Resource configuration and optimal scheduling method for supervision resources

Abstract

The invention provides a resource configuration and optimal scheduling method for supervision resources, and the method comprises the steps: building a cable laying three-dimensional visual model, enabling the attributes of supervision personnel to be matched with a supervision region, generating an initial configuration scheme, and carrying out the correlation analysis of a historical electrical engineering quality problem and the attributes of the supervision personnel, thereby achieving the optimal scheduling of the supervision resources. And dynamic adjustment of supervision personnel configuration is realized. And the construction process is monitored in real time by using the Internet of Things technology, and early warning and positioning are carried out on abnormal conditions. Quality of key processes is judged through image analysis, and professional supervisors are matched for disposal according to severity of problems. And skill training is carried out in combination with a virtual reality technology, and a personnel scheduling strategy is optimized. And through big data analysis, continuous evaluation and dynamic adjustment of supervision personnel configuration are carried out to form closed-loop optimization. According to the method, the supervisor and the supervision area are highly matched, the supervisor can quickly and efficiently complete the supervision work, the construction efficiency of the cable laying project is effectively improved, and the quality of the cable laying project is improved.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for resource allocation and optimal scheduling of supervision resources. Background Art

[0002] In cable laying projects, the scientific allocation and optimized scheduling of supervisors present technical challenges. The complex and variable nature of cable laying routes presents significant challenges for monitoring. Pipeline layouts and other factors vary significantly across different areas, resulting in highly nonlinear characteristics in cable laying routes. Supervisors must conduct detailed analysis of the characteristics of each route segment and dynamically adjust monitoring strategies based on actual conditions, placing high demands on their professional capabilities. There is also a conflict between the supervisors' professional qualifications and the resource allocation within the supervision area. Because cable laying involves multiple specialized fields, such as electrical, civil engineering, and mechanical engineering, supervisors of different specialties have varying technical backgrounds and working methods. How to rationally allocate supervisors of different specialties based on the specific conditions of the supervision area and ensure close collaboration and synergy on-site is a pressing issue. Summary of the Invention

[0003] The present invention provides a method for resource allocation and optimal scheduling of supervision resources, which mainly includes: Acquire cable laying path data, build a 3D visualization model, match the supervisor attributes with the supervision area in the model, and realize dynamic adjustment of supervisor configuration based on the correlation analysis between historical electrical engineering quality issues and supervisor attributes; in the hidden parts of the cable laying path, collect environmental parameters, equipment status and personnel location data of the cable laying site in real time, visually monitor the construction process, send warnings to the appropriate supervisors in case of abnormalities, and highlight the abnormal locations in the 3D model; analyze the images and videos of cable fireproof sealing and cable insulation resistance testing at the construction site, extract key information of sealing materials and resistance testing instruments, and compare the key information with the 3D model. The cable properties at the corresponding positions in the 3D visualization model are compared to determine whether the construction quality meets the specifications and standards; if the fireproof sealing material is found to be inconsistent or the insulation resistance test is abnormal, the appropriate supervisor will be assigned to handle the problem according to the severity of the problem, and the review situation will be fed back to the 3D visualization model and the cable status will be updated; the update instructions of the cable laying construction will be received, the update content will be semantically understood, and the feasibility of the update will be evaluated based on the skill level of the supervisor; the data of the cable laying construction will be digitized and recorded, the review opinions and operation logs of the supervisors will be recorded, and the supervision records of the changes in the cable laying construction will be included in the evidence, while the configuration of the supervisors will be continuously evaluated and dynamically adjusted to form a closed-loop optimization.

[0004] The technical solution provided by the embodiment of the present invention may have the following beneficial effects: The present invention discloses a method for resource allocation and optimal scheduling of supervision resources. The method matches the attributes of supervisors with the supervision areas by constructing a three-dimensional visualization model of cable laying, and generates an initial configuration plan. The Internet of Things technology is used to monitor the construction process in real time, and abnormal conditions are warned and located. The quality of key processes is judged through image analysis, and professional supervisors are matched to deal with the problems according to the severity of the problems. Skill training is carried out in combination with virtual reality technology to optimize personnel scheduling strategies. The decision tree method is used to evaluate the feasibility of design changes, and digital evidence storage and responsibility tracing are realized. The present invention continuously evaluates and dynamically adjusts the configuration of supervisors through big data analysis to form a closed-loop optimization, which effectively improves the intelligence level and quality control capabilities of electrical engineering supervision. Because the supervisors are highly matched with the supervision areas, the supervisors can complete the supervision work quickly and efficiently, effectively improving the construction efficiency of the cable laying project and improving the quality of the cable laying project. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] Figure 1 The present invention is a flowchart of the method for resource allocation and optimal scheduling of supervision resources. DETAILED DESCRIPTION

[0006] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0007] like Figure 1 The method for resource allocation and optimal scheduling of supervision resources in this embodiment may specifically include: S101. Obtain cable laying path data, build a three-dimensional visualization model, match the supervisor attributes with the supervision area in the model, and realize dynamic adjustment of supervisor configuration based on the correlation analysis between historical electrical engineering quality issues and supervisor attributes.

[0008] In this embodiment, cable laying path data is acquired, a three-dimensional visualization model is constructed, and supervisor attributes are matched with the supervision areas in the model to generate an initial supervisor allocation plan. Based on the correlation analysis between historical electrical engineering quality issues and supervisor attributes, the allocation plan is optimized, enabling dynamic adjustment of supervisor allocation. Specifically, cable laying path data is acquired based on electrical engineering design drawings, and a three-dimensional visualization model is constructed using AutoCAD software. The cable laying path data includes information on cable laying paths, cable specifications, pipeline layout, and grounding device layout for different areas. A risk level assessment is performed on the three-dimensional visualization model. The weights of the project difficulty coefficient and equipment complexity factors are calculated using the Analytic Hierarchy Process (AHP) to derive a risk score for each supervision area. Based on the risk scores, a K-means clustering algorithm is used to classify historical electrical engineering quality issues, and a correlation model is established between quality issues and supervisor attributes. Once the correlation model is established, an optimization strategy for supervisor allocation is derived through BP neural network training. The input layer of the BP neural network includes supervisor attribute features such as supervisor education, years of experience, and professional certifications, while the output layer represents the probability of quality issues occurring. Obtain supervisor attributes and match them with the supervision areas in the 3D visualization model to generate an initial supervisor allocation plan. Based on this initial supervisor allocation plan, use a sliding time window to calculate the frequency of quality issues in each supervision area and determine whether the frequency exceeds a preset threshold. If the frequency exceeds the preset threshold, the dynamic supervisor adjustment process is initiated.

[0009] Specifically, cable routing data was obtained from electrical engineering design drawings, and information on pipeline layout and grounding device placement was extracted. A 3D visualization model was constructed using AutoCAD software, with cable specifications and layout methods for different areas annotated within the model. A risk assessment was performed on the 3D visualization model. Factors such as project difficulty and equipment complexity were weighted using the analytic hierarchy process (AHP) to calculate the weights of each factor, resulting in a comprehensive risk score for each supervision area. A K-means clustering algorithm was used to classify historical electrical engineering quality issues, establishing a correlation model between quality issues and supervisor attributes. A BP neural network was trained to optimize supervisor allocation strategies. The input layer included supervisor education, years of experience, and professional certifications, while the output layer represented the probability of quality issues occurring. Supervisor attributes were obtained and matched with the supervision areas in the 3D visualization model using knowledge graph technology to generate an initial supervisor allocation plan. Correlations were calculated between the engineering characteristics and supervisor attributes within the supervision area, with minimum support and confidence thresholds set to identify strong association rules. If the match between a supervisor and their assigned area falls below a preset threshold, an adjustment mechanism is triggered, adding experienced supervisors to high-risk areas and assigning new supervisors to low-risk areas for training. Based on real-time project progress and quality feedback data, a sliding time window is used to calculate the frequency of quality issues in each supervision area. When the frequency exceeds a preset threshold, the supervisor dynamic adjustment process is initiated and the configuration optimization algorithm is re-executed. Image recognition technology is used to extract cable routing data from electrical engineering design drawings, with a recognition rate exceeding 95%. Pipeline layout and grounding device layout information are extracted from the drawing text using optical character recognition (OCR) technology, with an accuracy rate of 98%. The extracted data is converted into a 3D visualization model using the AutoCAD software API, with an accuracy error of within ±5mm. Risk assessment of the 3D visualization model considers the project difficulty coefficient (weighted 0.4), equipment complexity (weighted 0.3), and environmental factors (weighted 0.3). The analytic hierarchy process is used to calculate weights, construct a judgment matrix, and solve for the weight vector using the eigenvalue method. The risk score for each supervision area ranges from 0 to 100 points, with scores above 80 being considered high-risk areas and below 50 being considered low-risk areas. The K-means clustering algorithm was used to classify historical quality issues, with the number of clusters K set to 5 and the Euclidean distance used as the similarity metric. The BP neural network model consists of an input layer (10 nodes, corresponding to the attributes of the supervisors), a hidden layer of 20 nodes, and an output layer (5 nodes, corresponding to the types of quality issues). The Sigmoid activation function was used, the learning rate was set to 0.01, and the number of iterations was 1000. The knowledge graph technology was used to construct an association network between the attributes of supervisors and supervision areas. The graph contains node types, including supervisors, supervision areas, and relationship types, including suitable for supervision and qualified.A graph embedding algorithm was used to calculate node similarity, with cosine similarity selected as the matching metric. The minimum support was set to 0.1, and the minimum confidence was set to 0.7. The Apriori algorithm was used to mine association rules. The matching threshold between supervisors and assigned areas was set to 0.8; below this value, an adjustment mechanism was triggered. Real-time project progress and quality feedback data was used as a sliding window of one week to measure the frequency of quality issues. If the frequency exceeded three times per week, the dynamic supervisor adjustment process was initiated, and the configuration optimization algorithm was re-executed. Adjustments were made no more frequently than once a month to ensure project stability.

[0010] Skills training for supervisors is carried out through online learning platforms. The cable laying and grounding device construction process is simulated through virtual reality. Fault scenarios are set up for practical exercises. The supervisors' operations are scored, and the training results are linked to the supervisors' attributes to optimize personnel scheduling strategies.

[0011] Obtain the content of the supervisor skills training course, set up learning modules of multiple difficulty levels based on the course content, and use the course content association matrix to establish the association between learning content. Use 3D modeling software to convert the engineering design drawings into a 3D visualization model. Based on the 3D visualization model, construct the cable laying and grounding device construction scenarios, and convert the construction specifications into executable virtual operation processes. Record the supervisor's operation behavior in the virtual operation process and calculate the similarity between the operation behavior and the preset standard operation process. Based on the operation similarity, obtain the supervisor's training results, associate the training results with the supervisor's attribute data, and construct a supervisor skill assessment indicator system. Use a genetic algorithm to calculate the optimal personnel scheduling plan. If the project progress and quality feedback change, dynamically adjust the personnel allocation based on the changes to achieve continuous optimization of the supervisor's configuration.

[0012] Specifically, an online learning management system is used to create a skills training course for supervisors. Learning modules of multiple difficulty levels are set for the cable laying and grounding device construction process. The correlation relationship between learning contents is constructed through the course content association matrix. The correlation weight is calculated based on the similarity and difficulty level of the learning content to achieve adaptive learning path recommendation. 3D modeling software is used to construct cable laying and grounding device construction scenarios. Engineering design drawings are converted into 3D visualization models. Construction specifications are converted into executable virtual operation processes. A database of various common fault scenarios is imported. Practical exercise tasks are generated through random combinations. The fault scenarios are integrated with the virtual environment to form a complete virtual training environment. Motion capture technology is used to record the operating behavior of supervisors in the virtual environment. Combined with the preset standard operating procedures, an operation sequence comparison algorithm is used to calculate the operation similarity. The supervisors' practical performance is quantitatively scored from three dimensions: time, space, and operation sequence. Training results were linked to supervisor attribute data to construct a supervisor skill assessment index system. A genetic algorithm was used to calculate the optimal staffing plan. Training results, personnel attributes, and project requirements were converted into optimization objectives. A fitness function was designed to evaluate the scheduling plan's performance. Staffing allocation was dynamically adjusted based on project progress and quality feedback, achieving continuous optimization of supervisor staffing. An online learning management system was used to construct a course covering the construction process flow for cable laying and grounding devices, with three difficulty levels: beginner, intermediate, and advanced, each containing 10 learning modules. The course content association matrix used a 50x50 two-dimensional array. Cosine similarity was used to calculate learning content similarity, with difficulty levels weighted between 0.1 and 1.0. Adaptive learning path recommendations used a collaborative filtering algorithm, based on user learning history and content similarity, to recommend courses with a similarity score greater than 0.8. 3D modeling software imported CAD-format engineering drawings with a conversion accuracy of ±1mm, generating a construction scene containing 500 virtual objects. Construction specifications were converted into 100 standardized operating procedures, each containing 10-20 steps. The fault scenario database contains 200 common faults, randomly combined to generate 1,000 practical training tasks. The motion capture system has a sampling frequency of 60 Hz and records the spatial coordinates of 18 key nodes of the supervisor. The operation sequence comparison algorithm uses a dynamic programming method to calculate the edit distance between the standard operating procedure and the actual operation. The error threshold for the time dimension is set to ±5 seconds, and the error threshold for the spatial dimension is set to ±10 cm. The consistency requirement for the operation sequence is greater than 80%. Quantitative scoring uses a percentage system, with weights of 0.3, 0.3, and 0.4 for the time, space, and sequence dimensions, respectively. The supervisor skill assessment index system includes professional knowledge, operational skills, and emergency response, with weights of 0.4, 0.4, and 0.2, respectively. Each aspect has five secondary indicators. The genetic algorithm population size is set to 100, the crossover probability is 0.8, the mutation probability is 0.1, and the number of iterations is 1,000.The fitness function comprehensively considers personnel skill matching, workload balance, and cost factors, with weights of 0.5, 0.3, and 0.2, respectively. Staffing is dynamically adjusted based on weekly project progress reports and quality feedback. Adjustments are made every two weeks, with each adjustment not exceeding 20% ​​of the total headcount to maintain work continuity.

[0013] According to the specific situation of the supervision area, the matching degree between the attributes of the supervisors and the supervision tasks is analyzed. Combined with the importance weights of different professional fields, as well as the work experience and skill level factors of the supervisors, the objective function and constraints are constructed to solve the optimal supervision personnel configuration plan.

[0014] Acquire expert questionnaire data, calculate the importance weight of each professional field, and obtain the professional demand score vector of the supervision area; according to the professional demand score vector of the supervision area, obtain data from the professional qualification, work experience and skill level database of the supervisors, and use the multi-index evaluation method to quantify the comprehensive ability of each supervisor to obtain the supervisor ability score vector; use the weighted cosine similarity algorithm to calculate the similarity between the supervisor ability score vector and the professional demand score vector of the supervision area, and obtain the matching score matrix of the supervisor for each supervision area; after obtaining the matching score matrix, establish an integer programming model with maximizing the overall matching as the objective function, and the decision variable of the integer programming model is the distribution relationship between the supervisor and the supervision area; based on the integer programming model, use the branch and bound algorithm to solve the optimal supervisor configuration plan, and output the personnel allocation result for each supervision area.

[0015] Specifically, based on the geographical location, project scale, and technical difficulty of the supervision area, the Delphi method combined with the entropy weight method was used to calculate the importance weights of each professional field. Opinions were collected through multiple rounds of expert questionnaires, and the entropy weight method was used to calculate the objective weights of electrical, civil engineering, water supply and drainage, and other professional fields, thereby obtaining a professional demand score vector for the supervision area. The professional qualifications, work experience, and skill level data of supervisors were obtained, and a multi-index evaluation method was used to quantify the comprehensive capabilities of each supervisor. Specific indicators included academic qualifications, professional titles, project experience, and skill certifications. By setting a set of evaluation indicators and performing standardization, a weighted score was calculated to obtain a supervisor capability score vector. A matching function between supervisors and supervision tasks was constructed, and a weighted cosine similarity algorithm was used to calculate the similarity between the supervisor capability score vector and the supervision area professional demand score vector. Taking into account the importance weights of different professional fields, a matching score matrix for each supervisor for each supervision area was obtained. An integer programming model was established with the objective function of maximizing the overall matching degree. The decision variables were the allocation of supervisors to supervision areas. Constraints included personnel quantity limits, professional qualification requirements, and workload balance. A branch-and-bound algorithm was used to solve the optimal supervisory staffing solution. Branching strategies and pruning rules were set to improve solution efficiency, and the resulting staffing allocation for each supervision area was output. In supervision project management, the supervision area was first evaluated, taking into account geographic location (e.g., city center, suburbs, remote areas), project size (small, medium, large), and technical difficulty (low, medium, high). A three-round expert questionnaire survey using the Delphi method was conducted, involving 15 industry experts. The questionnaire rated the importance of 10 professional areas, including electrical engineering, civil engineering, and water supply and drainage. The questionnaire used a 5-point Likert scale, with 1-5 representing unimportant to extremely important. After collecting expert opinions, the entropy weighting method was used to calculate the weight of each professional area, resulting in a professional demand score vector for the supervision area. For example, the professional demand score vector for a supervision area is [0.3, 0.25, 0.2, 0.15, 0.1], corresponding to electrical engineering, civil engineering, water supply and drainage, HVAC, and fire protection, respectively. For supervisor assessment, data is collected, including academic qualifications (1 point for bachelor's degree, 2 points for master's degree, 3 points for doctorate), professional title (1 point for assistant, 2 points for intermediate, 3 points for senior), project experience (0.5 points per year, maximum 5 points), and skill certifications (0.5 points per certification, maximum 2 points). A weighted summation method is used to calculate the overall score, with weights of 0.2, 0.3, 0.3, and 0.2, respectively. The scores are normalized to obtain the supervisor competency score vector. For example, the competency score vector for a supervisor is [0.8, 0.7, 0.6, 0.5, 0.4], corresponding to the five aforementioned majors. When constructing the matching evaluation function, a weighted cosine similarity algorithm is used to calculate the similarity between the supervisor competency score vector and the professional demand score vector for the supervision area. The weights use the values ​​in the professional demand score vector to obtain the matching score matrix of each supervisor for each supervision area.Finally, an integer programming model is established, where the decision variable xij represents whether the i-th supervisor is assigned to the j-th supervision area (0 or 1). The objective function is to maximize the overall matching score. Constraints include assigning at least one supervisor to each supervision area, assigning each supervisor to at most one area, and meeting specific professional qualification requirements. A branch-and-bound algorithm is used to solve the problem, with the branching strategy set to minimum lower bound first and the pruning rule set to local upper bound. Through iterative calculations, the optimal supervisor allocation solution is obtained.

[0016] S102. In the hidden parts of the cable laying path, the environmental parameters, equipment status and personnel location data of the cable laying site are collected in real time to visually monitor the construction process. In case of abnormality, an early warning is sent to the appropriate supervisor and the abnormal location is highlighted in the three-dimensional model.

[0017] In this embodiment, real-time data on environmental parameters, equipment status, and personnel locations is collected from hidden locations, including cable trenches and wells, at the cable laying site. This allows for visual monitoring and early warning of the construction process. If an abnormality is detected, an alert is sent to qualified and experienced supervisors. The abnormal location is also highlighted in the 3D visualization model, guiding supervisors to quickly locate the problem. Specifically, the environmental parameters and equipment operating status data collected by the wireless sensor network are obtained. The wireless sensor network is deployed in hidden locations, such as cable trenches and cable wells. Based on the collected data, a data packet containing the sensor ID, timestamp, measurement value and battery power is generated, and the data packet is cleaned, normalized and feature extracted to obtain processed sensor data. The mean, standard deviation and rate of change of the processed sensor data are calculated using a sliding time window, and abnormal states are identified using preset threshold rules and the isolation forest algorithm. Based on the identified abnormal states, the supervisor information is queried from the Neo4j graph database, which stores the professional qualifications, work experience and skill ratings of the supervisors. The Cypher query language is used to screen out matching supervisors, and early warning information containing the abnormality type, location, severity and occurrence time is pushed to the supervisors. The abnormal status data is associated with the spatial position information in the three-dimensional visualization model and then the three-dimensional scene is rendered. The world coordinates of the abnormal position are calculated through the vertex shader, and the red highlight intensity is set in the fragment shader according to the distance from the abnormal point to achieve precise positioning and visualization of the abnormal position.

[0018] Specifically, a wireless sensor network was deployed within the cable trenches and cable shafts. Temperature, humidity, and gas concentration sensors were used to collect environmental parameters. Vibration and current sensors were used to monitor equipment operating status. Bluetooth location beacons were used to obtain construction worker location data. All sensor data was transmitted to a data acquisition gateway via the LoRaWAN protocol. The sensor sampling frequency was set to once per minute, and the data packet format included the sensor ID, timestamp, measurement value, and battery charge level. A real-time data processing platform was built, using Apache Flink to clean, normalize, and extract features from the sensor data. The mean, standard deviation, and rate of change of each metric were calculated using a 5-minute sliding window. Anomalies were identified using pre-set threshold rules and the isolation forest algorithm. Based on detected anomalies, warning messages were generated containing the anomaly type, location, severity, and time of occurrence. A knowledge base for supervisors was established, recording each supervisor's professional qualifications, work experience, and skill rating. Supervisor information was stored in a Neo4j graph database. Supervisors were screened based on their professional compatibility and location using the Cypher query language. The most suitable supervisors were selected and warning messages were pushed via a message queue. The 3D visualization model was updated, associating abnormal status data with the spatial location information within the model. WebGL technology was used to render the 3D scene in the browser. The vertex shader calculated the world coordinates of the abnormal location, and the fragment shader set the red highlight intensity based on the distance from the abnormal point. Combined with real-time location data, the supervisor's movement trajectory was displayed within the model, achieving precise location and visualization of the abnormal location. 100 wireless sensor nodes were deployed within the cable trenches and cable wells, including 30 temperature sensors (accuracy of ±0.5°C), 30 humidity sensors (accuracy of ±3%RH), 20 gas concentration sensors (detection range of 0-100ppm), 10 vibration sensors (frequency range of 0-1000Hz), and 10 current sensors (range of 0-100A). Each sensor collected data every 60 seconds and transmitted it to the gateway via the LoRaWAN protocol, operating in the 470-510MHz frequency band and at a transmission rate of 0.3-50kbps. The data consists of 20 bytes, including a 4-byte sensor ID, an 8-byte timestamp, a 4-byte measurement value, and a 4-byte battery level. The Apache Flink real-time processing platform receives the data stream and calculates the mean, standard deviation, and rate of change using a 5-minute sliding window with a 1-minute step size. Anomaly detection uses the Isolation Forest algorithm, with 100 decision trees, a sample size of 256, and an anomaly threshold of -0.5. Upon detecting an anomaly, an alert is generated in JSON format, including the anomaly type (e.g., excessive temperature, abnormal humidity), the location coordinates (x, y, z), a severity level (1-5), and the time of occurrence.The supervisor knowledge base is stored in a Neo4j graph database and contains 50 supervisor nodes, each with 10 attributes, such as professional qualifications, years of experience, and skill rating. Using the Cypher query language, the most suitable supervisors are selected based on professional compatibility (weighted 0.6) and location (weighted 0.4), with query times under 100ms. Selected supervisor information is pushed to mobile devices via a RabbitMQ message queue with a throughput of 10,000 messages per second. The 3D visualization model uses WebGL technology, calculating world coordinates through a vertex shader. A red highlight effect is set in the fragment shader, with RGB values ​​ranging from 255,0,0 to 255,200,200, varying with distance from the anomaly. The model updates at a frequency of 10 frames per second and supports real-time viewing by 30 concurrent users. Supervisor locations are updated every 5 seconds and are marked in the model as blue dots, with their movement trajectories for the last 10 minutes displayed.

[0019] S103. Analyze the images and videos of cable fireproof sealing and cable insulation resistance testing at the construction site, extract key information of the sealing materials and resistance testing instruments, compare the key information with the cable properties at the corresponding positions in the three-dimensional visualization model, and determine whether the construction quality meets the specifications and standards.

[0020] In this embodiment, images and videos of key processes of cable fireproof sealing and cable insulation resistance testing at the construction site are analyzed to extract key information such as the type, specification, quantity, construction process of the sealing material and the model, range, accuracy, and calibration status of the resistance testing instrument. The information is then compared with the cable properties at the corresponding positions in the three-dimensional visualization model to determine whether the construction quality meets the specifications and standards. Specifically, video data of the cable fire-proof sealing and insulation resistance testing process is obtained; key frames are extracted from the video data to obtain a representative image with one frame per second; the representative image is processed using a target detection algorithm to identify the sealing materials and test instruments in the representative image to obtain an annotated image; the annotated image is processed using an optical character recognition method to extract text information in the annotated image, which includes the type, specification, and quantity of the sealing material, as well as the model, range, accuracy, and calibration date of the resistance test instrument; based on the spatiotemporal information in the annotated image, the text information is associated with the cable position in the three-dimensional visualization model to obtain associated data; a rule engine based on the Rete algorithm is constructed, and the associated data is input into a rule base, which contains a set of rules for material specification matching, process parameter qualification, and instrument accuracy requirements; based on the output results of the rule engine, the construction quality is judged and a quality assessment report is generated.

[0021] Specifically, a high-definition camera was used to capture the cable fire-proof sealing and insulation resistance testing process in real time. Image preprocessing algorithms were used to denoise, enhance, and correct the captured images. Keyframes were extracted from the video data, with one representative frame selected every second. The YOLOv5 object detection algorithm was used to identify the sealing material and test instrument, and the test results were annotated on the original image to form an annotated image. Optical character recognition technology was used to extract textual information from the annotated image, including the type, specification, and quantity of the sealing material, as well as the model, range, accuracy, and calibration date of the resistance test instrument. The extracted text was semantically analyzed using the BERT-CRF named entity recognition model to identify key information and structured storage. The construction process was analyzed to identify the process sequence and key parameters, and process parameters such as sealing depth, compaction degree, and test voltage were extracted. Based on the spatiotemporal information in the image, the extracted construction process data was correlated with the cable locations in the 3D visualization model. The R-tree spatial indexing algorithm was used to quickly locate the construction area, mapping image coordinates to 3D model coordinates. Cable attribute information at the corresponding location, including cable model, cross-sectional area, and length, was retrieved from the model database. A rule engine based on the Rete algorithm was built. Extracted construction data and cable attribute information were input into a rule base containing rules such as material specification matching, process parameter compliance, and instrument accuracy requirements. Through rule matching and reasoning, the system compared the parameters of the plugging materials and test instruments to determine if they met regulatory requirements, determined the construction quality, and generated a quality assessment report containing the judgment results and detailed comparison data. Ten 4K HD cameras were deployed at the cable construction site, each capturing a 1080p resolution video stream at 30 fps. During video preprocessing, Gaussian filtering was used for noise reduction, histogram equalization was employed to enhance image contrast, and affine transformation was used to correct image tilt. A keyframe extraction algorithm selected one frame per second and used the inter-frame differencing method to calculate the difference between adjacent frames, with a difference threshold of 0.1. The YOLOv5 object detection model was trained on 5,000 annotated images, covering 20 types of plugging materials and 10 types of test instruments, achieving a detection accuracy of 95%. Optical character recognition, using the Tesseract engine, achieved a character recognition accuracy of 98%. The BERT-CRF named entity recognition model was trained using 10,000 annotated corpora, encompassing 15 entity types, with an F1 score of 0.92. Construction process analysis uses a rule-based sequence labeling algorithm to identify eight standard processes and extract 12 key process parameters. The R-tree spatial indexing algorithm constructs a five-layer tree structure, with each node containing up to 50 child nodes and a query time complexity of O(logn). Three-dimensional model coordinate mapping uses an affine transformation matrix, with mapping errors controlled within ±5cm. The rule engine, implemented based on the Rete algorithm, contains 100 rules covering five major categories of construction specifications. The rule matching speed reaches 10,000 times per second, and a single quality assessment takes no more than 100ms.The final quality assessment report contains 20 evaluation indicators. The qualification standard and actual value of each indicator are compared to provide quantitative scores and qualitative suggestions.

[0022] S104. If it is found that the fireproof sealing material is not in compliance or the insulation resistance test is abnormal, the supervisor will be assigned to handle the problem according to the severity of the problem, and the review results will be fed back to the 3D visualization model and the cable status will be updated.

[0023] In this embodiment, if it is found that the fireproof sealing material does not conform to the design or the insulation resistance test result is abnormal, a supervisor with strong professional ability is matched to deal with it according to the severity of the problem, the on-site review situation is fed back to the three-dimensional visualization model, and the cable status is updated. Specifically, a convolutional neural network is used to perform image recognition on the fireproof sealing material, and the material specification information is obtained from the design database; the similarity score is calculated based on the image recognition result and the material specification information; if the similarity score is lower than the preset threshold, the fireproof sealing material is determined to be inconsistent. Based on the test data collected by the insulation resistance tester, the ARIMA model is used to analyze the test data; the deviation between the test data and the standard value is calculated by the ARIMA model; the degree of deviation is judged based on the deviation to obtain the severity index. A supervisor capability assessment model is constructed, and the input features of the capability assessment model include the supervisor's professional qualifications, work experience and historical disposal effects; the supervisor's professional capability score is output through the supervisor capability assessment model. Based on the severity index of the abnormal situation information, match the supervisor with the corresponding professional ability score; push the abnormal situation information to the matched supervisor; use the real-time communication protocol to synchronize the on-site review data to the 3D visualization model and update the cable status information.

[0024] Specifically, the ResNet50 convolutional neural network is used to perform image recognition on fireproof sealing materials, obtain material specification information from the design database, and compare it with the material specifications in the database through feature extraction to calculate the similarity score. If the similarity is lower than the preset threshold of 0.85, it is determined that the material does not conform. Based on the data collected by the insulation resistance tester, the ARIMA model is used to analyze the test results, calculate the deviation between the test value and the standard value, set a 5-level warning threshold, and quantify the degree of deviation into a severity index of 1-5. Construct a supervisor's ability assessment model, use the supervisor's professional qualifications, work experience and historical disposal effects as input features, implement the gradient boosting decision tree algorithm through the XGBoost library, train the regression model to output the supervisor's professional ability score, and match supervisors with corresponding ability levels according to the problem severity index. A mobile application was developed to notify selected supervisors of abnormalities. Using GPS positioning and ARKit for augmented reality guidance, supervisors were directed to the problem location. On-site verification data was collected and uploaded to a central server. After data verification and conflict resolution, the results were synchronized to the 3D visualization model using the WebSocket real-time communication protocol, updating cable status information, including material type, construction quality, and insulation performance. Five high-definition cameras were deployed at the cable construction site to capture real-time images of fire-resistant sealing materials. A ResNet50 network, trained on 10,000 annotated images, achieved 98% recognition accuracy. The design database includes 500 standard material specifications, and feature vector matching was calculated using cosine similarity, with a threshold of 0.85. The insulation resistance tester collected data every five minutes. The ARIMA model used the previous 24 hours of data for prediction, with five warning thresholds set at ±5%, ±10%, ±15%, ±20%, and ±25% of the standard value. The supervisor competency assessment model is based on the XGBoost algorithm, using 100 decision trees with a learning rate of 0.1 and a maximum depth of 6. Parameters are optimized using 5-fold cross-validation. Input features include professional qualifications (1-5 points), work experience (years), and historical handling results (success rate), and the output is a competency score on a scale of 0-100. The mobile application is developed using the Flutter framework and supports both Android and iOS. GPS positioning accuracy is controlled to ±3 meters, and ARKit achieves centimeter-level spatial positioning accuracy. On-site review data includes 20 key parameters and is verified for data integrity using the SHA256 algorithm. WebSocket communication latency is kept within 100ms. The 3D visualization model is built using the Unity engine, supports 1,000 concurrent users, and updates 10 times per second. The average response time from anomaly detection to supervisor arrival is kept within 15 minutes, enabling real-time monitoring and rapid resolution of cable construction quality.

[0025] Obtain the schedule and milestone nodes of the cable laying project, use the critical path method and resource balancing method to optimize the allocation time and resources of the supervisors' work tasks, arrange the supervisors' on-site time and rotation plan accordingly, and make dynamic adjustments based on actual progress.

[0026] Obtain the schedule and milestone node data of the cable laying project. The schedule includes task names and estimated durations. Generate a Gantt chart to visualize the project timeline based on the schedule and milestone node data. The Gantt chart displays the start and end times of tasks. Use the critical path algorithm to calculate the critical path of the project, obtain the task sequence that has the greatest impact on the total construction period, and determine the earliest start time and latest completion time of each task. Use a genetic algorithm to optimize the work task allocation of supervisors. The genetic algorithm uses binary coding to represent the task allocation scheme and uses single-point crossover and bit flip mutation operations to generate new schemes. Collect actual project progress data through a real-time progress tracking system. The real-time progress tracking system records the task completion percentage and actual cost expenditure. Use earned value analysis to calculate the deviation between the plan and the actual. If the progress deviation exceeds the preset threshold, the rescheduling mechanism is triggered to dynamically adjust the work arrangements of the supervisors.

[0027] Specifically, the cable laying project schedule and milestone data were imported from the project management system. A Gantt chart visualization tool was used to display the project timeline. The critical path algorithm was used to calculate the project's critical path, identify the task sequence with the greatest impact on the total duration, and determine the earliest start time and latest completion time for each task. The PERT technique was used to handle complex dependencies, identify parallel tasks, and construct a complete task network diagram. A supervisor resource pool was established, recording each supervisor's professional skills, work experience, and current workload. A resource histogram balancing method was used to perform a preliminary allocation of supervisory tasks, calculating the resource requirements and available resources for each task. Resource utilization was smoothed by adjusting the start times of tasks on non-critical paths. A genetic algorithm was used to optimize the supervisor's work assignments. Binary encoding was used to represent the task assignment schemes, and new schemes were generated using single-point crossover and bit-flip mutation operations. Task completion time and resource utilization were used as fitness functions. The optimization process was iterative until the optimal solution was found, resulting in the specific work schedule for each supervisor. Based on the optimized task allocation plan, on-site schedules and rest rotations for supervisors were developed, taking into account statutory working hours and consecutive working day limits. Actual project progress data was collected through a real-time progress tracking system, and earned value analysis was used to calculate deviations from the plan. When the progress deviation exceeded 10% or the cost deviation exceeded 5%, a rescheduling mechanism was triggered to dynamically adjust the supervisors' work arrangements. If actual progress fell significantly behind schedule, an emergency scheduling plan was activated, temporarily increasing manpower or adjusting work priorities. For the cable laying project, the project management system imported a schedule containing 500 task nodes and 20 key milestones. A Gantt chart tool displayed the 180-day duration in daily units. The critical path algorithm identified 78 critical tasks, representing 15.6% of the total number of tasks. The PERT technique addressed 50 complex dependencies and identified 30 task groups that could be executed in parallel. The resource pool consisted of 50 supervisors, each possessing an average of three specialized skills and with work experience ranging from 1 to 20 years. The resource histogram balancing method controlled daily resource demand fluctuations within ±10%, increasing average resource utilization to 85%. A genetic algorithm used 100-bit binary code to represent task allocation plans, with a population size of 200, a crossover probability of 0.8, and a mutation probability of 0.05. Convergence was achieved after 500 iterations. The fitness function weighted work periods by 0.6 and resource utilization by 0.4. The optimized plan shortened the project's total duration by 5 days and increased resource utilization by 7%. The work shift schedule adhered to the principle of no more than 44 hours per week and no more than six consecutive days. A real-time progress tracking system updated progress data every four hours, and earned value analysis was used to calculate the schedule performance index (SPI) and cost performance index (CPI). Rescheduling was triggered when the SPI fell below 0.9 or the CPI fell below 0.95, with an average response time of two hours.The emergency dispatch plan presets five resource allocation modes, which can mobilize an additional 20% of human resources within 24 hours, realizing the full intelligent scheduling and optimization of the cable laying project with a team of 50 people and a construction period of 180 days.

[0028] S105. Receive update instructions for cable installation construction, perform semantic understanding on the update content, and evaluate the feasibility of the update based on the skill level of the supervisor.

[0029] In this embodiment, update instructions for cable installation are transmitted in real time through wireless communication, semantic understanding of the update content is performed, and specified information on the material and specifications of the grounding device is extracted. The feasibility of the update is evaluated based on the skill level and communication ability of the supervisor. Specifically, update instructions are received, AES-256 encryption is performed on the update instructions, and the SHA-256 algorithm is used to generate an integrity check value for the update instructions; natural language processing is performed on the encrypted update instructions, including word segmentation, part-of-speech tagging, and named entity recognition, to extract specified information from the design change instructions; a supervisory personnel capability assessment model is constructed based on the specified information, and the supervisory personnel capability assessment model adopts a multi-layer perceptron algorithm, with the input layer containing a preset number of nodes, the two hidden layers each containing a preset number of nodes, and the output layer containing a preset number of nodes, to obtain the supervisory personnel's skill level score and communication ability score; based on the specified information of the update instructions, the supervisory personnel's skill level score, and the supervisory personnel's communication ability score, a decision tree model is constructed, the decision nodes of the decision tree model include change complexity, material availability, and supervisory personnel capability matching, the optimal partitioning features are selected through the information gain ratio, and the pessimistic pruning method is used to prevent overfitting to obtain the change feasibility assessment result.

[0030] Specifically, a low-power wide-area network is deployed, using the LoRaWAN protocol to achieve real-time transmission of design change instructions. Data is uploaded to a cloud server via a gateway, and message queue technology is used to ensure reliable delivery and sequential processing of instructions. Received instructions are encrypted with AES-256 and integrity checked with SHA-256. A local cache mechanism is set up to automatically save instructions when communication is interrupted and retransmit them after the network is restored. Natural language processing technology is used to pre-process the content of the change instructions, including word segmentation, part-of-speech tagging, and named entity recognition. Domain-specific dictionaries and rules are used to identify keywords and phrases related to grounding devices. Dependency parsing and semantic role labeling are used to extract specific information about materials and specifications. Relationship extraction technology is used to identify associations between entities. The extracted specified information is matched with predefined supervisory skill requirements to generate a preliminary assessment of the required supervisory capabilities. A supervisor competency assessment model was constructed, using factors such as supervisors' professional qualifications, work experience, and historical performance as input features. A multi-layer perceptron algorithm was used to train the regression model. The network structure consisted of a 10-node input layer, two hidden layers with 20 nodes each, and an output layer with 2 nodes. Using the ReLU activation function and the Adam optimizer, the model output quantitative scores of skill level and communication ability, thereby establishing a supervisor competency profile database. Based on the extracted change information and supervisor competency scores, a C4.5 decision tree model was constructed to assess change feasibility. Change complexity, material availability, and supervisor competency match were used as decision nodes. The optimal partitioning features were selected using the information gain ratio, and pessimistic pruning was used to prevent overfitting. The model generated feasibility assessment results and corresponding decision-making basis. A feasibility threshold of 0.75 was set. When the assessment result exceeded this threshold, a change recommendation report was automatically generated, including specific implementation steps and a list of required resources. During the cable laying project, 50 LoRaWAN nodes were deployed, covering a 10-square-kilometer construction area. The data transmission rate was 5.5 kbps, and communication latency was kept within 100 ms. The cloud server uses Kafka message queues and can process 1,000 instructions per second. The AES-256 encryption key length is 256 bits, and the SHA-256 hash value length is 32 bytes. The local cache capacity is 1GB and can store instructions within 24 hours. The natural language processing module uses the Jieba word segmenter with an accuracy of 98% and a named entity recognition F1 score of 0.92. The domain dictionary contains 5,000 professional terms. Dependency parsing uses the Stanford parser with an accuracy of 85%. Semantic role labeling uses the BERT model with an F1 score of 0.88. The relationship extraction accuracy reaches 80%. The supervision capability assessment model is based on the PyTorch framework. The input layer contains 10 features, such as education, years of work experience, project experience, etc. The two hidden layers have 20 neurons each, and the two nodes in the output layer represent skill level and communication ability respectively.The ReLU activation function was used, with a learning rate of 0.001 and a batch size of 64. After 5000 training epochs, convergence was achieved, with the average error reduced to 0.05. The maximum depth of the C4.5 decision tree was set to 8, and the minimum number of sample splits was 10. Cross-validation was used to select the optimal parameters. The information gain ratio threshold was set to 0.1, and a pessimistic error rate estimate was used during pruning. The feasibility assessment achieved an accuracy of 85%. For changes with an assessment score exceeding 0.75, a recommendation report containing 10 implementation steps and 20 resource requirements was automatically generated.

[0031] S106. Digitally store the cable installation construction data, record the supervisor's review opinions and operation logs, and include the supervision records of cable installation construction changes in the records. At the same time, continuously evaluate and dynamically adjust the configuration of supervisors to form a closed-loop optimization.

[0032] In this embodiment, cable installation construction data is digitally archived, recording each supervisor's review opinions and operation logs to achieve accountability. Supervision records of cable installation construction changes are included in the digital archive. The supervisor configuration plan is continuously evaluated to ensure compliance with pre-set requirements, and the supervisor configuration is dynamically adjusted based on the evaluation results, forming a closed-loop optimization system. Specifically, cable installation construction data is obtained and stored in a distributed file storage system. Based on the construction data in the distributed file storage system, a hash value and timestamp are calculated for each document. The hash value and timestamp are recorded to create an immutable digital archive chain. Supervisor review operations and system interactions are received; click paths, dwell time, and operation frequency are recorded based on these operations. Change application, review process, and execution status information are obtained and included in the digital archive. The content of each change is recorded. Association rules between change items and supervisors are established using the FP-Growth algorithm. Historical performance data for supervisors is obtained and massively parallelized. Based on the parallel processing results, the ARIMA model is used to assess whether the supervisor configuration plan meets pre-set requirements. If not, the staff scheduling strategy is optimized through a deep Q-learning network, where the state space includes the current staff distribution and workload, and the action space is the staff deployment options.

[0033] Specifically, a distributed file storage system is used to digitize cable laying construction data. Image recognition, OCR, and audio and video transcription technologies are used to extract designated key information from drawings, texts, and multimedia files. Blockchain technology is used to record the hash value and timestamp of each document to build a tamper-proof digital evidence chain. The supervision operation record system is used, and Google Analytics for Firebase is used to capture the supervisor's review operations and system interaction behaviors, record click paths, dwell time, and operation frequency. Natural language processing technology is used to perform semantic analysis on the review opinions to extract key viewpoints and decision-making basis. Through the construction change management platform, change applications, review processes, and execution status are included in the scope of digital evidence. Version control technology is used to record the details of each change. The FP-Growth algorithm is used to establish association rules between change items and supervisors, and a change impact assessment report is generated. Apache Spark was used to perform large-scale parallel processing of historical supervisor performance data. The ARIMA model was employed to evaluate the rationality of supervisor configuration plans. A Deep Q-Learning Network (DQN) was used to optimize the staff scheduling strategy. The state space was defined to include the current staff distribution and workload, the action space to represent staff deployment options, and the reward function was designed based on project progress and quality indicators. Monthly evaluations were conducted to collect feedback from the construction and owner. Scheduling recommendations, combined with the DQN output, were reviewed and adjusted by the project management committee, forming a closed-loop optimization process. For the cable laying project, a distributed storage system with a capacity of 100TB was used, supporting 1,000 concurrent reads and writes per second. Image recognition used the ResNet50 network, achieving an accuracy of 95%. OCR used the Tesseract engine, achieving a recognition rate of 98%. Audio and video transcription employed the DeepSpeech model, maintaining a word error rate of less than 6%. The blockchain, utilizing the Hyperledger Fabric framework, was able to process 500 transactions per second. The supervision operation record system is based on the Firebase platform and collects approximately 100,000 pieces of user behavior data daily, including 5,000 review operations. The BERT model is used for natural language processing, achieving an F1 score of 0.85 for extracting key insights. The construction change management platform uses Git version control and supports 100 concurrent branches. The FP-Growth algorithm generates 300 association rules with a minimum support of 0.05 and a confidence level of 0.7. The Apache Spark cluster consists of 20 nodes and can process 1TB / hour. The ARIMA model has parameters of (2, 1, 2) and a prediction accuracy of 85%. The DQN network consists of three fully connected layers, each with 128 neurons. The ε-greedy strategy has an initial ε value of 0.9 and decays by 0.995 per round. The state space dimension is 50, representing the current five resource allocations in 10 regions; the action space dimension is 100, representing different personnel deployment plans.The reward function comprehensively considers the progress achievement rate and quality pass rate, with weights of 0.6 and 0.4, respectively. The model is updated weekly, and a monthly scheduling optimization report containing 20 specific recommendations is generated. The project management committee reviews the recommendations through a visual interface, with an average adoption rate of 80%.

[0034] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for resource allocation and optimal scheduling of supervision resources, characterized in that: Methods include: Acquire cable laying path data, build a 3D visualization model, match supervisor attributes with the supervision area in the model, and dynamically adjust supervisor configuration based on correlation analysis between historical electrical engineering quality issues and supervisor attributes. In hidden areas of the cable laying path, real-time data on environmental parameters, equipment status, and personnel location are collected at the cable laying site to visually monitor the construction process. In the event of anomalies, early warnings are sent to appropriate supervisors, and the abnormal locations are highlighted in the 3D model. Analyze images and videos of cable fireproofing and insulation resistance testing at the construction site, extract key information about the sealing materials and resistance test instruments, and compare this key information with the cable attributes at the corresponding locations in the 3D visualization model to determine whether the construction quality meets regulatory standards. If the fireproofing sealing materials are found to be inconsistent or the insulation resistance test is abnormal, the appropriate supervisor will be assigned to handle the issue based on its severity. The review results will be fed back to the 3D visualization model and the cable status will be updated. Receive update instructions from cable installation construction, conduct semantic understanding of the update content, and evaluate the feasibility of the update based on the skill level of the supervisors; digitally store the cable installation construction data, record the supervisors' review opinions and operation logs, and include the supervision records of cable installation construction changes in the storage. At the same time, continuously evaluate and dynamically adjust the configuration of supervisors to form a closed-loop optimization.

2. The method according to claim 1, characterized in that The method of acquiring cable laying path data, constructing a three-dimensional visualization model, matching supervisor attributes with the supervision area in the model, and dynamically adjusting supervisor configuration based on correlation analysis between historical electrical engineering quality issues and supervisor attributes includes: Obtain cable laying path data based on electrical engineering design drawings and use AutoCAD software to build a 3D visualization model; Conduct risk level assessment on the 3D visualization model, calculate the weights of engineering difficulty coefficient and equipment complexity factors through the analytic hierarchy process, and derive the risk score for each supervision area; Based on the risk score, the K-means clustering algorithm is used to classify historical electrical engineering quality issues and establish a correlation model between quality issues and supervisor attributes; After the association model is established, the optimization strategy for supervisor configuration is obtained through BP neural network training. The input layer of the BP neural network includes the characteristics of supervisor attributes, and the output layer is the probability of quality problems. Match the supervisor attributes with the supervision area in the 3D visualization model to generate an initial supervisor configuration plan; Based on the initial supervision staffing plan, count the frequency of quality problems in each supervision area; If the frequency of quality problems exceeds the preset threshold, the dynamic adjustment process of supervisors will be initiated.

3. The method according to claim 2, characterized in that The method further comprises: Conduct supervisor skills training through online learning platforms, conduct practical exercises through virtual reality simulations, score supervisors' operations, and correlate training results with supervisor attributes to optimize personnel scheduling strategies; According to the specific situation of the supervision area, the matching degree between the attributes of the supervisors and the supervision tasks is analyzed. Combined with the importance weights of different professional fields, as well as the work experience and skill level factors of the supervisors, the objective function and constraints are constructed to solve the optimal supervision personnel configuration plan.

4. The method according to claim 3, characterized in that The aforementioned method involves conducting supervisor skills training through an online learning platform, conducting practical exercises through virtual reality simulation, scoring supervisors' operations, associating training results with personnel attributes, and optimizing personnel scheduling strategies, including: Obtain the content of the supervisory personnel skill training course, set up learning modules of multiple difficulty levels based on the course content, and use the course content association matrix to construct the association relationship between the learning content; use 3D modeling software to convert the engineering design drawings into 3D visualization models, build cable laying and grounding device construction scenes based on the 3D visualization models, and convert the construction specifications into executable virtual operation processes; record the operation behavior of the supervisors in the virtual operation process, and calculate the operation similarity between the operation behavior and the preset standard operation process; obtain the training results of the supervisors based on the operation similarity, associate the training results with the supervisors' attribute data, and construct a supervisory personnel skill evaluation index system; use genetic algorithms to calculate the optimal personnel scheduling plan. If the project progress and quality feedback change, the personnel allocation will be dynamically adjusted according to the changes to achieve continuous optimization of the supervision personnel configuration.

5. The method according to claim 3, characterized in that Based on the specific conditions of the supervision area, the matching degree between the professional qualifications of the supervisors and the supervision tasks is analyzed. The objective function and constraints are constructed based on the importance weights of different professional fields, as well as the work experience and skill level of the supervisors, to solve the optimal supervision personnel configuration plan, including: Acquire expert questionnaire data, calculate the importance weight of each professional field, and obtain the professional demand score vector of the supervision area; according to the professional demand score vector of the supervision area, obtain data from the professional qualification, work experience and skill level database of the supervisors, and use the multi-index evaluation method to quantify the comprehensive ability of each supervisor to obtain the supervisor ability score vector; use the weighted cosine similarity algorithm to calculate the similarity between the supervisor ability score vector and the professional demand score vector of the supervision area, and obtain the matching score matrix of the supervisor for each supervision area; after obtaining the matching score matrix, establish an integer programming model with maximizing the overall matching as the objective function, and the decision variable of the integer programming model is the distribution relationship between the supervisor and the supervision area; based on the integer programming model, use the branch and bound algorithm to solve the optimal supervisor configuration plan, and output the personnel allocation result for each supervision area.

6. The method according to claim 1, characterized in that In the hidden parts of the cable laying path, the environmental parameters, equipment status and personnel location data of the cable laying site are collected in real time, the construction process is visually monitored, and an alarm is sent to the appropriate supervisor in case of an abnormality, and the abnormal location is highlighted in the 3D model, including: Acquiring environmental parameters and equipment operating status data collected by a wireless sensor network, where the wireless sensor network is deployed in hidden locations, including cable trenches and cable wells; Generate sensor data including sensor ID, timestamp, measurement value and battery level based on the collected data; The sliding time window is used to calculate the mean, standard deviation and rate of change of sensor data, and abnormal conditions are identified through preset threshold rules and the isolation forest algorithm; According to the identified abnormal state, the supervisor information is queried from the preset map database; Screen out matching supervisors and send them warning information including the abnormality type, location, severity, and occurrence time; The abnormal state data is associated with the spatial position information in the three-dimensional visualization model and then the three-dimensional scene is rendered; Calculate the world coordinates of the anomaly location and set the red highlight intensity based on the distance from the anomaly point.

7. The method according to claim 1, characterized in that The images and videos of cable fireproofing and cable insulation resistance testing at the construction site are analyzed to extract key information about the sealing materials and resistance test instruments. The key information is compared with the cable attributes at the corresponding positions in the 3D visualization model to determine whether the construction quality meets the specifications and standards, including: Obtain video data of cable fire sealing and insulation resistance testing processes; Extract key frames from video data to obtain a representative image at one frame per second; The representative images are processed using the target detection algorithm to identify the blocking materials and test instruments in the representative images and obtain the labeled images; Extract text information from the annotated image, including the type, specification, and quantity of the plugging material, as well as the model, range, accuracy, and calibration date of the resistance test instrument; According to the spatiotemporal information in the annotated image, the text information is associated with the cable position in the 3D visualization model to obtain associated data; Build a rules engine to input the associated data into a rule base containing rule sets for material specification matching, process parameter eligibility, and instrument accuracy requirements; Based on the output results of the rule engine, the construction quality is judged and a quality assessment report is generated.

8. The method according to claim 1, characterized in that If the fireproof sealing material is found to be inconsistent or the insulation resistance test is abnormal, the supervisor will be assigned to handle the problem according to the severity of the problem. The review results will be fed back to the 3D visualization model and the cable status will be updated, including: Use convolutional neural networks to perform image recognition on fireproof sealing materials and obtain material specification information from the design database; Calculate a similarity score based on the image recognition results and the material specification information. If the similarity score is lower than a preset threshold, the fireproof blocking material is determined to be inconsistent; Calculate the deviation between the test data and the standard value based on the test data collected by the insulation resistance tester; Based on the deviation, the degree of deviation is judged to obtain a severity index; Construct a supervisory personnel capability assessment model. The input features of the capability assessment model include the supervisor's professional qualifications, work experience, and historical disposal results; Output the professional ability score of supervisors through the supervisor ability assessment model; Match supervisors with corresponding professional competence scores based on the severity index of abnormal situation information; Push abnormal situation information to the matching supervisor; Utilize real-time communication protocols to synchronize field review data to the 3D visualization model and update cable status information.

9. The method according to claim 1, characterized in that The receiving of cable installation construction update instructions, semantic understanding of the update content, and assessment of the feasibility of the update based on the skill level of the supervisor include: Receive an update instruction, encrypt the update instruction, and generate an integrity check value for the update instruction; Perform natural language processing on the encrypted update instructions, including word segmentation, part-of-speech tagging, and named entity recognition, to extract specified information from the design change instructions; A supervisory personnel capability assessment model is constructed based on the specified information. The supervisory personnel capability assessment model adopts a multi-layer perceptron algorithm. The input layer contains preset nodes, the two hidden layers each contain preset nodes, and the output layer contains preset nodes. The skill level score and communication ability score of the supervisor are obtained. Based on the specified information of the update instructions and the skill level score and communication ability score of the supervisors, a decision tree model is constructed. The decision nodes of the decision tree model include the complexity of the change, the availability of materials, and the matching degree of the supervisor's ability. The optimal partitioning features are selected through the information gain ratio, and the pessimistic pruning method is used to prevent overfitting to obtain the feasibility assessment results of the change.

10. The method according to claim 1, characterized in that The aforementioned digital evidence storage of cable installation construction data, recording supervisors’ review opinions and operation logs, and incorporating supervisory records of cable installation construction changes into the evidence storage. At the same time, continuous evaluation and dynamic adjustment of supervisor personnel allocation will be conducted to form a closed-loop optimization, including: Obtain cable laying construction data and store the construction data in a distributed file storage system; Calculate the hash value and timestamp of each document based on the construction data in the distributed file storage system; Record hash values ​​and timestamps to obtain a tamper-proof digital evidence chain; Receive supervisors' audit operations and system interaction behaviors, and record click paths, dwell time, and operation frequency; Obtain information on change applications, review processes, and implementation status, and include this information in digital evidence storage; Record the content of each change; Establish rules for associating changes with supervisors; Obtain historical performance data of supervisors and perform large-scale parallel processing of historical performance data; Based on the parallel processing results, evaluate whether the supervision configuration plan meets the preset requirements; If not, the staff scheduling strategy is optimized through a deep Q-learning network, where the state space includes the current staff distribution and workload, and the action space is the staff deployment options.

Images