device are corrected in real time according to the AI ​​prediction model to achieve stable operation, low-disturbance acquisition and energy efficiency improvement in complex terrain.

[0015] The ultra-deepwater resource enhancement and tailwater treatment module includes a surface mother ship platform, a coarse-grained slurry lifting riser system, staged pump sets, tailwater purification devices, and a fluid circulation control module. This module connects the surface mother ship platform and the subsea factory via a vertical pipeline system, forming a complete circulation system. Physically, the surface mother ship platform serves as the central command center, with the fluid circulation control core deployed on its deck. The riser system, acting as the "main artery" of vertical transportation, is suspended below the hull, with several high-pressure variable frequency booster pump sets distributed at specific depth intervals along its pipe. These components form a closed-loop fluid system through the upward slurry pipeline and the tailwater reinjection and discharge pipe, and achieve real-time data interaction across the entire chain via a photoelectric composite umbilical cable on the riser's outer wall. In terms of interaction logic, the riser's ultrasonic concentration meter and pressure sensors along the pipeline capture the physical state of the slurry in real time and instantly upload the data to the ship's AI decision-making kernel. The algorithm dynamically adjusts the pump set's operating power based on the real-time pressure drop gradient to eliminate the risk of blockage; combined with the analysis results, it adaptively matches the tailwater treatment intensity to ensure that the discharge meets environmental stability requirements. Artificial intelligence algorithms are used to adjust the slurry conveying pressure, flow rate and power distribution in real time to achieve a two-way balance between the upward slurry and the downward tailings water. The tailings water purification device combines image recognition and spectral analysis technology to perform graded purification of the return water, and uses an AI model to achieve adaptive regulation based on emission concentration feedback to ensure green emissions.

[0016] The seabed environmental disturbance monitoring and remediation module uses multi-source sensors to collect data on plume velocity, concentration, and diffusion direction, and establishes a plume dynamics model through AI algorithms. The system automatically identifies plume distribution and deposition risk areas, and adjusts the dosage and timing of green flocculant based on AI analysis results to construct a closed loop for ecological restoration of the work area.

[0017] In this technology, environmental monitoring data is not only quantified, but also used as an "environmental signal" to directly trigger the seabed remediation module to dynamically release flocculants and control the deep-sea heavy-duty mining vehicle to enter a low-disturbance operation mode to complete the relevant remediation work.

[0018] Compared with existing technologies, this invention not only achieves intelligent operation of a single operational link, but also constructs a complete deep-sea industrial system. During operation, this system requires no real-time manual control of individual equipment, only limited human intervention during system startup or strategy update phases, thereby achieving long-term autonomous operation of deep-sea mining operations. This invention has the following significant advantages: (1) A system-level material and energy closed loop has been achieved. In response to the problems of equipment dispersion and poor coordination in traditional technologies, this system effectively solves the industry problems of pipeline blockage and independent power supply system overload through a dual closed-loop control mechanism, and realizes the continuity and self-sustainability of deep-sea mining operations.

[0019] (2) A qualitative leap has been achieved in ecological governance. Traditional environmental control relies on passive adjustments based on lagging data. This system achieves real-time prediction through an AI plume model and actively regulates green flocculants to keep operational disturbances within standard ranges.

[0020] (3) A dynamic closed-loop system for wastewater treatment has been achieved. The AI-driven purification unit ensures that the return water is far below international emission standards, truly building a "deep-sea green factory".

[0021] (4) A deep-sea factory system with "self-sustaining capability" has been constructed (system-level advantage). Addressing the pain points of existing deep-sea mining equipment being "discrete and single-machine," this invention, through an integrated architecture design, deeply integrates front-end data acquisition, vertical transportation, end-of-pipe treatment, and central control, constructing the first deep-sea automated production system with independent operational capabilities. This system can achieve autonomous closed-loop operation throughout the entire process without high-frequency intervention from the mother ship, significantly improving the industrialization level of deep-sea resource development.

[0022] (5) "Material-Energy" dual closed-loop control mechanism. This invention breaks through the limitations of the traditional open-loop operation mode: In the dimension of material flow, a "supply and demand matching" mechanism based on the transport capacity is established. By reversely constraining the production capacity of the deep-sea heavy-duty mining vehicle, the problem of blockage and slug flow in ultra-deepwater risers is effectively optimized; In the dimension of energy flow, a global allocation strategy under limited power is established, realizing the dynamic power game between mining, hoisting and treatment modules, ensuring the energy self-consistency and zero downtime of the system in the deep-sea independent power supply environment.

[0023] (6) Adaptive production based on "capacity priority and environmental constraints" is realized. Unlike the traditional operation mode, this invention establishes a hierarchical scheduling strategy. Under steady state, the goal is to maximize ore output rate. The optimal matching of path and parameters is achieved through a multi-objective optimization algorithm. When environmental disturbances approach the threshold, the "environmental damping" mechanism is automatically triggered. Through negative feedback adjustment, a dynamic balance between production and environmental protection is achieved, thereby maximizing capacity while ensuring the ecological bottom line.

[0024] (7) Optimized the “machine-soil-flow” coupling control problem under complex bottom conditions. For deep-sea soft bottom and complex terrain, the system introduces ground mechanical matching control and fluid-structure coupling regulation technology. By monitoring the slip ratio and sediment shear strength in real time, the optimal travel torque and jet pressure are dynamically calculated, which not only avoids vehicle slippage, but also achieves efficient stripping of ore and low-energy collection, significantly improving the robustness of the operation terminal.

[0025] (8) A full-process "fluid stability" and "flux balance" guarantee system was established, integrating tailwater treatment into the logistics closed loop. This not only achieved graded purification of pollutants, but more importantly, established a flux coupling mechanism between the upstream slurry and the downstream tailwater. This not only ensured the fluid stability of the lifting pipeline, but also maintained the hydrostatic balance of the seabed in the operating area, minimizing the disturbance of the deep-sea pressure field by industrial activities.

[0026] (9) The system has achieved a transformation from "passive maintenance" to "predictive operation and maintenance". Relying on the global knowledge graph of the central collaborative control center, the system can perceive the health status of all elements in real time. Through digital twin technology and historical data training, the system has the ability to provide fault warning and autonomous strategy migration. It can automatically generate preventive scheduling instructions for extreme deep-sea conditions, which greatly reduces the system's failure rate and the cost of operation and maintenance throughout its entire life cycle. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the system of the present invention; Figure 2 This is a schematic diagram illustrating the working principle of the present invention; Among them, 1. Submersible, 2. Coarse particle slurry lifting riser system, 3. Deep-sea heavy-duty mining vehicle, 4. Deep-sea heavy-duty mining vehicle cable and particle transmission pipeline, 5. Surface mother ship platform, 6. Tailwater reinjection and discharge pipe, 7. Photoelectric composite umbilical cable, 8. Marine environment three-dimensional perception module, 9. Intermediate pump station. Detailed Implementation

[0028] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0029] The structures, proportions, and sizes illustrated in the accompanying drawings are merely for illustrative purposes and to aid those skilled in the art in understanding and reading the invention. They are not intended to limit the scope of the invention and therefore have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to size, provided they do not affect the effectiveness or purpose of the invention, should still fall within the scope of the technical content disclosed herein. Furthermore, the terms "upper," "lower," "left," "right," "middle," and "one" used in this specification are merely for clarity and not intended to limit the scope of the invention. Changes or adjustments to their relative relationships, without substantially altering the technical content, should also be considered within the scope of the invention's implementation.

[0030] Figure 2In the system's physical architecture shown, the surface mothership platform 5 is deployed at sea level, serving as the surface support and command center for the entire plant system. The coarse-grained slurry lift-up system 2 is vertically suspended below the surface mothership platform 5, with its bottom connected to an intermediate pump station 9. This intermediate buffer station / pump station 9 acts as a relay node on the seabed, connected to the deep-sea heavy-duty mining vehicle 3 operating on the seabed via a flexible deep-sea heavy-duty mining vehicle cable and particle transmission pipe 4, enabling the collection and transmission of ore particles. A photoelectric composite umbilical cable 7 is laid along the coarse-grained slurry lift-up system 2, transmitting power and fiber optic signals from the surface to various underwater nodes. Around the operating area, a submersible 1 patrols to monitor plumes and conduct environmental surveys; while a marine environment stereoscopic sensing module 8 is installed at the front end or on an independent base of the deep-sea heavy-duty mining vehicle 3, responsible for topographic surveying. The treated tailwater is discharged back to a specific depth layer in the deep sea through the tailwater reinjection and discharge pipe 6.

[0031] like Figure 1 , Figure 2 As shown, the artificial intelligence-based green factory system for deep-sea mining includes: The marine environment three-dimensional perception module 8 is used to reconstruct the three-dimensional topography of the work area and generate a mineral abundance distribution map, providing spatiotemporal benchmarks and resource data for factory production. The intelligent control module of the deep-sea heavy-duty mining vehicle, as the core operating terminal of the system, is based on multi-source heterogeneous data from the operating area and solves the problem through a multi-objective optimization algorithm. This multi-objective optimization algorithm integrates multi-source heterogeneous data such as seabed topography slope, ore body abundance map, flow field vector, and local track slip state as input for the calculation. Its core logic is to construct a dynamic operation cost function, setting the maximization of ore output per unit time as the core optimization objective, and simultaneously coupling the minimization of ineffective energy consumption and the minimization of base disturbance as evaluation constraints. During the calculation process, the algorithm dynamically adjusts the weight coefficients of each objective function in real time according to different working condition commands such as steady-state production, environmental constraints, or energy efficiency optimization issued by the mining system integration and intelligent operation and maintenance module. By performing global optimization calculation in a multi-dimensional optimization space, the final output includes control variables including the expected travel speed of the deep-sea heavy-duty mining vehicle, driving torque, variable track tooth aspect ratio geometric parameters, and hydraulic jet injection pressure and incident angle. This enables trajectory planning and adaptive adjustment of acquisition posture in complex terrain while ensuring ecological red lines. It also outputs the travel parameters and collection operation parameters of the deep-sea heavy-duty mining vehicle, enabling efficient stripping and collection of ore; The ultra-deepwater resource enhancement and tailwater treatment module serves as the system's vertical logistics transport channel, used to establish a multiphase flow transport link for slurry from the seabed to the water surface and a tailwater return link, and to maintain the dynamic material balance between the upward solid-liquid flux and the downward liquid phase flux. The seabed environmental disturbance monitoring and remediation module, as the system's environmental constraint and protection unit, is used to quantify the plume diffusion flux in real time during production operations and proactively perform ecological restoration. The front-end sensing unit of the monitoring module is responsible for monitoring the turbidity and diffusion direction of the plume around the mining operation area in real time and transmitting the environmental data back to the intelligent operation and maintenance center to trigger the "environmental damping" mechanism or guide the release of flocculants.

[0032] The mining system integration and intelligent operation and maintenance module serves as the central collaborative control hub of the system, connecting with the aforementioned modules through a heterogeneous network. Each module achieves information sharing and dynamic optimization through the "artificial intelligence decision-making kernel" within the intelligent operation and maintenance module, forming a closed-loop system of "perception-decision-execution-feedback-repair," thereby enabling efficient and low-disturbance green mining operations in complex deep-sea environments.

[0033] The system's AI decision-making core adopts a distributed deployment architecture. Edge computing nodes are embedded within the titanium alloy pressure-resistant sealed compartments of the deep-sea heavy-duty mining vehicle 3, the intermediate pumping station 9, and the submersible 1; while the long-cycle optimization model is deployed in the central server array on the surface mother ship platform 5. This "cloud-edge collaboration" architecture ensures high real-time response in deep-sea operations.

[0034] The deployment method, parameter acquisition method, and global control logic of the mining system integration and intelligent operation and maintenance module (central collaborative control hub), which serves as the core command unit of the entire system.

[0035] 1. The installation and layout of the mining system integration and intelligent operation and maintenance module unit is not a single physical device, but a distributed control architecture. It is used to illustrate how the artificial intelligence decision-making and collaborative control subsystem, as the global control center, in the deep-sea mining green factory system can achieve cross-module collaborative regulation and system-level closed-loop operation based on multi-source state parameters.

[0036] (1) Central Decision-Making Station (Surface Section): Installed in the central control room of the surface mother ship platform 5, it is equipped with an earthquake-resistant and corrosion-resistant industrial server array. It is mainly responsible for long-cycle production planning and scheduling, energy consumption monitoring of the entire system, and global decision-making based on digital twins.

[0037] (2) Edge computing nodes (underwater component): These are installed in the electronic compartment of the deep-sea heavy-duty mining vehicle 3, the control cabinet of the intermediate pump station 9 of the hoisting system, and the submersible 1. They are encapsulated in a high-pressure resistant (withstanding water pressure at a depth of 5000 meters) sealed titanium alloy body. The underwater edge computing node is integrated in the pressure-resistant electronic compartment of the deep-sea heavy-duty mining vehicle 3. This node uses an embedded high-performance processor to locally process the raw data from the slip rate sensor and lidar, execute millisecond-level obstacle avoidance and power adjustment decisions, and only upload the processed state results to the surface via the optoelectronic composite umbilical cable 7.

[0038] (3) Physical connection: The surface and underwater nodes are connected by an optoelectronic composite umbilical cable 7 (as the backbone network) to achieve low-latency transmission of gigabit-level data. The underwater modules exchange data with each other through a local fiber optic network on the seabed or short-range high-frequency underwater acoustic communication.

[0039] 2. System Parameter Acquisition Methods and Meanings: The mining system integration and intelligent operation and maintenance module aggregates sensor data from the other four functional units in real time through an industrial Ethernet interface, constructing a complete "factory profile." Specific parameters are as follows: (1) The marine environment three-dimensional perception module obtains the following parameters and other marine environment-related parameters.

[0040] Topographic matrix M topo The data transmitted back by multibeam sonar represents geometric obstacles such as seabed slope and pits, and is used to determine path safety. Ore body abundance vector A nodule Determined by optical image recognition, it represents the density of mineral nodules per unit area and determines the upper limit of production efficiency; Flow field vector : Measured by the bottom-mounted ADCP, representing the direction and velocity of the ocean current, used to correct the power compensation of the deep-sea heavy-duty mining vehicle.

[0041] (2) The intelligent control module of the deep-sea heavy-duty mining vehicle obtains the following travel and collection parameters of the deep-sea heavy-duty mining vehicle 3.

[0042] Traveling torque T m : Measured by the drive shaft sensor, representing the resistance of the bottom material to the tracks; slip ratio S slip The risk level of a deep-sea heavy-duty mining vehicle getting stuck in silt is determined by comparing the drive wheel speed with the actual DVL speed. Instantaneous flow rate of the suction port Q mine Measured by an electromagnetic flowmeter, representing the upstream production intensity; Jet pressureP jet : Represents the energy removed by hydraulic stripping.

[0043] (3) Obtain lifting and tailwater treatment parameters from the ultra-deep water resource enhancement and tailwater treatment module.

[0044] riser pressure drop gradient Δ P pipe : Represents the fluid resistance inside the riser, used to monitor the pipe's siltation status; Increase traffic Q lift : Measure the sea surface outlet flux, and compare it with the front end Q mine Benchmarking to solve for the conservation of mass; Tailwater spectral characteristic values σ water : Measured by a spectrometer, representing the concentration of suspended solids and heavy metal ions in the treated effluent.

[0045] (4) Obtain environmental parameters from the seabed environment disturbance monitoring and remediation module.

[0046] plume turbidity field distribution C plume : Measured by a sensor array, representing the degree of environmental pollution caused by the operation; Residual amount of repair agent L chemical Used to predict the endurance of environmental protection systems.

[0047] (5) Obtain functional and capacity-related parameters from the mother ship's power system Total available power quota P total : The global energy boundary for factory operation.

[0048] 3. Overall Control Logic and Implementation Process The mining system integration and intelligent operation and maintenance module serves as the integrated brain of the entire system. Through built-in decision-making algorithms, it performs multi-dimensional analysis of the aggregated global data and executes the following four core control tasks to achieve the steady-state operation of the deep-sea factory: (1) Coordinated Control Logic of Production and Logistics Transportation: This aims to solve the capacity matching problem between front-end collection and intermediate vertical transportation in deep-sea mining operations. The mining system integration and intelligent operation and maintenance module acquires and compares in real time the instantaneous mining volume parameters at the suction port of the deep-sea heavy-duty mining vehicle, the slurry flux parameters at the outlet of the lifting pipeline, and the pressure drop gradient parameters inside the riser. When the system detects an abnormal increase in riser pressure drop or a significantly lower outlet flux than the front-end collection volume, it determines that there is a risk of siltation or blockage in the logistics link. At this time, the mining system integration and intelligent operation and maintenance module performs logical calculations through its internal decision kernel, outputting a travel speed limit command and a ore head suction pump frequency adjustment command to the deep-sea heavy-duty mining vehicle module, reducing the production intensity from the source until the riser pressure drop parameters return to the steady-state range, ensuring the smooth operation of the factory logistics system.

[0049] (2) Energy Efficiency Scheduling Logic under Constrained Power Boundaries: This aims to ensure that the plant system does not experience overload shutdowns within a limited energy supply under independent deep-sea operating conditions. The mining system integration and intelligent operation and maintenance module obtains the total power quota parameters provided by the surface mother ship platform 5 in real time, and simultaneously summarizes the real-time power consumption feedback parameters of each module, such as mining, hoisting, and environmental remediation. When the total power consumption of the entire system approaches the supply boundary, the module executes a priority-based scheduling decision: prioritizing the operation power of the hoisting system to prevent pipeline collapse, and maintaining the basic power of the environmental monitoring system. After decision analysis, the module outputs the upper limit command of the walking motor power and the power reduction command of the high-pressure jet pump to the deep-sea heavy-duty mining vehicle, thereby dynamically compressing unnecessary energy consumption at the production end and keeping the entire system operation within the energy safety red line.

[0050] (3) Production Damping Control Logic Based on Environmental Constraints: Environmental protection indicators are used as hard constraints for factory production. The mining system integration and intelligent operation and maintenance module acquires plume concentration distribution parameters from the environmental monitoring array and ocean current vector parameters measured by the Doppler current profiler in real time. When the real-time turbidity parameter of a certain area approaches the environmental safety threshold, the decision kernel determines that the current production intensity has exceeded the environmental carrying capacity. The module then outputs a work trajectory deviation correction command to guide the deep-sea heavy-duty mining vehicle to avoid highly disturbed sensitive areas, and outputs a command to increase the flow rate of the flocculant dosing pump. Through this "environmental damping" mechanism, environmental feedback is transformed into a negative feedback signal to the production mechanism, realizing the dynamic decoupling of production efficiency and ecological indicators.

[0051] (4) Resource Distribution-Based Output Optimization Scheduling Logic: This aims to improve the unit energy efficiency output ratio of the factory system in complex mining areas. The mining system integration and intelligent operation and maintenance module aggregates the ore body abundance distribution matrix parameters and the real-time energy consumption rate parameters of the current system provided by the sensing module in real time. By comparing the expected output and actual energy efficiency in different areas, when the decision kernel determines that the deep-sea heavy-duty mining vehicle is in a low-yield ore area and the unit energy consumption output is low, the module will automatically output path planning replanning coordinate instructions and operation mode switching instructions to guide the deep-sea heavy-duty mining vehicle to cross the low-yield ore area in an efficient cruising mode and move to the high-abundance rich ore zone detected by the sensing module. By adjusting the operation center in real time, the resource acquisition efficiency of the entire factory system throughout its entire life cycle is optimized.

[0052] The marine environment three-dimensional perception module 8, which serves as the "sensing terminal" of the deep-sea factory, is responsible for acquiring the underlying environmental data of the entire system and converting physical signals into structured instruction parameters that support factory decision-making through a multi-source information fusion algorithm.

[0053] 1. Hardware integration and spatial layout of the module The marine environment three-dimensional perception module 8 adopts an integrated detection platform architecture, deployed on the front-end sensing frame of the deep-sea heavy-duty mining vehicle 3 or on a standalone seabed observation tower. Its components include: a multi-beam sonar array (for acquiring large-scale terrain depth point clouds), a high-frequency scanning lidar (for high-precision reconstruction of short-range obstacles), an acoustic Doppler current profiler (ADCP, for flow field vector detection), and an optoelectronic imaging unit with a constant illumination system. Each sensor is connected to the module's internal edge computing core, enabling synchronous signal acquisition and heterogeneous preprocessing.

[0054] 2. The module for acquiring and defining raw environmental parameters acquires the following underlying sensing parameters in real time, which serve as input variables for subsequent sensing algorithms: D ( x , y ): Topographic depth parameter, the three-dimensional coordinate depth value of the seabed calculated from multibeam sonar echoes; S bs ( x , y ): Backscattering intensity, the energy of sound waves reflected from the seabed interface obtained by sonar, and its value represents the physical hardness characteristics of the seabed. G pixel Optical pixel matrix, which is the real-time grayscale and texture data of the working surface acquired by the photoelectric imaging unit; : Environmental flow vector, which is the water flow velocity and direction parameters at different depths measured by ADCP.

[0055] 3. To address the false alarm problem caused by environmental disturbances in single deep-sea sensors, this module implements a multi-source fusion logic based on spatial feature weighting. The specific implementation steps are as follows: Step 1: Terrain safety calculation. The system extracts depth parameters. D ( x , y Based on the local gradient changes, a slope evaluation operator is established. D The calculation method is as follows: , Step Two: Comprehensive Resource Abundance Estimation. To achieve reliable identification in low-visibility deep-sea environments, this module executes an acoustic-optical linkage verification logic: the system calculates the backscatter intensity... S bs The high-energy region was used to locate the connected domain of the distribution of hard targets on the seabed. Subsequently, in the optical pixel matrix... G pixel Extract the local pixel set corresponding to the connected component and calculate the reflectance contrast operator. K ref : , in, The average gray value of the pixels within the connected component. This represents the average grayscale value of the background region surrounding the connected domain. This operator characterizes the differences between the ore target and the sediment background. The ore body abundance parameters are then obtained through comprehensive calculation. A nod : , in, Ф For calibration coefficients, S area To detect the coverage area; Step 3: Output the global cost map. The system constructs the job cost function. C map This enables a grid-based hierarchical classification of the environment. , in, w 1 、w 2 、w 3 These are weighting factors for safety, production benefits, and energy efficiency costs, respectively. The module will generate... C map The matrix is ​​distributed to the mining system integration and intelligent operation and maintenance module and the deep-sea heavy-duty mining vehicle intelligent control module, guiding factory units to avoid obstacles and concentrate in high-yield areas through cost gradients. Resource feature vectorV ore The adaptive extraction and output logic guides the parameter presetting of the actuators of the deep-sea heavy-duty mining vehicle. This module extracts resource feature vectors through micro-topography calculation. V ore The system uses lidar to acquire the height of micro-undulations in the sampling area and calculates the surface roughness coefficient. R a : , Will R a Acoustic intensity S bs Perform mapping and output feature vectors. V ore =[ A nod , R a , Type id After receiving the vector, the deep-sea heavy-duty mining vehicle determines the roughness based on the data. R a The measurement value is automatically matched with the starting pressure of the jet pump to achieve precise collection of ores of different particle sizes.

[0056] The deep-sea heavy-duty mining vehicle 3 serves as the control logic for the factory's execution terminal. This module is responsible for achieving efficient, low-disturbance, and highly maneuverable autonomous operation in complex machine-soil-flow coupling environments.

[0057] 1. Internal structure and special actuators of the deep-sea heavy-duty mining vehicle module The deep-sea heavy-duty mining vehicle 3 integrates a local controller that executes global instructions. Its core hardware includes: (1) Adaptive variable geometry track mechanism: It has variable track tooth function and dynamically adjusts the height and width of the track teeth through hydraulic actuator; (2) Height-adjustable ore collection device: equipped with lifting push rods and double rows of Coanda jet nozzles, which utilize the wall pressure drop generated by the Coanda effect to strip the ore; (3) Local sensing unit: includes ground Doppler velocimeter (DVL), motor encoder, ground clearance ultrasonic probe and track support roller pressure sensor.

[0058] 2. The parameter acquisition and input interface is defined to achieve precise control. This module acquires and calculates the following three types of parameters in real time: (1) External environmental parameters from the sensing module: including the job cost map C map (Used to guide the direction of travel), resource feature vector V ore (Including ore roughness)R a Substrate type identification Type id ); (2) Global control signals from the mining system integration and intelligent operation and maintenance module: including environmental damping signals S env (Used to constrain work intensity, value range 0-1) and power quota instructions P limit ; (3) Operating status parameters collected by local sensors: including actual ground velocity V a Drive wheel speed n and real-time ground clearance of the ore collection head H real .

[0059] 3. A ground mechanics-based driving performance optimization algorithm is used to prevent the deep-sea heavy-duty mining vehicle from getting stuck in identified low-load-bearing soft-bottom areas by adjusting the track tooth geometry parameters. The specific steps are as follows: Step 1: Calculate the real-time slip ratio s: , Where r is the radius of the drive wheel.

[0060] Step 2: Adjust the track tooth geometry. When s exceeds a preset threshold (e.g., 20%), the local controller calculates the track tooth height-to-middle ratio correction ∆λ based on the slip deviation. HP : , in, α This is the gain coefficient. S set Target slip ratio. Logic output: The deep-sea heavy-duty mining vehicle outputs hydraulic commands to increase the extension height of the track teeth (increase the height-to-distance ratio) and simultaneously expand the width of the track teeth (increase the width-to-distance ratio), thereby increasing the displacement volume of the track teeth in the soft mud to obtain greater physical traction.

[0061] 4. An adaptive control logic for ore collection parameters based on the "Coanda effect" is used to achieve precise stripping of ore and suppress plume generation. The control process is as follows: Step 1: Calculate the jet velocity command U jet The system is based on ore roughness. R a (Characterizing ore grain size and occurrence depth) Establishing a dynamic model: , in, ρ s and ρw The density of the ore and seawater, k The adsorption coefficient of Coanda is given. C l This is the lift coefficient.

[0062] Step Two: Actuator Coordination and Control. The controller, according to... R a The output value of the lifting command will adjust the height of the ore head off the ground. H Lock in the target range (e.g., 1.2). R a -1.8R a This ensures that the Konda fluid effectively encapsulates the ore. Based on the environmental damping signal issued by the mining system integration and intelligent operation and maintenance module... S env Correcting the jet velocity: U ﹡ jet = U jet · S env Logic output: Outputs a frequency converter control signal to adjust the jet pump speed, thereby reducing the plume concentration by decreasing the direct impact of the jet on the bottom sediment while ensuring the stripping force.

[0063] 5. Production Status Feedback and Collaborative Closed-Loop: The intelligent control module of the deep-sea heavy-duty mining vehicle will feed back the actual acquisition intensity parameters (measured by the electromagnetic flowmeter at the suction port) and the track tooth extension and retraction status parameters in real time to the mining system integration and intelligent operation and maintenance module after executing the above algorithm. If the slip ratio... s If the cost cannot be reduced, the mining system integration and intelligent operation and maintenance module will recalculate the actual operation cost map from a global perspective. C map This system forces heavy-duty deep-sea mining vehicles to perform obstacle avoidance or path replanning, achieving a dynamic closed loop of "perception-decision-execution" across the entire plant system. The heavy-duty deep-sea mining vehicles integrate high-pressure ultrasonic or X-ray concentration meters at the slurry outlet pipe to obtain the initial concentration parameters of the discharged slurry in real time. C in .

[0064] Ultra-deepwater resource enhancement and tailwater treatment module: This module achieves efficient circulation of the plant's material flow and stability of the seabed pressure field through intelligent control of the transport pressure and precise reinjection of tailwater.

[0065] 1. Internal Structure and Equipment Composition of the Module The module adopts a vertical closed-loop logistics architecture, and its main components include: (1) Coarse-grained slurry riser system: including a vertical pipe connecting the seabed and the water surface, a distributed pressure gauge along the pipe, and an ultrasonic concentration meter integrated at the pipe inlet (used to obtain the initial slurry concentration). C in ); (2) Multi-stage variable frequency booster pump set: distributed at different depths in the riser to provide vertical conveying power; (3) Tailwater treatment and reinjection unit: includes a cyclone solid-liquid separator, a fine filtration device and a variable frequency reinjection pump that diverts water to the seabed, all deployed on a surface platform.

[0066] 2. Cross-module parameter interaction and acquisition logic: To achieve steady-state operation of the logistics chain, the module acquires and calculates the following parameters in real time: (1) Input parameters: Obtain the instantaneous flow rate of the suction port of the deep-sea heavy-duty mining vehicle. Q mine and the initial concentration measured in real time by an ultrasonic concentration meter C in ; (2) Local sensing parameters: Obtain the real-time total pressure drop measured by the riser friction pressure sensor group. ∆P The lift flow rate measured by the outlet flow meter Q lift And the spectral characteristic value σ of the purified effluent (characterizing the cleanliness of the water body).

[0067] 3. Core control algorithm and flow stability control during execution: Real-time comparison of local controllers ∆P With based ∆P and C in The calculated theoretical safe pressure drop. If ∆P An abnormal fluctuation rate exceeding 15% is identified as localized solid phase settling within the pipe. To increase the flow velocity, the variable frequency pump unit speed is increased, while a speed-limiting signal is sent to the deep-sea heavy-duty mining vehicle via the mining system integration and intelligent operation and maintenance module, reducing the solid phase flux entering the system at its source. Flux coupling balance control (maintaining seabed pressure): Based on the principle of mass conservation, the system calculates the flux command of the downstream reinjection pump in real time. Q return : Q return = Q lift - V ore + δQ make-up in, V ore The volume of solid phase separated from the water surface. δQmake-up This is a compensation amount based on seabed environmental pressure feedback. Control results: By precisely controlling the reinjection flow rate, the amount of water reinjected into the seabed and the amount of water extracted are kept in balance within a dynamic range, preventing negative pressure environments in the work area due to excessive water extraction.

[0068] Submarine Environmental Disturbance Monitoring and Remediation Module This module enables immediate management of environmental disturbances in the work area by monitoring plume intensity in real time and performing active chemical remediation.

[0069] 1. Internal Structure and Equipment Composition of the Module The module is deployed at the edge or downwind of the mining operation area, and its main components include: (1) Stream monitoring array: Consists of distributed optical turbidimeters used to capture the real-time concentration of sediment plumes. C real ; (2) Active repair actuator: includes a chemical storage tank, a high-precision variable frequency dosing pump and a dispenser equipped with atomizing nozzles, used for spraying environmentally friendly flocculants.

[0070] 2. Cross-module parameter interaction and acquisition of logical input parameters: Acquiring real-time ocean current vectors measured by ADCP. Used to pinpoint the core path of plume diffusion and its deployment coordinates; Local sensing parameters: Obtaining real-time plume concentration distribution transmitted back by the monitoring array. C real .

[0071] 3. Core control algorithm and execution process module implement dynamic repair strategy: The governance trigger mechanism establishes environmental safety thresholds through the system. C limit When local turbidity deviation is detected e c = C real - C limit If the value is greater than 0, the repair procedure will be initiated immediately. The dynamic deployment algorithm uses the system based on ocean current vectors. Feedforward correction is applied to the dosage to calculate the flocculant dosing command. Q dose : , in, k 1 For concentration feedback gain, k 2 This is a feedforward adjustment coefficient based on flow rate. Control result: The dosing pump adjusts according to the calculated... Qdose The system drives the delivery device. Through feedforward compensation logic, the system can increase the reagent flux in advance in response to changes in flow velocity, achieving rapid coagulation and forced sedimentation of suspended particles. Simultaneously, this module feeds back environmental exceedance status to the mining system integration and intelligent operation and maintenance module in real time. The mining system integration and intelligent operation and maintenance module then executes global "production damping" logic to reduce the operational intensity of the deep-sea heavy-duty mining vehicle, curbing further plume generation at its source. The environmental remediation module does not operate independently; its delivery trigger threshold and delivery parameters are uniformly controlled by the mining system integration and intelligent operation and maintenance module and serve as constraint inputs for production intensity adjustment, thus directly incorporating environmental remediation activities into the factory's production control closed loop.

[0072] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A deep-sea mining green factory system based on artificial intelligence, characterized in that, include: The marine environment three-dimensional perception module is used to reconstruct the three-dimensional topography of the work area and generate ore body abundance distribution maps, providing spatiotemporal benchmarks and resource data for factory production; The intelligent control module of the deep-sea heavy-duty mining vehicle, as the core operation terminal of the system, is based on multi-source heterogeneous data from the work area. It uses a multi-objective optimization algorithm to solve the problem and output the deep-sea heavy-duty mining vehicle's travel parameters and collection operation parameters, thereby achieving efficient stripping and collection of ore. The ultra-deepwater resource enhancement and tailwater treatment module serves as the system's vertical logistics transport channel, used to establish a multiphase flow transport link for slurry from the seabed to the water surface and a tailwater return link, and to maintain the dynamic material balance between the upward solid-liquid flux and the downward liquid phase flux. The seabed environmental disturbance monitoring and remediation module, as the system's environmental constraint and protection unit, is used to quantify plume diffusion flux in real time during production operations and proactively perform ecological restoration. The mining system integration and intelligent operation and maintenance module, as the central collaborative control hub of the system, is connected to the above modules through a heterogeneous network. Among them, each module achieves information sharing and dynamic optimization through the "artificial intelligence decision kernel" inside the mining system integration and intelligent operation and maintenance module, forming a closed-loop system of "perception-decision-execution-feedback-repair", thereby realizing green mining operations in complex deep-sea environments.

2. The deep-sea mining green factory system based on artificial intelligence as described in claim 1, characterized in that, The mining system integration and intelligent operation and maintenance module is equipped with an artificial intelligence decision kernel, which includes a model reasoning unit, a strategy generation unit, and a parameter distribution interface. The model inference unit serves as the core data entry point, receiving and aggregating sensor data from four modules in real time: three-dimensional perception of the marine environment, intelligent control of deep-sea heavy-duty mining vehicles, resource enhancement and tailwater treatment, and environmental monitoring and remediation. By using the built-in pre-trained model to extract and fuse features from these heterogeneous data, state parameters reflecting the current operating conditions of the factory are constructed, and the potential risks of overload, blockage or environmental exceedance of the system are assessed in real time. The system state evaluation results produced by the model inference unit are transmitted to the strategy generation unit. The strategy generation unit performs multi-objective optimization calculations within the red lines of total power quota and material flow balance based on the hierarchical scheduling logic of "production priority and environmental constraints". The generated logical strategy is transmitted to the parameter distribution interface through the internal interface. The parameter distribution interface is responsible for converting the "strategy" into physical parameters that can be recognized by each work module. By establishing a mapping relationship between the strategy and hardware frequency and torque, the final execution command is distributed to the execution mechanism of each work module, thereby realizing adaptive and self-stabilizing production operations of the entire factory system. The AI ​​decision kernel is deployed in a collaborative manner between edge computing nodes and surface computing nodes. Millisecond-level response decision logic is completed on the underwater edge nodes, while long-cycle optimization and model updates are completed on the surface nodes.

3. The deep-sea mining green factory system based on artificial intelligence as described in claim 1, characterized in that, The marine environment three-dimensional perception module adopts an integrated detection platform architecture, including an acoustic detection device, an optical imaging unit, a flow field monitor, a chemical sensor, and edge computing nodes. The edge computing nodes are embedded in the titanium alloy pressure-resistant sealed chambers of the deep-sea heavy-duty mining vehicle, the intermediate pumping station, and the submersible. Their hardware is uniformly connected to the internal edge computing nodes to achieve synchronous acquisition and preprocessing. The acoustic detection device is responsible for long-distance large-scale topographic reconstruction, the optical imaging unit works with a constant supplementary lighting system for close-range fine identification, the flow field monitor captures environmental flow vectors in real time, and the high-frequency scanning lidar is used for roughness calculation of micro-topography. All hardware components work together to lock the distribution of ore bodies through an "acoustic-optical linkage" verification logic. Each sensor is connected to the edge computing core inside the module via Ethernet to achieve synchronous signal acquisition and heterogeneous preprocessing. The system uses artificial intelligence algorithms to fuse data from multiple sources, achieving real-time perception and anomaly identification of seabed topography, ore body distribution, flow field structure, and environmental parameters. After AI analysis, the perceived data generates a high-resolution three-dimensional reconstruction model of the seabed environment, providing dynamic data support for the deep-sea heavy-duty mining vehicle's path planning and operation optimization.

4. The deep-sea mining green factory system based on artificial intelligence as described in claim 3, characterized in that, The system uses artificial intelligence algorithms to fuse data from multiple sensors. The specific steps are as follows: Step 1: Terrain safety calculation, system extracts depth parameters D ( x , y Based on the local gradient changes, a slope evaluation operator is established. D The calculation method is as follows: ； Step Two: Comprehensive Resource Abundance Estimation. To achieve reliable identification in low-visibility deep-sea environments, this module executes an acoustic-optical linkage verification logic: the system calculates the backscatter intensity... S bs The high-energy region is used to locate the connected domain of the distribution of hard targets on the seabed; Subsequently, in the optical pixel matrix G pixel Extract the local pixel set corresponding to the connected component and calculate the reflectance contrast operator. K ref : ， in, The average gray value of the pixels within the connected component. This is the average gray value of the background region surrounding the connected domain; this operator characterizes the feature differences between the ore target and the sediment background; then, the ore body abundance parameters are obtained through comprehensive calculation. A nod : ， in, Ф For calibration coefficients, S area To detect the coverage area; Step 3: Output the global cost map and construct the job cost function. C map This enables a grid-based hierarchical classification of the environment. ， in, w 1 、w 2 、w 3 The module will generate weighting factors for safety, production benefits, and energy efficiency costs, respectively. C map The matrix is ​​distributed to the mining system integration and intelligent operation and maintenance module and the deep-sea heavy-duty mining vehicle control module, guiding the factory units to avoid obstacles and concentrate in high-yield areas through cost gradients; resource feature vectors V ore The adaptive extraction and output logic guides the parameter presetting of the actuators of the deep-sea heavy-duty mining vehicle. This module extracts resource feature vectors through micro-topography calculation. V ore The system uses lidar to acquire the height of micro-undulations in the sampling area and calculates the surface roughness coefficient. R a : ， Will R a Acoustic intensity S bs Perform mapping and output feature vectors. V ore =[ A nod , R a , Type id After receiving the vector, the deep-sea heavy-duty mining vehicle determines the roughness based on the data. R a The measurement value is automatically matched with the starting pressure of the jet pump to achieve precise collection of ores of different particle sizes.

5. The deep-sea mining green factory system based on artificial intelligence as described in claim 1, characterized in that, The intelligent control module of the deep-sea heavy-duty mining vehicle uses deep learning and reinforcement learning algorithms to achieve terrain recognition, attitude adaptive adjustment, path optimization and conflict avoidance; the operating parameters of the acquisition device, the travel device and the jet device are corrected in real time according to the AI ​​prediction model to achieve stable operation, low-disturbance acquisition and energy efficiency improvement in complex terrain.

6. The artificial intelligence-based green factory system for deep-sea mining as described in claim 5, characterized in that, Terrain recognition is achieved through deep learning and reinforcement learning algorithms, including a driving passability optimization algorithm based on ground mechanics matching. For identified low-load-bearing soft-bottom areas, the deep-sea heavy-duty mining vehicle prevents itself from getting stuck by adjusting the track geometry parameters. The specific steps are as follows: Step 1: Calculate the real-time slip ratio s ， Where r is the radius of the drive wheel; Step 2: Adjust the track tooth geometry. When the real-time slip ratio s exceeds the preset threshold, the local controller calculates the track tooth height-to-middle ratio correction ∆λ based on the slip deviation. HP : ， in, α This is the gain coefficient. S set Target slip ratio; Logic output: The deep-sea heavy-duty mining vehicle outputs hydraulic commands to increase the extension height of the track teeth, increase the height-to-distance ratio, and simultaneously expand the width of the track teeth to increase the width-to-distance ratio. This increases the physical traction force by increasing the displacement volume of the track teeth in the soft mud. The adaptive attitude control achieved through deep learning and reinforcement learning algorithms includes an adaptive control logic for ore collection parameters based on the "Coanda effect." This logic is used to achieve precise ore stripping and suppress plume generation. The control process is as follows: Step 1: Calculate the jet velocity command U jet The system is based on ore roughness. R a Establish a dynamic model: ， in, ρ s and ρ w These are the densities of ore and seawater, respectively. k The adsorption coefficient of Coanda is given. C l Where g is the lift coefficient and g is the acceleration due to gravity. Step Two: The actuators coordinate and control the system; the controller, according to… R a The output value of the lifting command will adjust the height of the ore head off the ground. H Locked within the target range, ensuring that the fluid energy of the Konda system effectively envelops the ore; based on the environmental damping signal issued by the mining system integration and intelligent operation and maintenance module. S env Correcting the jet velocity: U ﹡ jet = U jet · S env ;Logic output: Output frequency conversion control signal to adjust the speed of jet pump, so as to reduce the plume concentration by reducing the direct impact of the jet on the bottom mud while ensuring the stripping force.

7. The deep-sea mining green factory system based on artificial intelligence as described in claim 1, characterized in that, The ultra-deepwater resource enhancement and tailwater treatment module includes a surface mother ship platform, a coarse-grained slurry lifting riser system, a staged pump set, a tailwater purification device, and a fluid circulation control module. It uses artificial intelligence algorithms to adjust the slurry delivery pressure, flow rate, and power distribution in real time, achieving a bidirectional balance between the upward-flowing slurry and the downward-flowing tailwater. The tailwater purification device combines image recognition and spectral analysis technology to perform staged purification of the return water, and uses an AI model to achieve adaptive regulation based on emission concentration feedback, ensuring green emissions.

8. The deep-sea mining green factory system based on artificial intelligence as described in claim 7, characterized in that, The method of adjusting the slurry delivery pressure, flow rate, and power distribution in real time through artificial intelligence algorithms specifically includes core control algorithms and flow stability control during the execution process. Local controller compares total voltage drop in real time ∆P Compared with total pressure drop ∆P and initial concentration C in The calculated theoretical safe voltage drop; if the total voltage drop ∆P If the abnormal fluctuation rate exceeds 15%, it is determined that local solid phase sedimentation has occurred within the pipe. Simultaneously, the variable frequency pump unit speed is increased to raise the flow rate, and a speed limit signal is sent to the deep-sea heavy-duty mining vehicle through the mining system integration and intelligent operation and maintenance module to reduce the solid phase flux entering the system from the source. Flux coupling balance control maintains seabed pressure: the system calculates the flux command of the downstream reinjection pump in real time according to the principle of mass conservation. Q return : Q return = Q lift - V ore + δQ make-up in, Q lift To increase traffic, V ore The volume of solid phase separated from the water surface. δQ make-up The compensation amount is based on the pressure feedback of the seabed environment; the control result is that by precisely controlling the reinjection flow rate, the amount of water reinjected into the seabed and the amount of water pumped out are kept in balance within the dynamic range, preventing the negative pressure environment in the work area from being caused by excessive water extraction.

9. The deep-sea mining green factory system based on artificial intelligence as described in claim 1, characterized in that, The seabed environmental disturbance monitoring and remediation module uses multi-source sensors to collect data on plume velocity, concentration, and diffusion direction, and establishes a plume dynamics model through AI algorithms. The system automatically identifies plume distribution and deposition risk areas, and adjusts the dosage and timing of green flocculant based on AI analysis results to construct a closed loop for ecological restoration of the work area.

10. The artificial intelligence-based green factory system for deep-sea mining as described in claim 9, characterized in that, The method of regulating the dosage and timing of green flocculant based on AI analysis results to construct a closed loop for ecological restoration of the work area includes: a core regulation algorithm and an execution process module that implements dynamic restoration strategies. The governance trigger mechanism establishes environmental safety thresholds through the system. C limit When local turbidity deviation is detected e c = C real - C limit If the value is greater than 0, the repair procedure will be started immediately. C real This indicates real-time concentration; the dynamic delivery algorithm uses ocean current vectors to determine the concentration. Feedforward correction is applied to the dosage to calculate the flocculant dosing command. Q dose : ， in, k 1 For concentration feedback gain, k 2 The feedforward adjustment coefficient is based on the flow rate; the control result: the dosing pump adjusts according to the calculated... Q dose The system drives the dispenser and, through feedforward compensation logic, can increase the reagent flux in advance in response to changes in flow velocity, thereby achieving rapid coagulation and forced sedimentation of suspended particles. At the same time, this module feeds back the environmental exceedance status to the mining system integration and intelligent operation and maintenance module in real time. The mining system integration and intelligent operation and maintenance module then executes the global "production damping" logic to reduce the operating intensity of the deep-sea heavy-duty mining vehicle and curb the further generation of plumes from the source.

Images