Intelligent inspection scheduling method based on 5G big data fusion

Abstract

The invention particularly relates to an intelligent routing inspection scheduling method based on 5G big data fusion, and relates to the technical field of water area drilling guiding and routing inspection scheduling, and the method comprises the steps: coupling geology, drill string mechanics and a guiding tool response model to construct a digital twin system, and combining an LSTM prediction model to quantify a rail deviation risk; and generating a grading task based on the trigger condition, and scheduling inspection resources. According to the method, multiple physics fields are coupled to construct a digital twin system, the LSTM prediction model is matched to realize accurate track prediction in the future 10 m, the track deviation risk is quantified through the three-dimensional Euclidean distance, post deviation correction is upgraded to pre-deviation prevention, an intervention time window is reserved in advance, a high-risk scene is effectively avoided, the strict requirement of underwater drilling engineering for the guiding precision is met, and the method is suitable for large-scale popularization and application. And the core pain points of fuzzy track and delayed deviation correction in the traditional technology are solved.

Description

Technical Field

[0001] This invention relates to the field of drilling guidance and inspection scheduling technology in water areas, and in particular to an intelligent inspection scheduling method based on 5G big data fusion. Background Technology

[0002] In water drilling projects, traditional techniques suffer from two deep-seated biases: First, it is believed that the complex and ever-changing aquatic environment, such as water fluctuations, signal attenuation, and invisible bottom sediment, makes high-precision automatic guidance impossible. Second, inspection resources (such as inspection vessels and buoys) are regarded as independent data acquisition units, and their work is separated from the drilling rig guidance control cycle, thus failing to form a synergistic effect.

[0003] In existing technologies, data acquisition often relies on a single data source, making it susceptible to interference from the aquatic environment. Data transmission suffers from delays and losses, making it difficult to build a high-fidelity data foundation. Multi-source data fusion uses only simple weighted averaging, failing to consider the dynamic error characteristics of each data source, resulting in insufficient trajectory accuracy and reliability. Guiding control is mostly reactive, lacking proactive prediction and intervention. Inspection resource scheduling uses a fixed plan model, unable to respond to the real-time needs of the guiding system. The system lacks self-optimization capabilities, making it difficult to adapt to different geological and environmental conditions. These problems lead to low guiding accuracy, insufficient utilization of inspection resources, and low construction efficiency in aquatic drilling, severely restricting the development of aquatic drilling projects.

[0004] Therefore, there is an urgent need for a method that can break through traditional technological biases and achieve intelligent inspection scheduling and high-precision guidance through the integration of 5G big data, thereby addressing the shortcomings of existing technologies. Summary of the Invention

[0005] The purpose of this invention is to propose an intelligent inspection and scheduling method based on 5G big data fusion in order to solve the above problems.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: The intelligent inspection and scheduling method based on 5G big data fusion includes: Data collection and deployment inside and outside the drill bit and in the environment, coupled with a 5G lossless transmission solution; A multi-source dynamic error model is constructed, and a fusion algorithm is used to generate a three-dimensional drill bit trajectory; A digital twin system is constructed by coupling geological, drill string mechanics, and steering tool response models, and the risk of deviation is quantified by combining an LSTM prediction model. Hierarchical tasks are generated based on trigger conditions, and inspection resources are scheduled.

[0007] Preferably, the deployment of data collection inside and outside the drill bit and the environment, combined with a 5G lossless transmission solution, specifically includes: Drill bit internal attitude acquisition: The measurement-while-drilling tool is embedded 1m from the top of the drill string, with a distance of ≤3m from the drill bit; Inclination measurement range -90°~+90°, tool face angle 0°~360°, sampling frequency 10ms / time; built-in temperature compensation module; External spatial position acquisition of drill bit: The transducer of the ultra-short baseline acoustic positioning system is deployed in the center of the bottom of the inspection vessel, and collects water temperature, salinity and depth data every 200ms, which are then input into the acoustic error model in real time. The MEMS inertial measurement unit is integrated inside the drill bit measurement-while-drilling tool and is mounted coaxially with the gyroscope sensor; The spacing between surface buoys is ≤5km, forming a regional geomagnetic monitoring network; the buoys are equipped with geomagnetic sensors, which are uploaded to the geomagnetic reference field server via 5G network; the magnetometer in the drill bit collects data synchronously, and the azimuth correction is calculated by comparing real-time data with grid map. Environmental data: The multibeam echo sounder transducers on the patrol vessel are installed on both sides of the hull; the echo data is analyzed to automatically identify the type of bottom sediment; the ADCP current profiler is deployed at the rear of the hull and calculates the three-dimensional velocity of the water flow through vector synthesis to generate a heat map of water flow interference. Three 5G macro base stations were deployed in the construction area to form a triangular coverage area.

[0008] Preferably, the construction of the multi-source dynamic error model, employing a fusion algorithm to generate the three-dimensional drill bit trajectory, specifically includes: The error function was fitted using indoor water tank experiments and field measurements: Establish a time-drift coupling model; dynamically correct the drift coefficient based on geomagnetic calibration results; Based on ADCP data, an error model for the interference of water flow on drill bit attitude is established: This error is directly added to the error term of the attitude measurement data; Constructing state vectors ; Process noise matrix The initial value is ; Observation noise matrix Dynamically assign values ​​based on the real-time error values ​​of each data source; The factor graph, which includes data nodes, error factors, and spatiotemporal factors, is solved using the Gauss-Newton method. When the error value of a certain data source exceeds a preset threshold, it is first marked as suspicious data; if three consecutive frames of data exceed the threshold, it is determined to be abnormal data; if the subsequent two frames of data return to normal, it is reconnected for fusion calculation. Output a reliable trajectory. Each trajectory node includes coordinates, attitude parameters, confidence interval, data source weight, and timestamp. Smooth the position coordinates; Generate a 3D trajectory model and a reliability heatmap. The heatmap uses red, yellow, and green to mark different sections, and simultaneously outputs an error analysis report.

[0009] Preferably, the construction of a digital twin system using coupled geological, drill string mechanics, and steering tool response models, combined with an LSTM prediction model to quantify deviation risk, specifically includes: Building a digital twin system: Based on geological exploration data, voxel modeling is adopted. During the construction process, the model is updated every 10m of drilling, and real-time geological identification results are incorporated. Create a 3D solid model of the drill string, with the element type being C3D8R; A linear response model is established based on the command-displacement characteristic curves provided by the tool manufacturer. LSTM prediction model: Collect historical construction data and divide it into training set, validation set, and test set; Input features; numerical features are processed using Min-Max normalization, and categorical features are processed using independent encoding; core features are selected by ranking features by importance using random forest. Input layer, 3-layer LSTM hidden layer, fully connected layer, output layer; Training parameters: 100 iterations, learning rate 0.001, batch size 32.

[0010] Preferably, the method further includes a quantitative assessment of deviation risk: The deviation between the predicted trajectory and the designed trajectory is calculated using three-dimensional Euclidean distance. ; Three threshold value ranges are preset, and each threshold value range matches a risk level. Matching with the value ranges of the three sets of thresholds yields... The corresponding risk levels, which include low risk, medium risk and high risk; Based on geological and hydrological models, the impact range of high-risk areas is predicted; risk cause analysis is output simultaneously.

[0011] Preferably, the step of generating hierarchical tasks based on trigger conditions and scheduling inspection resources specifically includes: Triggering conditions include: risk level is medium or above, the credibility of a certain segment in the trajectory credibility heatmap is less than the preset standard, and the data source is abnormal; Task types include location encryption tasks; environmental resurveying tasks; and geomagnetic calibration tasks. A weighted scoring method was adopted, with the weights allocated as follows: risk level 0.6 + impact on construction progress 0.2 + geological complexity 0.2. High-risk tasks are worth 10 points, medium-risk tasks are worth 7 points, and low-risk tasks are worth 3 points. A final score of ≥9 points is a Level 1 task, 7-8 points is a Level 2 task, and 5-6 points is a Level 3 task.

[0012] Preferably, the system further includes an intelligent scheduler: Improved genetic algorithm parameters: population size 50, number of iterations 30, crossover probability 0.8, mutation probability 0.05; Determine the optimization objective function ; Real-time updates on the status information of inspection vessels / drones, including location, endurance, onboard equipment, current mission progress, and operational capabilities; Plan the optimal route for the inspection vessel and determine the route cost function; The planning process avoids no-navigation zones and maintains safe distances. The scheduling process includes: Receive the task list; match available resources; calculate the cost of each resource to complete the task. The cost of choosing a task Output the scheduling scheme for the resource allocation task with the smallest value; After receiving instructions, the inspection vessel automatically parses the parameters and generates an execution path; it reports that the instructions have been received and the estimated arrival time; during the voyage, it reports its position and progress every 5 minutes; upon reaching the target area, it first performs equipment calibration; it collects data as required and reports the data quality every 10 minutes; after completing the task, it reports that the task has been completed and the total number of data sets collected is [number missing].

[0013] Preferably, the method further includes the following steps: The system updates the trajectory and digital twin system using inspection resource feedback data, verifies the effectiveness by issuing guidance commands after simulation, and achieves system self-optimization through reinforcement learning and adaptive parameter adjustment; specifically including: After the inspection vessel completes its mission, the newly collected data is transmitted back in real time via the 5G network and integrated into the fusion model; the fusion model is recalculated and the trajectory data of the target section is updated; the physical mapping of the digital twin system is updated synchronously. Based on the updated trajectory and prediction results, the system generates guidance instructions.

[0014] Preferably, the method further includes: The database is partitioned and stored according to data type and time dimension; Deep reinforcement learning is used to optimize the prediction model and scheduling algorithm. The state space includes geological conditions, environmental parameters, resource status, and guidance error, while the action space includes the adjustment of prediction model parameters, selection of scheduling strategy, and reward function. The system automatically adjusts the thresholds of key parameters; the adjustment is based on the statistical results of construction data over the past 30 days to ensure that the parameters are adapted to different scenarios. The system monitors the status of each module in real time, and automatically triggers the diagnostic process if a module failure is detected.

[0015] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. This invention constructs a digital twin system by coupling multiple physics fields, and uses an LSTM prediction model to achieve accurate prediction of the trajectory in the future within 10m. It quantifies the risk of deviation by using three-dimensional Euclidean distance, upgrades post-departure correction to pre-departure prevention, reserves an intervention time window in advance, effectively avoids high-risk scenarios, meets the stringent requirements of underwater drilling engineering for guidance accuracy, and solves the core pain points of traditional technology such as fuzzy trajectory and delayed correction.

[0016] 2. This invention constructs an intelligent scheduling system by employing an improved genetic algorithm and path planning, transforming inspection resources from passively operating according to a plan to proactively responding to demand. The scheduling algorithm aims to minimize task completion time and resource consumption, dynamically allocating tasks based on environmental constraints and resource status, thereby shortening inspection task completion time and reducing ineffective resource consumption. Through data feedback closed-loop trajectory updates and a digital twin system, the model and parameters are adaptively optimized via reinforcement learning, making the system adaptable to different geological and water flow scenarios. Attached Figure Description

[0017] Further details, features, and advantages of this application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which: Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0018] Several embodiments of this application will now be described in more detail with reference to the accompanying drawings to enable those skilled in the art to implement this application. This application may be embodied in many different forms and for various purposes and should not be limited to the embodiments set forth herein. These embodiments are provided to make this application thorough and complete, and to fully convey the scope of this application to those skilled in the art. The embodiments described do not limit this application.

[0019] Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It will be further understood that terms such as those defined in commonly used dictionaries shall be interpreted as having a meaning consistent with their meaning in the relevant field and / or the context of this specification, and shall not be interpreted in an idealized or overly formal sense unless expressly defined herein.

[0020] Example 1

[0021] Its specific implementation method is combined with the appendix Figure 1 Please provide a detailed explanation.

[0022] Appendix Figure 1 The flowchart of the intelligent inspection and scheduling method based on 5G big data fusion provided for the embodiments of the present invention shows the complete steps from data collection and deployment inside and outside the drill bit and the environment to achieving system self-optimization through reinforcement learning and adaptive parameter adjustment.

[0023] In this embodiment, it includes: By deploying three-dimensional anti-interference data collection inside and outside the drill bit and in the environment, and combining it with a 5G lossless transmission solution (including network, protocol, and fault tolerance mechanism), a full-dimensional high-fidelity uninterrupted data foundation is constructed. Specifically, it includes: Drill bit internal attitude acquisition: The measurement while drilling tool (fiber optic gyroscope type) is embedded 1m from the top of the drill string, with a distance of ≤3m from the drill bit, to avoid measurement deviations caused by drill string deformation; the tool shell is made of titanium alloy, with a pressure resistance rating of ≥100MPa and a waterproof rating of IP68, suitable for underwater operation depths of 0-100m. Acquisition parameters: tilt angle measurement range -90°~+90°, tool face angle 0°~360°, sampling frequency 10ms / time; built-in temperature compensation module (operating temperature -20℃~80℃) to eliminate the influence of water temperature changes on gyroscope accuracy; It adopts an RS485 industrial bus interface with a transmission rate of 115200bps and a data frame format of “start bit 1 + data bits 8 + parity bit 1 + stop bit 1” to ensure stable data transmission. External spatial position acquisition of drill bit: Acoustic positioning: The transducer of the ultra-short baseline acoustic positioning system is deployed in the center of the bottom of the inspection vessel at a draft of 1.5m to avoid signal obstruction by the hull structure; the operating frequency is 10-30kHz, the ranging range is 0-500m, and the ranging accuracy is ±0.2%×slant distance; it is equipped with a temperature, salinity, and depth sensor (CTD) to collect water temperature (±0.1℃), salinity (±0.1‰), and depth (±0.01m) data every 200ms and input them into the acoustic error model in real time; The MEMS inertial measurement unit (IMU) is integrated inside the drill bit measurement-while-drilling tool and is coaxially mounted with the gyroscope sensor; a 30-second forced calibration mechanism is set. When calibration is triggered, the IMU pauses independent calculations and receives geomagnetic reference data to correct azimuth drift. The spacing between surface buoys is ≤5km, forming a regional geomagnetic monitoring network. The high-precision geomagnetic sensors (resolution ≤0.1nT) on the buoys collect data every 100ms and upload it to the geomagnetic reference field server via 5G network to generate a 10m resolution regional geomagnetic grid map. The magnetometer in the drill bit collects data synchronously, and the azimuth correction is calculated by comparing the real-time data with the grid map (correction formula: corrected azimuth = original IMU azimuth + (grid geomagnetic value - real-time geomagnetic value) × calibration coefficient 0.001° / nT). Calibration is completed every 15 seconds. Environmental data: The multibeam echo sounder transducers on the patrol vessel are installed on both sides of the hull at a 30° angle to the centerline, covering a width ≥ twice the water depth. Echo intensity analysis of the depth data automatically identifies the bottom sediment type (echo intensity >60dB indicates rock, 40-60dB indicates sand, <40dB indicates silt). An ADCP current profiler is deployed at the stern of the hull, sampling at least 10 layers with a sampling time of 1 second per layer. Vector synthesis is used to calculate the three-dimensional velocity of the water flow (eastward, northward, and vertical), generating a heat map of water flow interference (red indicates high interference zone, velocity >2m / s; green indicates low interference zone, velocity <0.5m / s). Three 5G macro base stations were deployed in the construction area to form a triangular coverage and ensure signal strength ≥ -85dBm; SA standalone networking mode was adopted, and edge computing nodes (MEC) were enabled and deployed in shore-based data centers ≤ 10km away from the construction area to reduce data transmission latency; It adopts UDP protocol and RTP real-time transmission encapsulation, with a single frame data size of 512KB, including 23 types of data fields such as attitude, positioning, and environment; it sets dual identifiers of data frame sequence number and timestamp, with timestamp accuracy ≤1ms (based on GPS time synchronization) to ensure spatiotemporal synchronization of heterogeneous data.

[0024] A multi-source dynamic error model is constructed, and a deep coupling fusion algorithm combining adaptive Kalman filtering and factor graph optimization is adopted to generate accurate, reliable, and visualized 3D drill bit trajectories.

[0025] Specifically, it includes: The error function was fitted using indoor water tank experiments and field measurements: ; in, For transmission distance, Water temperature Salinity The error model updates the error value every 10 seconds based on real-time environmental data, representing the water flow velocity. Establish a time-drift coupling model: ; in, For continuous working hours, This refers to the acceleration of the drill string; Based on the geomagnetic calibration results, the drift coefficient is dynamically corrected (drift coefficient after calibration = original coefficient × (1 - calibration reliability)). Based on ADCP data, an error model for the interference of water flow on drill bit attitude is established: ; in, For water flow velocity, The angle between the water flow and the drill string axis; This error is directly added to the error term of the attitude measurement data; Constructing state vectors : ; in, These represent the east, north, and vertical coordinates of the drill bit in three-dimensional space (usually using the WGS-84 coordinate system), respectively, and are used to describe the spatial position of the drill bit. Inclination angle: The angle between the drill bit axis and the horizontal plane, reflecting the degree of inclination of the drill bit; Facing angle: The azimuth angle of the drill bit in the horizontal plane, used to determine the direction of the drill bit; These represent the drill bit's speed in the east, north, and vertical directions, respectively, and are used to describe the drill bit's motion state; Process noise matrix The initial value is The output is dynamically adjusted every 5 seconds based on the error model (when the error increases). When the error decreases after being magnified by 1.2 times. (Shrink to 0.8 times); Observation noise matrix Dynamically assign values ​​based on the real-time error values ​​of each data source; The factor graph includes data nodes (original data from each data source), error factors (constraints based on the error model), and spatiotemporal factors (coordinate system transformation constraints). It is solved using the Gauss-Newton method, with ≤20 iterations and a convergence threshold of residual sum of squares <1e-6. Local optimization is completed every 100ms, and global optimization is completed every 1 second to eliminate accumulated errors. When the error value of a certain data source exceeds the preset threshold (acoustic positioning > 50cm, IMU > 1°, geomagnetic > 5nT), it is first marked as suspicious data; if three consecutive frames of data exceed the threshold, it is determined to be abnormal data, and the data source is suspended from participating in fusion; if the subsequent two frames of data return to normal, it is reconnected to the fusion calculation, and the error model parameters are updated at the same time. Output a reliable trajectory, with a trajectory node every 50cm, including coordinates ( ), attitude parameters (tilt angle ±0.1°, face angle ±0.5°), confidence interval, data source weight, timestamp; A 5-point sliding window averaging method (window size = 5 nodes) is used to smooth the position coordinates and avoid trajectory jitter caused by data fluctuations. Generate a 3D trajectory model and a confidence heatmap. The heatmap marks different segments with red (confidence <80%), yellow (80%-90%), and green (>90%). Simultaneously output an error analysis report, explaining the main sources of error in each segment (e.g., "The error in segment X mainly comes from acoustic signal attenuation").

[0026] A digital twin system is constructed by coupling geological, drill string mechanics, and guidance tool response models. Combined with an LSTM prediction model, the risk of deviation is quantified to achieve AI-driven predictive guidance and pre-departure prevention. Specifically, it includes: Building a digital twin system: Based on previous 1:5000 geological exploration data, voxel modeling (voxel size 0.1m×0.1m×0.1m) was adopted, including parameters such as stratigraphic lithology (sandstone / mudstone / granite), compressive strength (5-100MPa), porosity (5%-30%), and friction coefficient (0.2-0.8); the model error was ≤5% verified by borehole core data; during construction, the model was updated every 10m of drilling, incorporating real-time geological identification results; A three-dimensional solid model of the drill string was established using ABAQUS finite element software secondary development. The element type was C3D8R (8-node hexahedral element) with a mesh density of 0.5m / element. Input parameters include drill pressure (5-20kN), rotational speed (50-200rpm), torque (10-50kN·m), and formation reaction force (calculated based on a geological model); real-time output includes drill string deformation (≤5cm), drill bit forces (axial force, radial force, tangential force), and drill bit attitude changes; Based on the command-displacement characteristic curves provided by the tool manufacturer, a linear response model is established: ;in, This represents the change in the facing angle. To control the voltage, For the response coefficient, This is the error term; Model coupling logic: The geological model outputs formation parameters to the drill string mechanics model; the drill string mechanics model outputs drill bit stress and deformation data to the steering tool response model; the steering tool response model feeds back the actual attitude adjustment effect to the digital twin system; the coupling update frequency of the three is 10Hz to ensure real-time synchronization with the physical world; LSTM prediction model: Collect 1000+ sets of historical construction data, covering 5 geological types, 3 water flow conditions, and 4 drilling parameter combinations, with a total data volume of ≥100GB; divide the data into training set, validation set, and test set in a 7:2:1 ratio; Input features include current trajectory coordinates, attitude, drilling pressure, rotation speed, torque, formation compressive strength, water flow velocity, data reliability, etc.; Min-Max normalization (normalization range [0,1]) is used to process numerical features, and one-hot encoding is used to process classification features (such as substrate type); core features are selected by ranking the importance of features through random forest. Input layer (15 neurons), 3 LSTM hidden layers (64 neurons per layer, activation function tanh, dropout rate 0.2), fully connected layer (32 neurons, activation function ReLU), output layer (6 neurons, corresponding to the next 5m and 10m). coordinate); Training parameters: 100 iterations, learning rate 0.001 (using Adam optimizer, learning rate decays by 10% every 20 iterations), batch size 32; A prediction is triggered every 1m of drilling, with the core feature data within the current 10m range input; the model outputs the predicted trajectory coordinates for the next 5m and 10m, with prediction accuracy: ≤15cm within 5m and ≤30cm within 10m; if the prediction error exceeds the threshold (5m>20cm), the model is automatically recalibrated (incorporating the latest 5 sets of actual trajectory data).

[0027] It also includes a quantitative assessment of deviation risk: The deviation between the predicted trajectory and the designed trajectory is calculated using three-dimensional Euclidean distance. : ; in, This represents the three-dimensional deviation between the predicted trajectory and the designed trajectory, i.e., the straight-line distance between the two in space; , , These represent the east, north, and vertical coordinates of the predicted trajectory in the WGS-84 coordinate system; , , These represent the east, north, and vertical coordinates of the design trajectory in the WGS-84 coordinate system. Three threshold value ranges are preset, and each threshold value range matches a risk level. Matching with the value ranges of the three sets of thresholds yields... The corresponding risk levels, which include low risk, medium risk and high risk; Based on geological and water flow models, the impact range of high-risk areas is predicted (e.g., "The current high-risk area will affect the subsequent 3m drilling, with an estimated cumulative deviation of 70cm"); risk cause analysis is output simultaneously (e.g., "Due to a sudden increase in the compressive strength of the strata in section X (from 30MPa to 80MPa), the drill bit is deflected under stress, causing the risk of deviation").

[0028] Based on triggering conditions such as guidance risk and data credibility, hierarchical tasks are generated. By improving the genetic algorithm and path planning to schedule inspection resources, resources can be replenished on demand and precisely empowered. Specifically, it includes: Triggering conditions include: risk level is medium or above, the credibility of a certain segment in the trajectory credibility heatmap is less than the preset standard, and the data source is abnormal; Task type and parameters: Location encryption task: target area coordinate range, acquisition frequency (10Hz), duration (30 minutes), target credibility (≥95%), required equipment (inspection vessel equipped with an ultra-short baseline acoustic positioning system).

[0029] Environmental retesting task: target section, retesting parameters (water flow velocity, sediment type), number of retests (3 times), data accuracy requirements (flow velocity error ≤ 0.1 m / s).

[0030] Geomagnetic calibration task: buoy movement coordinates, calibration range, number of calibrations (5 times), and accuracy of the target geomagnetic reference field (±0.5nT).

[0031] A weighted scoring method was adopted, with the weights allocated as follows: risk level 0.6 + impact on construction progress 0.2 + geological complexity 0.2. High-risk tasks are worth 10 points, medium-risk tasks are worth 7 points, and low-risk tasks are worth 3 points. Impact on construction progress (2 points for each hour of delay), geological complexity (1 point for each rock formation); a final score of ≥9 points is a Level 1 task (to be executed immediately), 7-8 points is a Level 2 task (to be executed within 30 minutes), and 5-6 points is a Level 3 task (to be executed within 1 hour).

[0032] It also includes an intelligent scheduler: Improved genetic algorithm parameters: population size 50, number of iterations 30, crossover probability 0.8 (using single-point crossover), mutation probability 0.05 (using uniform mutation); Determine the optimization objective function : ;in, For task completion time, For resource consumption, As a task priority, It is the minimum target value; , , These are the corresponding weighting factors; Real-time updates of the status information of inspection vessels / drones, including location (latitude and longitude ±1m), endurance (remaining battery ≥30% is usable), onboard equipment (acoustic positioning / geomagnetic sensor / multibeam echo sounder), current mission progress (completion rate ≥90% is schedulable), and operational capabilities (e.g., can operate in wind and wave levels ≤4). Plan the optimal route for the inspection vessel and determine the route cost function: ;in, The straight-line distance. At the cost of the environment, This is the environmental weighting coefficient; During the planning process, avoid no-navigation zones (such as reef areas and aquaculture areas) and reserve a 100m safety distance; The scheduling process includes: Receive the task list (including priority); match available resources (meeting device requirements and having sufficient battery life); calculate the cost of each resource to complete the task. The cost of choosing a task The task with the smallest resource allocation value; output the scheduling scheme (resource ID, task ID, path, completion time limit); Command issuance and execution: After receiving instructions, the inspection vessel automatically parses the parameters and generates an execution path; it reports that the instructions have been received and the estimated arrival time; during the voyage, it reports its position and progress every 5 minutes; upon reaching the target area, it first performs equipment calibration; it collects data as required and reports the data quality every 10 minutes; after completing the task, it reports that the task has been completed and the total number of data sets collected is [number missing]. If the operation encounters wind and waves ≥ level 5, equipment failure, or other situations during execution, the operation should be immediately suspended, and an emergency trigger should be reported, with the reason being: wind and waves exceeding the standard. After receiving the report, the scheduler will re-match backup resources and generate a new scheduling plan.

[0033] The system updates the trajectory and digital twin system using inspection resource feedback data, verifies the effectiveness by issuing guidance commands after simulation, and achieves system self-optimization through reinforcement learning and adaptive parameter adjustment; specifically including: After the inspection vessel completes its mission, the newly collected high-precision data (such as encrypted acoustic positioning data with an accuracy of ±5cm) is transmitted back in real time via the 5G network and integrated into the fusion model within 1 second; the fusion model is recalculated and the trajectory data of the target section is updated, increasing the positioning confidence from 92% to over 96%; the physical mapping of the digital twin system is updated simultaneously (update frequency 10Hz). Guiding command generation and verification: Based on the updated trajectory and prediction results, the system generates guiding commands (such as "adjust the facing angle to 205°, maintain drilling pressure at 12kN, and rotation speed at 150rpm"). After the commands are generated, they are first verified by simulation in the digital twin system (simulation duration is 5 seconds). If the simulation results show that the deviation is ≤5cm, the command is issued; if the deviation is >5cm, the command parameters are re-optimized.

[0034] Command execution and closed-loop verification: In manual mode, commands are pushed to the driller through a visual interface and executed after confirmation by the driller; in automatic mode, commands are sent to the drilling rig control system through an API interface (RESTful protocol), with a command execution delay of ≤3s; within 5 seconds after execution, the drilling measurement tool collects new attitude data to verify the adjustment effect (e.g., face angle adjustment error ≤0.3° is considered qualified); if unqualified, the adjustment command is regenerated until the requirements are met.

[0035] It uses a distributed PostgreSQL database, with data partitioned and stored according to data type (raw data / fusion results / instruction data) and time dimension; the data retention period is ≥3 years, and it includes 50+ dimensions (such as raw data, error value, prediction result, scheduling scheme, execution effect, and deviation); it supports multi-dimensional retrieval by geological type, environmental conditions, and task type. Deep reinforcement learning (DRL) is used to optimize the prediction model and scheduling algorithm. The state space includes geological conditions, environmental parameters, resource status, and guidance error; the action space includes prediction model parameter adjustment and scheduling strategy selection; and the reward function is: ;in, For the deviation amount, For task completion time, For resource consumption; The system automatically adjusts the threshold values ​​of key parameters, such as risk classification standards under different geological conditions and the initial weight values ​​of data fusion (the initial weight of acoustic positioning in deep water areas is increased from 0.3 to 0.4); the adjustment is based on the statistical results of construction data over the past 30 days to ensure that the parameters are adapted to different scenarios. Real-time monitoring of the status of each module (data acquisition, transmission, fusion, prediction, scheduling). If a module fault is detected (such as data transmission interruption), the diagnostic process is automatically triggered: locate the faulty node (such as base station signal interruption); activate the backup plan (switch to the backup base station); calibrate the data after recovery (such as resynchronizing trajectory data after transmission is restored); the fault recovery time is ≤30 seconds to ensure that the closed loop is not interrupted.

[0036] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0037] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0038] It should be noted that, in this document, the use of relational terms such as "first" and "second" is merely for distinguishing one entity or operation from another, and does not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0039] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0040] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0041] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0042] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0043] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0044] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0045] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. The intelligent inspection scheduling method based on 5G big data fusion, characterized in that, Comprise: Deployment of data acquisition inside and outside the drill bit and the environment, combined with 5G lossless transmission scheme; Construction of multi-source dynamic error model, generation of three-dimensional drill bit trajectory by fusion algorithm; Coupling of geological, drill string mechanics and guide tool response model to construct digital twin system, combined with LSTM prediction model to quantify the risk of off-track; Generation of hierarchical tasks based on trigger conditions, through scheduling of inspection resources. 2.The 5G big data fusion-based intelligent inspection scheduling method according to claim 1, characterized in that, Deployment of data acquisition inside and outside the drill bit and the environment, combined with 5G lossless transmission scheme, specifically comprising: Drill bit internal attitude acquisition: Measurement-while-drilling tool is embedded in the top of the drill string 1m away from the drill bit, with a distance of ≤3m; Dip angle measurement range -90°~+90°, tool face angle 0°~360°, sampling frequency 10ms / time; built-in temperature compensation module; Drill bit external space position acquisition: Transducers of the ultra-short baseline acoustic positioning system are deployed on the central bottom of the inspection ship, collecting water temperature, salinity and depth data every 200ms, which are input into the acoustic error model in real time; MEMS inertial measurement unit is integrated inside the measurement-while-drilling tool of the drill bit, coaxially installed with the gyro sensor; Water surface buoys are spaced ≤5km apart, forming a regional geomagnetic monitoring network; the buoys are equipped with geomagnetic sensors, which are uploaded to the geomagnetic reference field server through 5G network; the internal magnetometer of the drill bit synchronously collects data, which are compared with the grid map in real time to calculate the azimuth correction amount; Environmental data: The transducers of the multibeam echo sounder carried by the inspection ship are installed on both sides of the ship; the echo intensity analysis automatically identifies the bottom type based on the depth data; the ADCP current profiler is deployed at the back of the ship bottom, which calculates the three-dimensional velocity of the water flow by vector synthesis to generate the water flow interference thermal map; Three 5G macro base stations are deployed in the construction area to form a triangular coverage. 3.The 5G big data fusion-based intelligent inspection scheduling method of claim 1, wherein, Construction of multi-source dynamic error model, generation of three-dimensional drill bit trajectory by fusion algorithm, specifically comprising: Through indoor pool experiment and field measurement, error function is fitted: Establishment of time-drift coupling model; combined with geomagnetic calibration results, dynamically correct the drift coefficient; Based on ADCP data, establish the interference error model of water flow on drill bit attitude: The error is directly superimposed on the error term of attitude measurement data; Constructing state vector ; Process noise matrix Initial value is ; Observation noise matrix Dynamic assignment according to real-time error value of each data source Gaussian Newton method is used to solve the factor graph containing data nodes, error factors and space-time factors; When the error value of a certain data source exceeds the preset threshold, it is first marked as suspicious data; if 3 consecutive data exceed the threshold, it is determined as abnormal data; if the subsequent 2 frames of data return to normal, it is reconnected to the fusion calculation; Output the credible trajectory, each trajectory node contains coordinates, attitude parameters, confidence interval, data source weight and timestamp; Smooth the position coordinates; Generate three-dimensional trajectory model and credibility thermal map, the thermal map is marked with red, yellow and green for different sections, and the error analysis report is output synchronously. 4.The 5G big data fusion-based intelligent inspection scheduling method of claim 1, wherein, Coupling of geological, drill string mechanics and guide tool response model to construct digital twin system, combined with LSTM prediction model to quantify the risk of off-track, specifically comprising: Digital twin system construction: Based on geological exploration data, voxel modeling is adopted, the model is updated every 10m during construction, and real-time geological identification results are integrated; Establish a three-dimensional entity model of the drill string, with C3D8R as the unit type; Based on the instruction-displacement characteristic curve provided by the tool manufacturer, a linear response model is established; LSTM prediction model: Collect historical construction data, divide into training set, validation set, test set; Input features; numerical features are processed by Min-Max normalization, and categorical features are processed by independent encoding; through random forest feature importance sorting, core features are selected; Input layer, 3-layer LSTM hidden layer, full connection layer, output layer; Training parameters: 100 iterations, learning rate 0.001, batch size 32. 5.The 5G big data fusion-based intelligent inspection scheduling method according to claim 4, characterized in that, It also includes bias risk quantitative evaluation: A three-dimensional Euclidean distance is used to calculate the deviation of the predicted trajectory from the design trajectory ; preset three groups of threshold value ranges, each group of threshold value ranges matches a risk level, and the obtained is matched with the three groups of threshold value ranges, and a corresponding risk level is obtained corresponding risk level, wherein the risk level includes low risk, medium risk, and high risk. Based on the geological model and the water flow model, the influence range of the high-risk area is predicted; Synchronous output risk reason analysis. 6.The 5G big data fusion-based intelligent inspection scheduling method according to claim 1, characterized in that, Based on the trigger condition, generate a hierarchical task, and through the scheduling of inspection resources, including: Trigger conditions include: risk level is in medium risk and above, certain section of the trajectory reliability in the trajectory reliability heat map is less than the preset standard, data source anomaly; Task types include positioning encryption tasks; environmental re-measurement tasks; geomagnetic calibration tasks; Using the weighted scoring method, the weight distribution is risk level 0.6 + construction progress impact 0.2 + geological complexity 0.2; The base score of high-risk tasks is 10 points, medium-risk tasks is 7 points, and low-risk tasks is 3 points; The final score is ≥ 9 for first-level tasks, 7-8 for second-level tasks, and 5-6 for third-level tasks. 7.The 5G big data fusion-based intelligent inspection scheduling method according to claim 6, characterized in that, It also includes an intelligent scheduler: Improved genetic algorithm parameters: population size 50, iteration number 30, crossover probability 0.8, mutation probability 0.05; determining an optimization objective function ; Real-time update of inspection ship / unmanned aerial vehicle state information, including position, endurance time, carried equipment, current task progress, and operation capacity; Planning the optimal path of the inspection ship, determining the path cost function; Planning Avoiding forbidden areas and reserving safety distances during the process; The scheduling process includes: Receiving a task list; matching available resources; calculating the cost of each resource to complete the task ; selecting the task with the lowest cost resource allocation, outputting a scheduling scheme; After receiving the instruction, the inspection ship automatically analyzes the parameters, generates the execution path, and feeds back that the instruction has been received and the estimated arrival time; every 5 minutes during navigation, it feeds back the position and progress; after arriving at the target area, it first performs equipment calibration; it collects data as required, feeding back data quality every 10 minutes; after completing the task, it feeds back that the task has been completed, and a total of data groups are collected. 8.The 5G big data fusion-based intelligent inspection scheduling method according to claim 1, characterized in that, It also includes the following steps: Update the trajectory and digital twin system with the feedback data from the inspection resources, verify the effect of the guidance instructions through simulation, and realize system self-optimization through reinforcement learning and parameter self-adaptation; including: After the inspection ship completes the task, the newly collected data is transmitted back in real time through the 5G network and accessed to the fusion model; the fusion model recalculates and updates the trajectory data of the target section; the physical mapping of the digital twin system is updated synchronously; According to the updated trajectory and prediction results, the system generates guidance instructions. 9.The 5G big data fusion-based intelligent inspection scheduling method according to claim 8, characterized in that, It also includes: The database is stored in partition according to data type and time dimension; Optimize the prediction model and scheduling algorithm using deep reinforcement learning, the state space includes geological conditions, environmental parameters, resource state, and guidance error, the action space includes prediction model parameter adjustment and scheduling strategy selection, and the reward function; The system automatically adjusts the key parameter threshold; the adjustment is based on the construction data statistics of the past 30 days to ensure that the parameters adapt to different scenarios; Real-time monitoring of each module state, if detected module failure, automatically trigger diagnostic process.

Images