Intelligent integrated management method and platform system for infrastructure projects

Abstract

The application discloses a kind of wisdom integrated management method and platform system for infrastructure project, it is related to infrastructure management technical field, its method includes: obtaining project planning data;Build double-layer association fusion graph model, extract work order plan information and personnel static information from project planning data Generation plan reference vector;Collect human-computer interaction data;According to the first cognitive safety index calculated from human-computer interaction data;The work order is obtained by combining the first cognitive safety index and disassembling to get work order set;The work order management index vector of operation personnel is obtained by the application program of operation personnel terminal to complete work order set;According to work order management index vector and plan reference vector, use enhanced management entropy vector fusion module and double-layer association fusion graph model to obtain global risk influence probability distribution;According to global risk influence probability distribution, generate project-level risk early warning information and planning strategy.The application can effectively improve the management efficiency of infrastructure project.

Description

Technical Field

[0001] This application relates to the field of infrastructure management technology, and in particular to a smart integrated management method and platform system for infrastructure projects. Background Technology

[0002] Infrastructure projects, as vital material carriers supporting modern society and economic development, encompass multiple key areas such as transportation, energy, water conservancy, communications, and urban construction. The quality and efficiency of these projects directly impact the sustained growth of the regional economy, the improvement of social productivity, and the enhancement of residents' quality of life. Therefore, ensuring the effective management of infrastructure projects is of paramount importance.

[0003] In traditional infrastructure project management models, operators often need to process a large amount of information simultaneously when performing tasks, including construction plans, equipment status, work instructions, and safety regulations. Due to the large volume of information, the complexity of tasks, and the variable environment, operators are prone to cognitive biases or insufficient attention. For example, operators may misunderstand and execute steps inconsistent with the expected operational goals, or they may overlook critical operational steps due to distraction, both of which can affect work quality and safety levels. Traditional management models mainly rely on on-site supervision or experience-based judgment to identify and correct such biases. However, there is a lack of standardized methods for identifying and quantifying cognitive biases, making it difficult to develop quantifiable risk assessment indicators. Therefore, existing technologies struggle to dynamically monitor and quantify the cognitive state of operators in infrastructure projects, and cannot effectively predict potential risks or implement targeted management measures, thus limiting the safety of project management. Summary of the Invention

[0004] This application provides a smart integrated management method and platform system for infrastructure projects, which improves the management efficiency of infrastructure projects.

[0005] To achieve the above objectives, the embodiments of this application adopt the following technical solutions:

[0006] Firstly, a smart integrated management method for infrastructure projects is provided, including:

[0007] Obtain project planning data for infrastructure projects;

[0008] A two-layer relational fusion graph model is constructed based on project planning data, and work ticket plan information and personnel static information are extracted from the project planning data to generate a plan baseline vector.

[0009] Human-computer interaction data is collected through an application deployed on the operator's terminal;

[0010] The first cognitive safety index is calculated based on human-computer interaction data;

[0011] When the first cognitive safety index exceeds the preset dynamic threshold, the work ticket is decomposed based on the first cognitive safety index to obtain a set of work tickets.

[0012] A set of work tickets is pushed to the operator. The operator obtains the work ticket management indicator vector of the operator's completion of the work ticket set through the application on the operator's terminal. The work ticket management indicator vector includes real-time execution data and a second cognitive safety index.

[0013] Based on the work ticket management indicator vector and the plan baseline vector, the global risk impact probability distribution is obtained using the enhanced management entropy vector fusion module and the two-layer association fusion graph model;

[0014] Project-level risk warning information and planning strategies are generated based on the probability distribution of global risk impact.

[0015] In another possible implementation of the first aspect, the construction of a two-layer relational fusion graph model based on project planning data, and the extraction of work order plan information and personnel static information from the project planning data to generate a plan baseline vector, includes:

[0016] Extract project work breakdown structure data, human resource organizational structure and human resource data from project planning data;

[0017] Analyze the project work breakdown structure data and establish a task dependency graph in the graph database. Each task node in the task dependency graph contains planned man hours and budgeted cost.

[0018] Analyze human resource organizational structure and human resource data, and establish a personnel organizational relationship diagram in a graph database. The personnel organizational relationship diagram includes basic skills, cognitive load and dynamic effective performance for each personnel node.

[0019] The allocation relationship is determined by associating the task dependency graph with the personnel organization graph, and a two-layer association fusion graph model is constructed in the graph database through the allocation relationship between tasks and personnel.

[0020] The project planning data is used to extract work order plan information and personnel static information to generate a plan baseline vector.

[0021] In another possible implementation of the first aspect, the step of determining the allocation relationship based on the association between the task dependency graph and the personnel organization graph, and constructing a two-layer association fusion graph model in the graph database through the allocation relationship between tasks and personnel, includes:

[0022] For any task node, the candidate set corresponding to the task node is obtained by filtering personnel nodes based on basic skills in the graph database.

[0023] For each personnel node in the candidate set, the matching score between the personnel node and the task node is calculated using cognitive load, dynamic effective performance, planned working hours, and budgeted cost.

[0024] A preliminary allocation scheme list is determined based on the matching scores and using preset target constraints.

[0025] Iterate through each preliminary allocation scheme list, create allocation relationships in the graph database for each task node and each personnel node, and add allocation attributes to each allocation relationship;

[0026] The cognitive load of the assigned personnel nodes is updated based on the allocation relationship to obtain a two-layer relational fusion graph model.

[0027] In another possible implementation of the first aspect, calculating the first cognitive safety index based on human-computer interaction data includes:

[0028] Information flow data and operation flow data are obtained through human-computer interaction data. The information flow data includes operation information of work tickets pushed by the system to the operators, and the operation flow data includes the sequence of actions of the operators executing work tickets and feeding back to the system.

[0029] Extract the timestamp of each message in the information stream data and construct the information stream time sequence according to the push time order;

[0030] The expected operation target and sub-tasks are obtained by parsing the operation instructions of each piece of information using natural language processing algorithms;

[0031] Extract the complexity elements of each message based on the operation instructions for each message, and calculate the complexity value of each message based on the complexity elements;

[0032] Extract the timestamp of each operation step from the operator's operation flow data, and construct the operation flow time sequence according to the execution order;

[0033] The timing relationship is determined based on the information flow timing sequence and the operation flow timing sequence, and the operation instruction corresponding to each operation step is determined through the timing relationship.

[0034] The matching degree between each operation step and the operation instruction is measured in conjunction with the expected operation objectives, and the operation quality value of each operation step is determined based on the matching degree;

[0035] Calculate the timing coupling mismatch by combining the information flow timing sequence and the operation flow timing sequence;

[0036] The semantic consistency mismatch is obtained based on the semantic similarity between the actual operation content and the expected operation goal of each operation step.

[0037] The complexity fit / misfit of each piece of information is determined by the complexity value of each piece of information and the quality of operation.

[0038] The first cognitive security index is obtained by combining temporal coupling mismatch, semantic consistency mismatch, and complexity adaptation mismatch.

[0039] In another possible implementation of the first aspect, the step of extracting the complexity element of each piece of information based on the operation instruction of each piece of information, and calculating the complexity value of each piece of information using the complexity element, includes:

[0040] Retrieve historical operation data;

[0041] The number of subtasks of an operation instruction is identified by parsing the operation instruction.

[0042] Parse operation instructions to identify conditional statements within them, and determine the nesting level of the operation instructions based on these conditional statements.

[0043] The average response time and error rate corresponding to the number of subtasks and the number of nested operation levels are determined by statistical analysis of historical operation data.

[0044] The historical execution difficulty coefficient is determined by combining the average response time and error rate;

[0045] The complexity value of each piece of information is calculated by taking into account the historical execution difficulty coefficient, the number of nested operation levels, and the number of subtasks.

[0046] In another possible implementation of the first aspect, the step of decomposing the work ticket into a work ticket set by combining the first cognitive safety index includes:

[0047] Game theory algorithms were used to calculate the main contributions of temporal coupling mismatch, semantic consistency mismatch, and complexity fit mismatch to the first cognitive safety index, in order to determine the main causes of the operator's current cognitive bias.

[0048] Construct a task dependency graph based on the complexity value, expected operation goal, and subtask of each piece of information;

[0049] Identify the subtask decomposition trigger points of the job ticket based on the task dependency graph;

[0050] A task decomposition strategy is generated by combining the trigger points of subtask decomposition with the main causes of cognitive bias.

[0051] The task decomposition strategy is used to intelligently break down the task tickets to obtain a set of task tickets.

[0052] In another possible implementation of the first aspect, the subtask decomposition trigger points include semantic task decomposition points, temporal task decomposition points, and complexity task decomposition points. The step of combining the subtask decomposition trigger points and the root cause of cognitive bias to generate a task decomposition strategy includes:

[0053] When the main contribution value of semantic consistency mismatch is the largest, the main cause of the operator's current cognitive bias is determined to be semantic comprehension bias.

[0054] The work order is analyzed and identified using a pre-defined semantic parsing model, and key phrases with semantic ambiguity are located.

[0055] Semantic error patterns are extracted based on key phrases to identify sources of semantic bias.

[0056] A semantic reconstruction and refinement strategy is adopted, and semantic deviation sources are combined to transform operation instructions into executable structured statements;

[0057] The work ticket is decomposed into tasks using executable structured statements and semantic task decomposition points;

[0058] When the main contribution value of the timing coupling mismatch is the largest, the operator's current cognitive bias is determined to be primarily due to timing cognitive bias.

[0059] Identify timing misalignment nodes by combining information flow timing sequences and operation flow timing sequences;

[0060] Determine the timing misalignment characteristics based on the timing misalignment nodes, and reconstruct the task timing logic based on the timing misalignment characteristics;

[0061] The work order is decomposed by combining the task sequence logic and the time sequence task decomposition points;

[0062] When the main contribution value of complexity fit mismatch is the largest, the main cause of the operator's current cognitive bias is determined to be excessive cognitive load, and the operator's cognitive ability baseline is obtained.

[0063] Based on the information complexity value and cognitive ability baseline, the overload target point is located, and a targeted load reduction strategy is generated based on the overload target point;

[0064] The work order is decomposed into tasks by using targeted workload reduction strategies and complexity task decomposition points.

[0065] In another possible implementation of the first aspect, the step of obtaining the global risk impact probability distribution based on the work ticket management indicator vector and the planning baseline vector using an enhanced management entropy vector fusion module and a two-layer correlation fusion graph model includes:

[0066] The second cognitive security index is converted into a forward-looking security risk index using a pre-defined nonlinear mapping function;

[0067] Real-time execution data is converted into dynamic and effective performance coefficients using a preset decay function;

[0068] A cognitive dimension feature vector is obtained by combining a forward-looking safety risk index and a dynamic effective performance coefficient.

[0069] The execution deviation feature vector is calculated by real-time execution data and the planned baseline vector.

[0070] The cognitive dimension feature vector and the execution deviation feature vector are fused using the enhanced management entropy vector fusion module to obtain the cognitive enhanced management entropy vector;

[0071] The cognitive-enhanced management entropy vector and the two-layer association fusion graph model are used as input graph structures to a pre-defined graph neural network prediction model, and the local risk probability is obtained through the pre-defined graph neural network prediction model.

[0072] The probability distribution of global risk impact is obtained from the probability of local risk.

[0073] Secondly, this application provides a machine-readable storage medium storing instructions that cause a machine to execute the aforementioned intelligent integrated management method for infrastructure projects.

[0074] Thirdly, this application provides a smart integrated management platform system for infrastructure projects, comprising:

[0075] The memory is configured to store instructions; and

[0076] The processor is configured to retrieve the instructions from the memory and, when executing the instructions, to implement the aforementioned intelligent integrated management method for infrastructure projects.

[0077] Through the aforementioned technical solutions, by acquiring project planning data, the entire lifecycle data of infrastructure projects is aggregated, reducing the problem of information silos. By constructing a two-layer correlation and fusion graph model, different information can be correlated and modeled, generating a planning baseline vector, effectively improving the comparison accuracy of planning execution. Deploying an application on the operator's terminal enables real-time collection of interactive data such as the operator's operational behavior during execution, achieving observable behavior and improving monitoring accuracy. The first cognitive safety index is calculated through dual-channel calculation of information flow and operation flow, which can quantify the degree of uncertainty in information processing and cognitive bias of operators during task execution, improving the accuracy of assessment. By identifying situations where operators have excessive cognitive load or comprehension bias and triggering a task decomposition mechanism, tasks can be matched with personnel's cognitive abilities, effectively reducing operational risks and improving the safety of task execution. The work tickets after task decomposition are re-pushed to operators for execution, and a work ticket management indicator vector is generated based on execution data and cognitive status feedback, which not only has real-time performance but also lays the foundation for subsequent risk fusion analysis. By generating a global risk impact probability distribution through an enhanced management entropy vector fusion module, the risk transmission paths between different levels of tasks, personnel, and organizations can be identified, enabling the prediction of risks from local to global levels and improving the accuracy of project risk assessment. Finally, the results of the global risk impact probability distribution are used to generate project-level risk warnings, task adjustment suggestions, and resource reallocation strategies, which can significantly improve the safety management level and responsiveness of infrastructure projects.

[0078] Other features and advantages of the embodiments of this application will be described in detail in the following detailed description section. Attached Figure Description

[0079] Figure 1 A flowchart illustrating a smart integrated management method for infrastructure projects provided in this application embodiment;

[0080] Figure 2 This is a schematic diagram illustrating a process for constructing a two-layer relational fusion graph model and generating a planning baseline vector, as provided in an embodiment of this application. Detailed Implementation

[0081] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for illustration and explanation of the embodiments of this application and are not intended to limit the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0082] It should be noted that if the embodiments of this application involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0083] Furthermore, if the embodiments of this application involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0084] Figure 1 This illustration schematically depicts a process diagram of a smart integrated management method for infrastructure projects according to an embodiment of this application. Figure 1 As shown in the figure, this application provides a smart integrated management method for infrastructure projects, which may include the following steps.

[0085] S110. Obtain project planning data for infrastructure projects;

[0086] S120. Construct a two-layer relational fusion graph model based on project planning data, and extract work ticket plan information and personnel static information from the project planning data to generate a plan baseline vector;

[0087] S130. Collect human-computer interaction data through an application deployed on the operator's terminal;

[0088] S140. Calculate the first cognitive safety index based on human-computer interaction data;

[0089] S150. When the first cognitive safety index exceeds the preset dynamic threshold, the work ticket is decomposed based on the first cognitive safety index to obtain a set of work tickets.

[0090] S160. Push a set of work tickets to the operator and obtain the work ticket management indicator vector of the operator's completed work ticket set through the application on the operator's terminal. The work ticket management indicator vector includes real-time execution data and second cognitive safety index.

[0091] S170. Based on the work ticket management indicator vector and the plan baseline vector, the global risk impact probability distribution is obtained using the enhanced management entropy vector fusion module and the two-layer association fusion graph model.

[0092] S180. Generate project-level risk warning information and planning strategies based on the global risk impact probability distribution.

[0093] First, acquire project planning data for infrastructure projects. This data includes, but is not limited to, project work breakdown structure (WBS) data, building information model (BIM) data, schedule data, construction safety measures, and staffing and job description information. This can be achieved by interfaceing with the company's existing project management system, schedule management system, and human resource management system to extract structured and semi-structured data generated during the project planning phase.

[0094] Secondly, a two-layer relational fusion graph model is constructed based on the acquired project planning data. This model represents the multi-dimensional relationships between tasks, resources, and personnel in infrastructure projects, including a task dependency layer, a resource allocation layer, and a personnel organization layer. Specifically, the project work breakdown structure data is analyzed, and a task dependency graph is established in the graph database. The task dependency graph represents the hierarchy and dependencies between project tasks, with each task node containing information such as planned working hours and budgeted costs. Thirdly, human resource organizational structure data and human resource data are analyzed, and a personnel organizational relationship graph is established in the graph database. This graph depicts the hierarchical relationships and personnel characteristics within the organization, with each personnel node containing information such as basic skills, cognitive load, and dynamic effective performance. The two-layer relational fusion graph model, based on the task-personnel allocation relationship between the task dependency graph and the personnel organizational relationship graph, enables the structured expression of project tasks and the fusion of multi-dimensional information. Furthermore, work order planning information and personnel static information are extracted from the project planning data. The work order planning information includes planned tasks, duration, and resource requirements, while the personnel static information includes each operator's job category, basic skills, job responsibilities, skill level, and available working hours. Based on the work order planning information and personnel static information, a planning benchmark vector is constructed. This vector describes the multi-dimensional standard benchmark state of the engineering tasks during the planning phase. Through this planning benchmark vector, the system can provide a clear reference framework for subsequent work execution, allowing operators' actual performance to be compared with the standard plan, thereby enabling quantitative assessment of task deviations.

[0095] Subsequently, human-machine interaction data is collected through a terminal application. The operator terminal refers to various intelligent devices used by operators at the construction site. The application running on the terminal records the operator's actions, information reception, and feedback processes during the execution of work orders. Human-machine interaction data includes operation command sequences, information reading time, number of task confirmations, number of operational errors, and abnormal reporting information during the work process.

[0096] The first cognitive safety index is calculated based on the collected human-computer interaction data. This index assesses and characterizes the operator's initial understanding and accurate grasp of the instructions, information, and operational requirements presented by the system during the execution of a work order. The first cognitive safety index quantifies the operator's cognitive safety level and potential comprehension risks at the beginning of a task, providing a basis for pre-training, task allocation, and cognitive load prediction. The collected human-computer interaction data is analyzed and divided into two categories: information flow data and operation flow data. Information flow data characterizes the work order information, operational prompts, and safety instructions pushed to the operator by the system; operation flow data characterizes the operator's actual response behavior and operational feedback. By matching and analyzing the temporal, semantic, and complexity characteristics of the information flow and operation flow, the degree of deviation between information reception and execution is identified. Based on this, the system calculates the temporal coupling mismatch, semantic consistency mismatch, and complexity adaptation mismatch between the information flow and operation flow, and obtains the first cognitive safety index from these mismatches. This entropy value reflects the operator's cognitive uncertainty during task execution. When the entropy value is high, it indicates that there is a significant deviation between the operator's cognitive understanding and actual execution. The system can trigger intelligent task breakdown or safety warning mechanisms based on this, thereby achieving dynamic monitoring and control of human-machine cognitive risks.

[0097] When the first cognitive safety index exceeds a preset dynamic threshold, it is determined that the operator has a high risk of cognitive bias in the current task scenario. The preset dynamic threshold can be determined based on the operator's average in historical task executions, for example, by calculating the average μ of the first cognitive safety index of the most recent N task executions. Subsequently, based on the characteristic distribution of the first cognitive safety index, the current work order is intelligently decomposed to generate a work order set. The system first analyzes the dominant factors causing the increase in entropy value according to the components of the first cognitive safety index. When the entropy value exceeds the limit, the task decomposition strategy is determined in combination with the characteristics of cognitive bias. If the temporal coupling mismatch contributes the most, the process sequence reorganization is triggered; if the semantic consistency mismatch contributes the most, the task semantic reconstruction and instruction optimization are performed; if the complexity adaptation mismatch contributes the most, the complex task is refined into multiple low-complexity sub-tasks. The system also uses the original work order as a basis to re-divide the task content according to the task logical relationship, operational dependency, and personnel skill matching degree, generating a work order set composed of multiple sub-task tickets. The new set of work orders, while maintaining the overall work objectives, reduces the cognitive load and execution complexity of individual tasks, thereby improving the operator's accuracy in understanding the tasks and the stability of safe execution.

[0098] Next, the set of work tickets obtained from the intelligent decomposition is pushed to the operator's terminal. The terminal application interacts with the system in real time, recording the completion status, operational behavior, and feedback information of each task. Based on the real-time execution data collected from the operator's terminal, the system generates a work ticket management indicator vector. This vector quantifies the status and deviations of each work ticket during actual execution. Through this vector, the system can comprehensively evaluate the work execution status, providing foundational data for calculating the cognitive safety index, task optimization strategies, and risk management. The work ticket management indicator vector includes real-time execution data and a second cognitive safety index. Real-time execution data reflects the operator's operation time, step completion rate, operational deviations, and abnormal event records for each work ticket task. The second cognitive safety index quantifies the operator's cognitive state and misunderstanding of safety information during task execution. Combining real-time execution data, operational deviations, and workload information, the second cognitive safety index reflects misunderstandings and operational risks caused by cognitive fatigue, information complexity, or unexpected task situations during execution, providing real-time data support for task adjustment, dynamic intervention, and safety warnings. Specifically, the system collects real-time execution data from operators' terminals. This data reflects the actual execution trajectory of operators while performing tasks on work orders. Based on this data, the system compares the operator's actual execution trajectory with the sequence of operation steps recorded in the work order. This means comparing the operator's actual sequence with the standard sequence and time recorded in the work order. The standard sequence and time are derived from historical data showing the expected completion order and time. The system determines whether the operator's actual sequence is correct and whether the operation time has exceeded the limit, identifying deviations in both the sequence and time. Deviation in the sequence refers to the number of times the operator skips, omits, performs steps in reverse order, or fails to complete tasks according to the expected logic. Deviation in the time refers to the difference between the actual execution time of each operation step and the standard planned working hours recorded in the work order. Subsequently, the second cognitive safety index is calculated using the values ​​corresponding to deviations in the order of operation steps and deviations in operation time. The second cognitive safety index = (value of deviation in step order × w) seq ) + (operation time deviation value × w time ), where w seq Indicates the order weight of the steps; w time Indicates the weight of operation time. w seq and w timeThis is determined through historical execution data. For each historical execution data point, the correlation between the deviation of the step sequence and the deviation of the operation time on the completion effect of the work order is calculated. The Pearson correlation coefficient can be used to calculate the individual influence coefficients of step sequence deviation and operation time deviation on the completion effect of the work order. These individual influence coefficients are then normalized, and the normalized correlation coefficient of step sequence deviation is divided by the sum of the correlation coefficients of step sequence deviation and operation time deviation to obtain w. seq Divide the correlation coefficient of the normalized operation time deviation by the sum of the correlation coefficients of the step sequence deviation and the operation time deviation to obtain w. time By unifying real-time execution data and the second cognitive safety index into a work ticket management indicator vector, the system can achieve dynamic monitoring and quantitative evaluation of operators' execution behavior and cognitive state.

[0099] The system utilizes an enhanced management entropy vector fusion module and a two-layer correlation fusion graph model to obtain the global risk impact probability distribution based on the work ticket management indicator vector and the planned baseline vector. The work ticket management indicator vector includes real-time execution data and a second cognitive safety index after operators complete a set of work tickets, reflecting the actual state of project task execution and the operators' latest cognitive safety level. The planned baseline vector represents the key performance indicators and safety benchmarks of the project under the expected state. The system further utilizes the enhanced management entropy vector fusion module. This module is a computational unit used to receive and process the work ticket management indicator vector and the planned baseline vector to identify and quantify the deviation between actual execution and the plan. Subsequently, the second cognitive safety index in the work ticket management indicator vector is input into a preset nonlinear mapping function to convert it into a forward-looking safety risk index, reflecting the potential impact of operators' cognitive biases on future task safety. Simultaneously, real-time execution data is input into a preset decay function to generate a dynamic effective performance coefficient, representing the real-time efficiency and accuracy of operators during task execution. Finally, the forward-looking safety risk index and the dynamic effective performance coefficient are combined to form a cognitive dimension feature vector. Simultaneously, by comparing real-time execution data with the planned baseline vector, an execution deviation feature vector is calculated to reflect the degree to which the operation deviates from the planned target. Subsequently, the cognitive dimension feature vector and the execution deviation feature vector are input into the enhanced management entropy vector fusion module for fusion, generating a cognitive enhanced management entropy vector. This cognitive enhanced management entropy vector, together with the two-layer correlation fusion graph model, serves as the input graph structure for the preset graph neural network prediction model. Through the node feature aggregation and message passing mechanism of the graph neural network, the local risk probability of each task node and personnel node is calculated. Finally, based on the propagation and correlation of local risk probabilities on the two-layer correlation fusion graph model, the global risk impact probability distribution is calculated. This is used to quantify the risk impact degree of each task, personnel, and resource in the entire infrastructure project, providing data support for project-level risk warning and intelligent planning strategy generation.

[0100] The system generates project-level risk warning information and planning strategies based on the global risk impact probability distribution. It analyzes and aggregates the global risk impact probability distribution, identifying high-risk tasks, key nodes, and areas of personnel cognitive bias within the project, and generates corresponding risk warning information, including risk type, risk level, potential impact scope, and relevant responsible personnel. Simultaneously, based on the global risk probability distribution and the dependencies between tasks, planning strategies are generated. These strategies include, but are not limited to, adjusting task execution order, reallocating personnel and resources, and breaking down or merging complex tasks. By combining risk warning information with planning strategies, project managers can intervene and optimize before task execution, thereby reducing potential risks, improving operational safety and execution efficiency, and achieving intelligent, refined, and dynamic management of infrastructure projects.

[0101] By acquiring project planning data, the entire lifecycle data of infrastructure projects was aggregated, reducing the problem of information silos. A two-layer relational fusion graph model was constructed to correlate and model different information, generating a planning baseline vector, effectively improving the accuracy of planning execution comparisons. Deploying an application on operator terminals enabled real-time collection of interactive data such as operator behavior during execution, achieving observable behavior and improving monitoring accuracy. A first cognitive safety index was calculated through dual-channel calculations of information flow and operational flow, quantifying the uncertainty and cognitive bias of operators' information processing during task execution, improving assessment accuracy. By identifying situations where operators' cognitive load is too high or their comprehension is flawed, and triggering a task decomposition mechanism, tasks can be matched to operators' cognitive abilities, effectively reducing operational risks and improving task execution safety. The decomposed work orders were then re-pushed to operators for execution, and a work order management indicator vector was generated based on execution data and cognitive status feedback. This not only provided real-time updates but also laid the foundation for subsequent risk fusion analysis. By generating a global risk impact probability distribution through an enhanced management entropy vector fusion module, the risk transmission paths between different levels of tasks, personnel, and organizations can be identified, enabling the prediction of risks from local to global levels and improving the accuracy of project risk assessment. Finally, the results of the global risk impact probability distribution are used to generate project-level risk warnings, task adjustment suggestions, and resource reallocation strategies, which can significantly improve the safety management level and responsiveness of infrastructure projects.

[0102] In one embodiment of this example, Figure 2 This illustration schematically depicts a process for constructing a two-layer relational fusion graph model and generating a planning baseline vector according to an embodiment of this application. The method may include the following steps.

[0103] S210. Extract project work breakdown structure data, human resource organization structure and human resource data from project planning data;

[0104] S220. Analyze the project work breakdown structure data and establish a task dependency graph in the graph database. Each task node in the task dependency graph contains planned man hours and budgeted cost.

[0105] S230. Analyze the human resource organizational structure and human resource data, and establish a personnel organizational relationship diagram in the graph database. The personnel organizational relationship diagram includes basic skills, cognitive load and dynamic effective performance for each personnel node.

[0106] S240. Determine the allocation relationship based on the association between the task dependency graph and the personnel organization graph, and construct a two-layer association fusion graph model in the graph database through the allocation relationship between tasks and personnel.

[0107] S250: Extract work order plan information and personnel static information from project planning data to generate a plan baseline vector.

[0108] In this embodiment, the system obtains project work breakdown structure data, human resource organizational structure data, and human resource data from the project management system and the human resource management system. The project work breakdown structure data includes information such as tasks, sub-tasks, and dependencies between tasks at each stage, which is used to characterize the hierarchical structure and logical order of engineering tasks. At the same time, the system extracts human resource organizational structure data, including organizational relationship information of each team, department, and position, as well as the job allocation of personnel. The human resource data includes information such as skill level, qualification type, work experience, and available working hours for each operator.

[0109] Next, a task dependency graph is established in the graph database by parsing the project work breakdown structure (WBS) data. The task dependency graph represents the hierarchical relationships and logical dependencies between project tasks. Each task node represents a specific task within the project, and the edges between nodes represent the dependencies between tasks. The graph database is a data storage and management system centered on a graph structure. During the construction process, each task node includes attributes such as planned working hours and budgeted cost. Planned working hours represent the estimated time required to complete the task, and budgeted cost reflects the economic input required to complete the task. Specifically, the project WBS data is used to identify tasks at all levels and their hierarchical relationships, including project phases, subtasks, and specific tasks. For each task, information such as planned working hours, budgeted cost, and task category is extracted. Then, a corresponding node is created for each task in the graph database. Node attributes include the aforementioned planned working hours and budgeted cost information, used to represent the task's time, cost, and constraint characteristics. Edges between nodes represent dependencies between tasks, including sequential dependencies, parallel dependencies, and conditional dependencies. Edge attributes can include dependency type, latency, and resource constraint information. Subsequently, the system organizes the nodes and edges into a topological structure according to the task hierarchy and dependency logic, forming a complete task dependency graph.

[0110] Subsequently, by analyzing the project's work breakdown structure data, a task dependency graph and a personnel organizational graph were established in the graph database to represent the personnel, their organizational relationships, and skill characteristics. Each node corresponds to an operator, and each edge represents the collaborative relationship between personnel. During the construction process, each personnel node includes basic skills, cognitive load, and dynamic effective performance. Basic skills describe the types of jobs and professional abilities that the personnel are competent in; cognitive load quantifies the information processing pressure and task complexity adaptability of personnel during task execution; and dynamic effective performance reflects the efficiency and accuracy of personnel in actual operations. These three types of attributes are all numerical feature parameters after feature extraction and standardization to achieve comparability and calculability of attributes between different personnel nodes. Specifically, the hierarchical relationships of each team and department in the project were identified through human resource organizational structure data, and personnel affiliation information was extracted. Simultaneously, information such as basic skills and work experience of each operator was obtained through human resource data. Next, a corresponding node was created for each operator in the graph database, with each personnel node containing attribute information such as basic skills, cognitive load, and dynamic effective performance. Subsequently, based on the human resource organizational structure information, personnel nodes are assigned organizational affiliation edges according to their respective work groups, departments, or positions, and collaborative relationships or job dependencies are established as edge attributes, forming a complete personnel organizational relationship diagram.

[0111] Next, based on the attribute information of each task node in the task dependency graph, the basic skills, cognitive load, and dynamic effective performance of each personnel node in the personnel organization graph are analyzed. Then, based on the matching of task requirements and personnel capabilities, the optimal personnel for each task are determined, and the allocation relationship between task nodes and personnel nodes is established. The allocation relationship can include various types such as single-person tasks and multi-person collaborative tasks, and can be stored as edge attributes in the graph database, representing the execution dependency, collaboration relationship, and load allocation information between tasks and personnel. Simultaneously, the task dependency graph, personnel organization graph, and task-person allocation relationship are unified into a two-layer association fusion graph model. That is, for each determined allocation relationship, a new edge is created in the graph database. For example, this edge connects the corresponding operator node and task node, and is marked as the allocation relationship type. The allocation relationship edge can also allocate start time, expected end time, allocation ratio, and other information. The two-layer association fusion graph model is formed by associating and fusing the task dependency graph, personnel organization graph, and task-person allocation relationship edges in the graph database.

[0112] Finally, by parsing the project planning data, the system obtains planning information such as task content, planned working hours, budgeted cost, and task priority for each work order. In this embodiment, static personnel information refers to the basic attributes and capabilities related to operators extracted from human resource organizational structure and human resource data; work order planning information refers to the execution standards and planning information related to the work tasks extracted from the project work breakdown structure data. Simultaneously, the system extracts static personnel information, including each operator's job category, basic skills, job responsibilities, skill level, and available working hours. Subsequently, the work order planning information and static personnel information are integrated to generate a planning baseline vector. This involves associating and matching the work order planning information and static personnel information. Specifically, the system associates and matches the planned operator ID in the work order planning information with the personnel ID in the static personnel information. Through this matching, each work order is bound to the job category, basic skills, job responsibilities, skill level, and available working hours of the operator scheduled to execute that work order, thus forming a comprehensive dataset containing both expected task attributes and executor attributes, used to generate the planning baseline vector. For example, by matching the operator IDs planned to be assigned in the work order with the personnel IDs in the personnel static information, a specific work order is bound to the attributes of the operators who are scheduled to execute that work order. This forms a comprehensive dataset containing both the expected attributes of the task and the expected attributes of the executors, thus yielding the planning baseline vector. The planning baseline vector can simultaneously characterize the time, cost, and resource requirements of the task plan, as well as the personnel skills, availability, and job matching capabilities.

[0113] By constructing a two-layer relational fusion graph model and generating a planning baseline vector, the complex relationships between tasks and personnel can be intuitively displayed, and tasks and personnel can be closely linked, which promotes the synergy between project tasks and human resources, improves overall work efficiency, and thus maximizes the overall benefits of the project.

[0114] In one embodiment of this invention, the allocation relationship is determined based on the association between the task dependency graph and the personnel organization graph, and a two-layer association fusion graph model is constructed in the graph database through the allocation relationship between tasks and personnel, including:

[0115] S310. For any task node, filter personnel nodes based on basic skills in the graph database to obtain the candidate set corresponding to the task node.

[0116] S320. For each personnel node in the candidate set, calculate the matching score between the personnel node and the task node using cognitive load, dynamic effective performance, planned working hours, and budgeted cost.

[0117] S330. Determine a preliminary allocation scheme list based on the matching scores and using preset target constraints.

[0118] S340. Traverse each preliminary allocation scheme list, create allocation relationships in the graph database for each task node and each personnel node, and add allocation attributes to each allocation relationship;

[0119] S350. Update the cognitive load of the assigned personnel nodes according to the allocation relationship to obtain the two-layer association fusion graph model.

[0120] For any given task node, the system retrieves the skill requirements of the target task node from the task dependency graph, including information such as the required professional category. Subsequently, the system searches the personnel organizational graph for the basic skill attributes of all personnel nodes and matches them with the skill requirements of the target task. Matching rules may include conditions such as complete skill category matching and qualification type compliance. Through matching calculations, a set of personnel nodes that meet the task requirements is selected to form a candidate set. Alternatively, vector similarity matching can be used, where task skill requirements and personnel skill attributes are vectorized, and the degree of matching between personnel and tasks is calculated using vector similarity (such as cosine similarity). Candidates are then ranked and selected based on similarity. The candidate set is used for subsequent task allocation analysis and optimization calculations to determine the optimal task-personnel allocation scheme.

[0121] Once the candidate set corresponding to the task node is determined, for each personnel node in the candidate set, the system calculates a matching score by considering key factors for each personnel node and the task node to be assigned. These factors include cognitive load, dynamic effective performance, planned working hours, and budgeted cost. Cognitive load refers to an assessment of the operator's current level of concentration, fatigue, and multitasking ability. Dynamic effective performance refers to the operator's actual work efficiency, task completion quality, and safety compliance in recent or similar tasks; this performance is dynamic and can be comprehensively evaluated based on historical work order completion status, real-time execution data, and supervisor assessments. Planned working hours are an inherent attribute of the task node, representing the estimated working time required to complete the task. Budgeted cost is also an inherent attribute of the task node, representing the acceptable cost range for completing the task. First, the cognitive load index and planned working hours are normalized, and then a pre-set matching algorithm is used to calculate the matching score between the personnel node and the task node. The pre-set matching algorithm can be a weighted summation model, for example, the matching score... It can be represented as:

[0122]

[0123] in: This represents the matching score between personnel node i and task node j; For the cognitive load index of personnel nodes, Its normalized score function; For the dynamic and effective performance of personnel nodes, Its normalized score function; Planned work hours for task nodes Its normalized influence function; The budget cost for the task node. The normalized influence function is defined as follows: w1, w2, w3, and w4 are preset weights that reflect the importance of each indicator in the matching calculation. They can be adjusted according to the project management strategy and satisfy w1+w2+w3+w4=1.

[0124] By using a weighted summation model, the system can quantify the suitability of each candidate for the task and rank the candidates according to their matching scores, thus providing a basis for selecting the optimal personnel for task allocation.

[0125] Subsequently, the system first sorts the candidate set for each task node according to the matching score, and takes into account constraints such as task execution priority, planned working hours, budget cost and personnel availability.

[0126] In this embodiment, the candidate set is sorted according to matching scores. Then, filtering and constraint processing are performed based on preset target constraints. These preset target constraints include, but are not limited to, the maximum available working hours for each person within a specific time period and the upper limit of the budget cost for the task. Under the premise of satisfying the above constraints, the system uses an optimal allocation decision algorithm, such as linear programming or maximum matching algorithm, to allocate tasks to the sorted candidates, generating a preliminary allocation scheme list. Each scheme clearly lists the matching relationship and attribute information between task nodes and corresponding personnel nodes, providing basic data support for subsequent dynamic optimization, real-time execution monitoring, and analysis of the two-layer relational fusion graph model.

[0127] The process iterates through each preliminary allocation scheme list. For each task node and each personnel node in the scheme, a task-person allocation relationship is established in the graph database. This involves sequentially reading each task-person pair in the preliminary allocation scheme list and creating an edge connecting the task node and its corresponding personnel node in the graph database, indicating that the personnel are assigned to perform the task. Simultaneously, allocation attributes are added to each allocation relationship to describe the characteristics of the task allocation. These attributes may include task priority, allocation start time, planned working hours, budgeted cost, and personnel role. By structurally storing the task-person allocation relationship and its attributes in the graph database, tasks, personnel, and their allocation relationships can be fully represented in a two-layer relational fusion graph model, providing data support for subsequent execution monitoring, cognitive safety assessment, and global risk prediction. Finally, by constructing task-person relationships with allocation attributes in the graph database, a multi-layer fusion graph structure that can be used for intelligent analysis and graph neural network operations is formed.

[0128] For each assigned personnel node, the updated cognitive load value is calculated based on factors such as the number of tasks already assigned, task complexity, and execution time. Cognitive load quantifies the information processing load that personnel may experience during task execution. By dynamically updating the cognitive load of all personnel nodes and storing the updated attributes in a graph database, the two-layer relational fusion graph model can simultaneously reflect task dependencies, personnel organizational structure, and the impact of task allocation on personnel cognitive states, achieving multi-dimensional relational modeling of tasks, personnel, and allocation relationships.

[0129] By constructing a two-layer relational fusion graph model, task allocation can be completed more accurately, improving the efficiency of the entire project team. Resource allocation and task arrangements can be adjusted in a timely manner to deal with potential problems, which can effectively improve the management efficiency of infrastructure projects.

[0130] In one embodiment of this example, calculating the first cognitive safety index based on human-computer interaction data includes:

[0131] S401. Information flow data and operation flow data are obtained through human-computer interaction data. The information flow data includes the operation information of the work ticket pushed by the system to the operator, and the operation flow data includes the sequence of actions of the operator executing the work ticket and feeding back to the system.

[0132] S402. Extract the timestamp of each message in the information stream data and construct the information stream time sequence according to the push time order;

[0133] S403. The expected operation target and sub-task are obtained by parsing the operation instructions of each piece of information through natural language processing algorithms;

[0134] S404. Extract the complexity elements of each message based on the operation instructions for each message, and calculate the complexity value of each message based on the complexity elements.

[0135] S405. Extract the timestamp of each operation step from the operator's operation flow data, and construct the operation flow time sequence according to the execution order;

[0136] S406. Determine the timing relationship based on the information flow timing sequence and the operation flow timing sequence, and determine the operation instruction corresponding to each operation step through the timing relationship;

[0137] S407. Combine the expected operational objectives to measure the degree of matching between each operational step and the operational instructions, and determine the operational quality value of each operational step based on the degree of matching.

[0138] S408. Calculate the timing coupling mismatch by combining the information flow timing sequence and the operation flow timing sequence;

[0139] S409. Obtain the semantic consistency mismatch degree based on the semantic similarity between the actual operation content and the expected operation goal of each operation step.

[0140] S410. Determine the complexity fit / misfit of each piece of information based on its complexity value and operation quality.

[0141] S411. The first cognitive security index is obtained by combining the temporal coupling mismatch, semantic consistency mismatch, and complexity adaptation mismatch.

[0142] In one embodiment of the present invention, information flow data and operation flow data are obtained based on human-computer interaction data. The information flow data characterizes various work order operation information pushed by the system to the operators, including but not limited to work task instructions, work environment status information, and risk warning information. Each piece of information in the information flow data is accompanied by a timestamp to identify the push sequence of the information. The operation flow data characterizes the sequence of actual operational behaviors performed by the operators in response to the work order, including interactive actions such as confirmation, execution, feedback, and verification performed by the operators after receiving instructions. Each operation in the operation flow data also includes a timestamp and a behavior identifier to describe the behavioral response trajectory during the work process.

[0143] Next, the acquired information flow data undergoes time-series processing to construct an information flow time-series sequence for subsequent analysis. This involves extracting the timestamp corresponding to each message from the information flow data. The timestamp characterizes the specific time the system pushes the information to the operator. When pushing work order operation information, the system generates a unique identifier for each message and records its sending time simultaneously. The sending time is marked by the system's built-in clock module and written into the information flow data structure as a timestamp. The information flow data can be stored in a structured log format, with each record containing a message content field, a message type field, a recipient field, and a timestamp field. During the data extraction phase, the system parses the information flow log file or database records, reading and extracting data by calling the timestamp field. Then, the information flow data is sorted by time according to the timestamps of each message, arranged sequentially according to the order of push times to form the information flow time-series sequence.

[0144] Subsequently, natural language processing algorithms are used to semantically analyze each piece of operational information in the information stream data to extract the corresponding operational instructions and determine the expected operational goals and corresponding sub-tasks. First, semantic layering analysis of the push notification text content is performed using word vector embedding to identify action verbs, object entities, and other information. Then, based on a pre-defined task knowledge base, the obtained semantic elements are matched with standard operation templates to determine the expected operational goal corresponding to the operational information. The expected operational goal can be a task action goal, such as instructing operators to perform specific actions like "closing the valve," "disconnecting the power," or "installing a guardrail." The pre-defined task knowledge base or operational ontology can be set based on historical operational data, standard operating procedures, and safety management regulations. The system performs semantic mining and task classification on work tickets generated by the company over the years, extracting operational verbs and other information to form a preliminary semantic entity set. Subsequently, based on the company's internal management regulations, the semantic entities are relationally defined to obtain a multi-level associated operational ontology structure, thus obtaining the pre-defined task knowledge base. Further, based on the operational step structure corresponding to the expected operational goal, the operational information is decomposed to obtain the corresponding set of sub-tasks. Based on the expected operation goal, the system calls the preset task knowledge base to extract the operation step sequence contained therein; then, it uses a syntax analyzer to perform structured parsing of the operation step sequence, dividing the overall task into staged operation units, and each operation unit corresponds to an independently executable subtask.

[0145] After determining the expected operational goals and subtasks, the system first acquires historical operational data to provide statistical characteristics and performance indicators of the operational behavior. Then, it performs semantic parsing on each operational instruction in the information flow data, identifying the number of subtasks contained within the instruction and further identifying conditional statements or logical branches within the instruction structure to determine the nesting level of the operational instruction. The nesting level refers to the depth of the conditional statements or logical branches contained in a single operational instruction, used to quantify the complexity of the logical branches that the instruction needs to follow during execution. Next, based on historical operational data, it statistically analyzes the average response time and error rate under different combinations of subtask numbers and nesting levels, reflecting the actual complexity level of the operation execution. Based on this, it determines the historical execution difficulty coefficient by combining the average response time and error rate. Finally, it uses the historical execution difficulty coefficient, the nesting level, and the number of subtasks as complexity factors to calculate the complexity value of each piece of information.

[0146] The system extracts timestamps from each operation step in the operator's workflow data to construct an operation flow time sequence. Specifically, the system parses each recorded operation in the workflow data and obtains the timestamp of each operation step, which represents the specific time the operation step actually occurs during execution. Then, the timestamps are sorted according to the execution order of the operation steps, generating the operation flow time sequence. By constructing this operation flow time sequence, the system can accurately reflect the order and timing of operations performed by operators during the execution of work orders.

[0147] The timestamp of each push message in the information flow time sequence is compared and matched with the timestamps of each operation step in the operation flow time sequence. The sequence and interval patterns of the operation steps and the system push instructions are analyzed to determine the temporal mapping relationship between operation steps and their corresponding instructions. For example, if the system pushes operation instruction A at time t1, and the operator subsequently executes step a1 at time t2 and step a2 at time t3, then by comparison, steps a1 and a2 can be determined to be operation steps corresponding to instruction A. If operation instruction B is pushed at time t4, and operation step b1 is executed at time t5, step b1 can be mapped to the execution step of instruction B based on the time interval and sequence relationship. By establishing temporal correlation, the specific operation steps executed by the operator after receiving a specific operation instruction can be accurately identified, ensuring that each operation step clearly corresponds to the operation instruction it triggered.

[0148] Subsequently, based on the preset expected operational goals, the actions and sequences required by the operational instructions are represented as a standard operational template. Then, the actual executed operational steps are compared with the standard template to calculate a matching degree index. This can be calculated using cosine similarity, and the matching degree can be quantitatively evaluated based on factors such as action consistency, correct execution sequence, and fulfillment of operational conditions. A higher matching degree indicates a greater degree of conformity between the operational steps and the expected operational instructions; a lower matching degree reflects deviations in the operational steps. Based on this, the matching degree of each operational step is mapped to an operational quality value, with the calculated matching degree directly used as the operational quality value.

[0149] The time stamps of each push notification in the information flow time sequence and each operation step in the operation flow time sequence are extracted and arranged chronologically. A time-series mapping relationship is established between the information flow and operation flow time sequences to determine the operation instruction and expected execution time interval for each operation step. Then, the deviation between the actual execution time and the expected execution time interval for each operation step is calculated, and the time offset of each operation step is recorded. Further, based on the time offsets of all operation steps, a weighted average is used to calculate the temporal coupling mismatch degree of the overall operation flow relative to the information flow, thus obtaining the temporal coupling mismatch degree. The temporal coupling mismatch degree is used to quantify the degree of matching between the operation behavior and the system instruction in the time dimension. Finally, the calculated temporal coupling mismatch degree is used to evaluate the timeliness of the operator's operation response and the standardization of the operation process.

[0150] Next, the expected operational goals are structurally represented according to elements such as action type and operation object to form a standardized operational template. Then, the actual execution content of each operational step in the operational flow time sequence is analyzed to extract the action type, execution object, and related condition information, and converted into a comparable semantic representation vector. Next, a natural language processing algorithm is used to calculate the semantic similarity index between each operational step and the corresponding expected operational goal. Natural language processing can be word vector, sentence vector, or other algorithms. Further, the semantic consistency mismatch degree is obtained based on the semantic similarity index value. This can be quantified by subtracting the semantic similarity from 1 to reflect the degree of semantic deviation between the actual operational steps and the expected goal. Finally, the accuracy of the operator's understanding of the operational instructions is assessed through the semantic consistency mismatch degree.

[0151] Based on the complexity value of each piece of information, the complexity value quantifies the cognitive load and operational difficulty required to understand and execute the operation instructions. Simultaneously, combined with the operation quality value of the corresponding operation steps for each piece of information, the operation quality value reflects the accuracy, standardization, and responsiveness of the operator's actual execution of the operation instructions. Then, the complexity value and the operation quality value are compared or mapped to calculate the complexity fit-mismatch, which quantifies the degree of match between the operator's actual operational performance and the information complexity requirements. For example, if the system pushes a series of instructions involving multi-device linkage and complex operation steps (high complexity value), but the operator makes multiple errors, omits key steps, or the response time significantly exceeds expectations (low operation quality value), the calculated complexity fit-mismatch will be high, indicating that the operator has failed to effectively adapt to the requirements of high-complexity information. In other words, when the operation quality value is lower than the requirements of high-complexity instructions, the complexity fit-mismatch is high, reflecting potential comprehension or operational deviations by the operator; conversely, when the operation quality value meets or exceeds the complexity requirements, the complexity fit-mismatch is low.

[0152] Subsequently, according to preset weighting rules, the temporal coupling mismatch, semantic consistency mismatch, and complexity adaptation mismatch can be comprehensively calculated to obtain the first cognitive safety index for each piece of information. The weighting coefficients can be adjusted according to the type of task, risk level, or historical execution experience to reflect the relative impact of different dimensions on cognitive load. For example, the weights can be dynamically adjusted according to the criticality or sequence of operation steps in the overall task. For example, key operation steps can be given higher weights, while secondary or auxiliary operation steps have lower weights. The calculated first cognitive safety index is used to quantify the degree of cognitive uncertainty and comprehension deviation of operators when receiving and executing specific operation information.

[0153] By calculating the first cognitive safety index, potential risks in human-computer interaction can be identified and warned in a timely manner, thereby improving the safety of human-computer interaction. It can also effectively avoid cognitive discrepancies among operators when processing work orders, reduce safety hazards caused by errors or non-standard operations, and thus improve the overall operational quality of the work process.

[0154] In one embodiment of this invention, the complexity elements of each piece of information are extracted based on the operation instructions for each piece of information, and the complexity value of each piece of information is calculated using these complexity elements, including:

[0155] S510, Obtain historical operation data;

[0156] S520: Identify the number of subtasks of the operation instruction by parsing the operation instruction;

[0157] S530. Parse the operation instructions to identify the condition judgment instructions in the operation instructions, and determine the number of operation nesting levels of the operation instructions through the condition judgment instructions;

[0158] S540. Determine the average response time and error rate corresponding to the number of subtasks and the number of nested operation levels by statistically analyzing historical operation data.

[0159] S550, combined with average response time and error rate, determines the historical execution difficulty coefficient;

[0160] S560: The complexity value of each piece of information is calculated by considering the historical execution difficulty coefficient, the number of nested operation levels, and the number of subtasks.

[0161] In one embodiment of the present invention, historical operation data is first acquired, which involves collecting and organizing actual operational information from past work records regarding operators' execution of work orders or related tasks. This information is then used for subsequent operational complexity analysis, cognitive load assessment, and risk prediction. Historical operation data includes, but is not limited to, records of the execution steps, operation sequence, and operation timestamps for each operation instruction. Data can be obtained through work management system logs, digital work order records, sensor monitoring data, or on-site feedback information.

[0162] Subsequently, the system performs semantic analysis on the content of each operation instruction to determine the number of independent execution steps or decomposable subtasks contained within the instruction. The text information of the operation instruction is segmented, part-of-speech tagging is performed, and dependency parsing is conducted to identify action verbs, operation objects, and conditional constraints within the instruction. Then, based on a preset task template, the parsed semantic elements are matched and mapped with standard operation steps to distinguish each independent operation step. The preset task knowledge base or operation ontology can be set based on historical task data and standard operating procedures. Finally, the number of mapped independent operation steps is counted, which represents the number of subtasks for that operation instruction. The number of subtasks reflects the execution complexity and cognitive load of the operation instruction, providing foundational data for subsequent identification of nested operation levels, calculation of historical execution difficulty, and determination of information complexity values.

[0163] The system parses and identifies conditional statements within operation instructions, determining the nesting level of each instruction based on these statements. Specifically, the system performs natural language processing on the text of each instruction, including word segmentation, part-of-speech tagging, and dependency parsing, to identify conditional statements or logical branches, such as structures like "if...then...otherwise...", "execute when...condition is met...", and "loop judgment". Then, based on the identified hierarchical relationships of the conditional statements, the system calculates the depth of logical nesting in each instruction, i.e., the nesting level. The nesting level quantifies the logical complexity and number of decisions required during instruction execution, reflecting the cognitive load on the operator when understanding and executing the instructions.

[0164] Once the number of subtasks and the nesting level of operations are determined, the system retrieves the actual execution data for each operation instruction from the historical job records, including the completion time and execution order of the operation steps. Then, the historical execution data is categorized and summarized according to the number of subtasks and the nesting level of each operation instruction, grouping operations with the same number of subtasks and nesting levels into the same category. Next, the average response time (the average time interval from instruction push to operation completion) and the average error rate (the proportion of errors or deviations occurring in the operation steps of that category) are calculated for each operation step within each category. Finally, the average response time and average error rate are used as fundamental indicators to measure the execution difficulty of operation instructions of different complexities, and are used to calculate the historical execution difficulty coefficient and the complexity value of each piece of information.

[0165] Finally, the historical execution difficulty coefficient, the number of nested operation levels, and the number of subtasks are used as inputs, and a preset weighting function is used to calculate the complexity value of each piece of information. The preset weighting function can be a linear weighting function, assigning fixed weights to the historical execution difficulty coefficient, the number of nested operation levels, and the number of subtasks, respectively. The complexity value is obtained by linear summation. The fixed weights can be set according to the relative influence of each factor on the operation complexity. The historical execution difficulty coefficient, the number of nested operation levels, and the number of subtasks are assigned weights a1, a2, and a3, respectively, and the complexity value is calculated by linear summation: Complexity value = a1·historical execution difficulty coefficient + a2·number of nested operation levels + a3·number of subtasks.

[0166] By calculating the complexity value of each piece of information, the complexity of operations can be quantified and the management of operation tasks can be optimized. This helps to allocate operation tasks reasonably, identify high-complexity and high-risk tasks in advance and monitor them in a focused manner, ensure the safety and reliability of operations, and improve the overall efficiency and effectiveness of the operating system.

[0167] In one embodiment of this example, the work order is broken down into a work order set by combining the first cognitive safety index, including:

[0168] S610. Calculate the main contributions of temporal coupling mismatch, semantic consistency mismatch, and complexity fit mismatch to the first cognitive safety index using game theory algorithms, in order to determine the main cause of the operator's current cognitive bias.

[0169] S620. Construct a task dependency graph based on the complexity value, expected operation objective, and subtask of each piece of information;

[0170] S630. Identify the subtask decomposition trigger points of the work ticket based on the task dependency graph;

[0171] S640. Combine subtask decomposition trigger points and cognitive bias main causes to generate task decomposition strategies;

[0172] S650: The work tickets are intelligently decomposed into a set of work tickets through a task decomposition strategy.

[0173] In one embodiment of the present invention, the main contributions of temporal coupling mismatch, semantic consistency mismatch, and complexity fit mismatch to the first cognitive security index are analyzed using the Shapley value calculation algorithm in game theory. Temporal coupling mismatch, semantic consistency mismatch, and complexity fit mismatch are considered as multiple game-playing factors affecting the first cognitive security index, and their marginal contribution to the first cognitive security index is quantified by calculating the Shapley value of each game-playing factor. Specifically, let the set of game-playing factors be N={T,S,C}, where T represents temporal coupling mismatch, S represents semantic consistency mismatch, and C represents complexity fit mismatch. Then, the Shapley value of any game-playing factor i∈N is calculated using the following formula:

[0174]

[0175] in, This represents the Shapley value of the game participation factor i. and These are used to calculate the weights of the permutations and combinations of game participants, ensuring a fair average across all permutations.

[0176] Using the formula above, the average marginal contribution of each mismatch factor across all possible combinations is calculated, thus obtaining the independent contribution of each factor to the change in the first cognitive safety index. This independent contribution is used as the primary contribution value, which is then used to determine the root cause of the operator's current cognitive bias. The root cause of cognitive bias can include semantic comprehension bias, temporal cognitive bias, and excessive cognitive load. For example, when the primary contribution value of the semantic consistency mismatch is the largest, the system determines that the operator's bias mainly stems from a misunderstanding of the work instructions.

[0177] Subsequently, a task dependency graph is constructed based on the complexity value, expected operation objective, and subtasks of each piece of information. Each operation information is treated as a task node in the graph, with each node accompanied by a corresponding complexity value. Simultaneously, combined with the parsed expected operation objective, the logical order and dependencies between tasks are determined, such as "prerequisite tasks," "parallel tasks," or "condition-triggered tasks." Then, based on the set of subtasks obtained from the semantic decomposition of the operation instructions, subtask-level nodes are established in the task dependency graph, and edge relationships are used to describe the dependency paths between subtasks and with the main task, thus obtaining the task dependency graph.

[0178] Once the task dependency graph is obtained, the dependencies, complexity values, and cognitive load distribution of each task node are analyzed. Subtask decomposition trigger points are determined through the temporal dependencies and operational logic between nodes. For example, if a task node's execution order is adjacent to a critical safety control task, and the current node has a high cognitive load, it indicates that the node is highly susceptible to attentional fluctuations or operational deviations during execution, and is thus marked as a subtask decomposition trigger point. When tasks in the task dependency graph are executed in parallel but with significantly different complexities (e.g., one parallel task is a high-risk electrical operation, and the other is a low-risk inspection task), the system identifies the intersection of these parallel nodes as potential trigger points based on the complexity gradient change.

[0179] Next, based on the subtasks identified by the task dependency graph, trigger points are broken down, and the main cognitive problems of the operators at the current stage are identified, such as comprehension bias, execution bias, or cognitive overload. Trigger point features are matched with bias types, and corresponding task decomposition strategies are dynamically generated based on a pre-defined task decomposition rule base. These strategies may include semantic reconstruction and explicit processing of task instructions when the primary cause of cognitive bias is comprehension bias; time-series optimization or step-by-step splitting of operation steps when the primary cause is execution bias; and reducing task complexity or adding auxiliary prompts when the primary cause is cognitive overload.

[0180] After obtaining the task decomposition strategy, the complex or high-risk task nodes are broken down into several sub-task units according to the decomposition rules defined in the task decomposition strategy. The operation objectives, safety measures, execution conditions and personnel matching information are reconfigured for each sub-task unit. Based on the decomposed task logic and dependencies, multiple work tickets with clear execution boundaries and independent risk control units are generated, thus forming a new set of work tickets.

[0181] By generating intelligent task decomposition strategies, work orders can be scientifically broken down to obtain a set of work orders that better suits the operator's cognitive abilities and operational procedures. This not only reduces the cognitive burden on operators and minimizes operational errors caused by cognitive biases, but also makes the execution of work orders more aligned with actual operational needs and safety requirements. This improves the overall safety, efficiency, and quality of the operation, ensuring the smooth progress of the operation and the accurate achievement of operational objectives.

[0182] In one embodiment of this example, the subtask decomposition trigger points include semantic task decomposition points, temporal task decomposition points, and complexity task decomposition points. A task decomposition strategy is generated by combining the subtask decomposition trigger points and the root cause of cognitive bias, including:

[0183] S701. When the main contribution value of semantic consistency mismatch is the largest, the main cause of the operator's current cognitive bias is determined to be semantic comprehension bias.

[0184] S702. Analyze and identify the work ticket and locate semantically ambiguous key phrases through a preset semantic parsing model;

[0185] S703. Extract semantic error patterns based on key phrases to identify sources of semantic bias;

[0186] S704. Employ semantic reconstruction and refinement strategies, and combine semantic deviation sources to transform operation instructions into executable structured statements;

[0187] S705. Task decomposition of work tickets using executable structured statements and semantic task decomposition points;

[0188] S706. When the main contribution value of the timing coupling mismatch is the largest, the operator's current cognitive bias is determined to be the timing cognitive bias.

[0189] S707. Identify timing misalignment nodes by using information flow timing sequence and operation flow timing sequence;

[0190] S708. Determine the timing misalignment characteristics based on the timing misalignment nodes, and reconstruct the task timing logic based on the timing misalignment characteristics;

[0191] S709. Decompose the work order into tasks by combining the task sequence logic and the task decomposition points.

[0192] S710. When the main contribution value of complexity fit mismatch is the largest, the main cause of the operator's current cognitive bias is determined to be excessive cognitive load, and the operator's cognitive ability baseline is obtained.

[0193] S711. Based on the complexity value of information and the cognitive ability baseline, locate the overload target point and generate a targeted load reduction strategy according to the overload target point;

[0194] S712. Task decomposition of work orders is carried out through targeted burden reduction strategies and complexity task decomposition points.

[0195] In this embodiment, semantic task decomposition points refer to nodes that are suitable for decomposition based on the semantic integrity or functional independence of the task content; temporal task decomposition points refer to nodes that are suitable for decomposition based on the temporal order or logical dependency of task execution; and complexity task decomposition points refer to nodes that are suitable for decomposition based on the inherent complexity of the task itself.

[0196] When the semantic consistency mismatch contributes the most, it indicates a significant misunderstanding of the semantics of the operational instructions or tasks pushed by the system during the current operation. In this case, the system determines that a semantic deviation has occurred during the information reception or parsing phase, meaning there is a semantic inconsistency between the actual operational steps and the expected operational goals. This cognitive bias typically manifests as misunderstanding of operational instructions, ignoring safety prompts, or misinterpreting the task intent.

[0197] When semantic comprehension discrepancies occur, a pre-defined semantic parsing model is invoked to perform word segmentation, dependency parsing, and semantic role labeling on the work order text to extract semantic structural elements from each operational instruction and task description. Subsequently, the model calculates the semantic clarity index for each phrase or term and compares it with a standard operational terminology database to identify ambiguous, unclear, or semantically polysemous key phrases, such as vague expressions like "tighten appropriately," "ensure safety," and "operate as required." Based on the semantic position of these key phrases in the work order and their dependency relationship with the operational target, the task nodes that may cause comprehension discrepancies are located. The pre-defined semantic parsing model can be a rule-based model that uses a domain semantic rule base and a safety terminology ontology to identify the semantic structure and potentially ambiguous phrases in the work order through keyword matching, syntactic dependency analysis, and other methods.

[0198] Subsequently, the semantic error patterns of each key phrase are matched and analyzed using the standard semantic library of the work order to identify potential misunderstandings. For example, "tighten appropriately" might lead to inconsistent understanding of the required force, while "operate as required" might result in deviations in the order of steps. The standard semantic library of the work order can be customized and constructed according to the actual requirements of the enterprise. By classifying and statistically analyzing the semantic error patterns of these key phrases, potential sources of deviation during the operator's comprehension phase are identified—the semantic deviation sources. These semantic deviation sources provide a precise basis for subsequent semantic reconstruction, explicit instruction processing, and correction of cognitive biases, thereby reducing safety and operational risks caused by improper comprehension during work execution.

[0199] The system utilizes semantic reconstruction algorithms to perform semantic parsing, structured decomposition, and rearrangement of the original operation instructions. This transforms ambiguous descriptions into explicit operation steps and execution conditions. For multi-step tasks, the system appends the operation objectives, execution order, and safety measures of each sub-task unit to the structured statements, forming a standardized set of directly executable instructions. For example, "tighten the bolt appropriately" is reconstructed as "Use a torque wrench to tighten the bolt to 25±2 N·m"; and "operate the valve as required" is reconstructed as "Following section 3.2 of the operating procedure, rotate the valve clockwise to the fully closed state and confirm that the indicator shows closed."

[0200] By combining the semantic task decomposition points identified through task dependency mapping, key nodes requiring further refinement or decomposition in the work process are determined. During the decomposition process, complex tasks are broken down into multiple independent sub-task units according to the location of the decomposition points and the content of structured statements. Each sub-task unit contains clear operational steps, execution conditions, parameter ranges, and safety measures. For example, for a task described in an original work order as checking and adjusting valve pressure, the key decomposition points in this task are first identified as valve pressure measurement and valve adjustment. Subsequently, it is broken down into two sub-task units: Sub-task unit 1 involves using a pressure gauge to measure the valve inlet pressure, recording the pressure value, and ensuring that the measurement error is within ±0.5 bar; Sub-task unit 2 involves adjusting the valve opening based on the measurement results to achieve the target value of 10 bar ± 0.5 bar at the valve outlet pressure, and confirming that the pressure indicator is displaying normally. Each sub-task unit contains clear operational steps and safety measures.

[0201] When the main contribution value of the timing coupling mismatch is the largest, it indicates that the operator has significantly deviated from the time sequence, sequential dependencies, or task rhythm of the operation steps during task execution. In this case, the system judges that the operator's cognitive bias is mainly reflected in insufficient understanding of the temporal relationship of the task or an error in the execution order, i.e., timing cognitive bias.

[0202] Subsequently, the system analyzes the sequence and interval patterns of the operation steps and the system's push instructions, calculating the time deviation between each operation step and its corresponding instruction. When the execution time of an operation step deviates significantly from the expected push time, and this deviation may affect the execution order or security of subsequent tasks, the system marks the task node corresponding to that operation step as a timing misalignment node. Timing misalignment nodes are used to identify timing perception biases that occur during the operation process.

[0203] For each identified timing misalignment node, its specific type and degree of deviation are analyzed to determine the timing misalignment characteristics. For example, if an operation step corresponding to an instruction is missing in the operation flow, the characteristic is operation omission; if an operation step occurs before the instruction is pushed, the characteristic is operation premature. Further, the identified timing misalignment characteristics are analyzed, and combined with the hierarchical decomposition structure of the task, the dependencies between the operation steps involved in the misalignment node and other steps are identified. For example, if a critical pre-operation is omitted or delayed, the validity or timing of all its subsequent dependent operations may be affected. Based on the timing misalignment characteristics and their impact analysis, the original task timing logic is reconstructed. For example, for omitted operations, if the omitted operation is non-critical, the reconstructed logic can mark it as optional; if it is a critical operation, a mandatory checkpoint may be introduced.

[0204] Subsequently, by combining the temporal task decomposition points identified through the task dependency graph, key task nodes that need to be refined or decomposed are determined. During the task decomposition process, based on the location of the decomposition points and the temporal logic, the complex task is divided into multiple independent sub-task units, each of which contains clear operation steps, execution conditions, time windows, etc.

[0205] When the main contribution value of complexity mismatch is at its maximum, it indicates that the current operator's cognitive load is too high. This means the inherent complexity of the system-pushed instructions or tasks exceeds the operator's current or inherent cognitive processing capabilities, thus affecting the accuracy of the operation. Based on this diagnosis, the operator's cognitive ability baseline will be further obtained. The cognitive ability baseline refers to the average level of the operator's attention, information processing speed, and multitasking ability when handling tasks of different complexities.

[0206] The work order is broken down into tasks using targeted workload reduction strategies and complexity task decomposition points. This involves comparing the complexity value of each piece of information with the operator's cognitive ability baseline to identify specific points of significant mismatch, i.e., overload targets. For example, if a piece of information has a high "operation nesting level," but the operator's working memory baseline is low, then the high nesting level may be identified as a target causing cognitive overload. Once the specific targets causing cognitive overload are identified, targeted workload reduction strategies are generated based on these targets. If the target is a high nesting level, the strategy might be to flatten the complex nested instructions. Specific targeted burden reduction strategies could be as follows: When the target is a high level of nesting, decompose complex nested instructions into multiple peer-level subtasks using semantic hierarchical parsing algorithms (such as context-free grammars) and present them in a parallel structure; when the target is a high information density, break down long sentence-like tasks into several semantically short sentences and present them step by step in logical order; when the target is a temporal conflict, adjust the execution order of tasks or set mutual exclusion dependencies between tasks to reduce the burden of task switching; when the target is a high level of operational complexity, decompose complex tasks containing multiple sub-steps into several sub-work tickets; when the target is a high level of semantic ambiguity, generate standard operation instructions, key prompts, or graphical operation guidance.

[0207] Once the targeted load reduction strategy is determined, the work order is broken down into tasks based on the strategy and the complexity task breakdown points. For example, the work order instruction is: Complete a comprehensive overhaul of equipment A, including: 1. Power off; 2. Inspect the wiring; 3. Replace component X, which includes multiple nested steps such as disassembly, installation, and testing; 4. Record the overload results. If the identified overload target is a high-level nested task, replacing component X might be identified as a complexity task breakdown point because it contains multiple nested steps. After breakdown, the original work order might be split into: Work Order 1: Power off, inspect wiring. Work Order 2: Replace component X (internal steps are flattened or further broken down into smaller sub-work orders). Work Order 3: Record the overload results.

[0208] By generating task decomposition strategies, the feasibility of work orders can be improved, operational risks can be reduced, the work efficiency and quality of operators can be enhanced, and the smooth progress of the work process can be ensured.

[0209] In one embodiment of this example, the global risk impact probability distribution is obtained based on the work ticket management indicator vector and the plan baseline vector using an enhanced management entropy vector fusion module and a two-layer correlation fusion graph model, including:

[0210] S810. The second cognitive security index is converted into a forward-looking security risk index using a preset nonlinear mapping function.

[0211] S820: Convert real-time execution data into dynamic effective performance coefficients using a preset decay function;

[0212] S830, combining the forward-looking safety risk index and the dynamic effective performance coefficient to obtain the cognitive dimension feature vector;

[0213] S840. The execution deviation feature vector is calculated by real-time execution data and the planned baseline vector.

[0214] S850. The cognitive dimension feature vector and the execution deviation feature vector are fused using the enhanced management entropy vector fusion module to obtain the cognitive enhanced management entropy vector.

[0215] S860. The cognitive-enhanced management entropy vector and the two-layer association fusion graph model are used as input graph structures to the preset graph neural network prediction model, and the local risk probability is obtained through the preset graph neural network prediction model.

[0216] S870. Obtain the global risk impact probability distribution based on the local risk probability.

[0217] The system obtains a second cognitive safety index calculated by operators during the execution of work orders. This entropy value reflects the operator's cognitive load and comprehension biases under the current task execution state. Subsequently, through a pre-defined nonlinear mapping function, this cognitive safety entropy value is mapped to a forward-looking safety risk index to quantify the safety risks that the operator may cause. The nonlinear mapping function can be designed based on historical work data, task complexity, operating environment, and safety accident statistics, so that even small changes in the cognitive safety entropy value can sensitively reflect the changing trends of potential risks. For example, when the second cognitive safety index exceeds a certain threshold, the mapped forward-looking safety risk index will increase significantly, indicating that the operator may be in a high-risk operating situation.

[0218] Converting real-time execution data into dynamic effective performance coefficients using a preset decay function involves weighting the real-time execution data based on this function. This results in a higher impact from recent task performance and a gradually decreasing impact from earlier historical operation data, thus yielding a dynamic effective performance coefficient that reflects the operator's current performance capability. The decay function can employ exponential decay, weighted moving average, or other non-linear decay methods to ensure that the dynamic performance coefficient sensitively reflects recent operational deviations or performance changes. For example, if an operator experiences delays or errors in consecutive subtasks, the dynamic effective performance coefficient will drop rapidly; conversely, the contribution of earlier successful operations will decay over time.

[0219] The forward-looking safety risk index and dynamic effective performance coefficient are integrated according to a preset feature fusion method to generate a cognitive dimension feature vector that reflects the operator's current cognitive state, execution ability and potential risk level. This feature vector can be spliced ​​together in a linear combination manner. The cognitive dimension feature vector can be used for subsequent work order management entropy vector calculation, two-layer correlation fusion graph model analysis and global risk impact probability distribution assessment.

[0220] Next, the real-time execution data of the operator's work order, including the actual completion time and sequence of each operation step, is compared with the planned baseline vector. This vector includes the expected execution sequence, time nodes, operational standards, and safety constraints of the task. The timing deviation, semantic deviation, complexity adaptation deviation, and other execution deviation indicators for each operation step are calculated and integrated into a multi-dimensional vector form, forming an execution deviation feature vector. This execution deviation feature vector comprehensively reflects the degree to which the operator deviates from the expected plan during task execution. The two types of vectors are fused using an enhanced management entropy vector fusion module. The fusion process can employ weighted superposition to generate a cognitive enhanced management entropy vector that comprehensively reflects both cognitive state and execution deviation. In this embodiment, the enhanced management entropy vector fusion module specifically includes an input interface layer, a standardization and preprocessing layer, a weight allocation layer, and a fusion calculation layer. The input interface layer receives cognitive dimension feature vectors and execution deviation feature vectors. The standardization and preprocessing layer standardizes the received cognitive dimension feature vectors and execution deviation feature vectors (e.g., Min-Max standardization). The weight allocation layer assigns corresponding weights to the standardized cognitive state vectors and execution deviation feature vectors, which can be based on historical data analysis. The fusion calculation layer uses a weighted superposition algorithm to linearly combine the standardized and weighted cognitive state components and execution deviation components. According to the formula... Calculate and generate the cognitively enhanced management entropy vector, where E p w represents the entropy vector of cognitively enhanced management. p,i Indicates weight; C i D represents the standardized cognitive state vector; n represents the total number of dimensions of the execution bias feature vector; j This represents the standardized execution deviation feature vector. The cognitively enhanced management entropy vector can more comprehensively quantify operators' cognitive risks and operational deviations during task execution, providing data support for two-layer correlation fusion graph model analysis, global risk probability distribution calculation, task adjustment, and safety early warning.

[0221] This paper combines cognitively enhanced management entropy vectors with a two-layer relational fusion graph model to form an input graph structure containing task nodes, personnel nodes, their dependencies, and cognitive state information. This input graph structure is then fed into a pre-defined graph neural network prediction model, which may include graph convolutional networks or other deep learning networks for multi-layer graph structures. The graph neural network iteratively propagates and aggregates the node features and graph topology to output the local risk probability of each task node or key node. This local risk probability quantifies the potential risk level that operators may encounter when performing tasks at specific nodes. Finally, the local risk probabilities of each task node or key node predicted by the graph neural network reflect the potential risk level that operators may encounter during the execution of specific tasks or nodes. Subsequently, the task dependencies, personnel organizational relationships, and operational logic of the two-layer relational fusion graph model are combined to aggregate the local risk probabilities, including risk propagation along task dependency paths, cross-node risk coupling, and the weighted distribution of the contribution of task nodes to the overall project risk. The final result is a global risk impact probability distribution reflecting the entire workflow or project level.

[0222] This application provides a machine-readable storage medium storing instructions that cause a machine to execute the aforementioned intelligent integrated management method for infrastructure projects.

[0223] This application also provides a smart integrated management platform system for infrastructure projects, including:

[0224] The memory is configured to store instructions; and

[0225] The processor is configured to retrieve instructions from memory and, when executing instructions, to implement the aforementioned intelligent integrated management method for infrastructure projects.

[0226] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0227] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0228] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0229] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0230] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0231] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0232] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0233] It should also be noted that 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 process, method, article, or apparatus. Unless otherwise specified, 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 that element.

[0234] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A smart integrated management method for capital construction projects, characterized in that, The method comprises the following steps: acquiring project planning data of a capital construction project; extracting project work breakdown structure data, human organization structure and human resource data from the project planning data; analyzing the project work breakdown structure data to establish a task dependency graph in a graph database, wherein each task node in the task dependency graph contains planned working hours and budget cost; analyzing the human organization structure and human resource data to establish a personnel organization relationship graph in the graph database, wherein each personnel node in the personnel organization relationship graph contains basic skills, cognitive load and dynamic effective performance; determining an allocation relationship based on the association of the task dependency graph and the personnel organization relationship graph, and constructing a double-layer association fusion graph model in the graph database through the allocation relationship between tasks and personnel; extracting work order plan information and personnel static information from the project planning data to generate a plan benchmark vector; collecting human-computer interaction data through an application program deployed on an operator terminal; obtaining information flow data and operation flow data from the human-computer interaction data, wherein the information flow data includes work order operation information pushed by the system to the operator, and the operation flow data includes a behavior sequence of the operator executing the work order and feeding back to the system; extracting the time stamp of each piece of information in the information flow data, and constructing an information flow time sequence according to the push time sequence; obtaining an expected operation target and subtasks by analyzing the operation instruction of each piece of information through a natural language processing algorithm; extracting the complexity element of each piece of information according to the operation instruction of each piece of information, and calculating the complexity value of each piece of information through the complexity element; extracting the time stamp of each operation step through the operation flow data of the operator, and constructing an operation flow time sequence according to the execution sequence; determining the time sequence association relationship according to the information flow time sequence and the operation flow time sequence, and determining the operation instruction corresponding to each operation step through the time sequence association relationship; measuring the matching degree between each operation step and the operation instruction in combination with the expected operation target, and determining the operation quality value of each operation step based on the matching degree; calculating the time sequence coupling mismatch degree in combination with the information flow time sequence and the operation flow time sequence; obtaining the semantic consistency mismatch degree according to the semantic similarity between the actual operation content of each operation step and the expected operation target; determining the complexity adaptation mismatch degree of each piece of information through the complexity value and the operation quality of each piece of information; obtaining a first cognitive safety index by comprehensively considering the time sequence coupling mismatch degree, the semantic consistency mismatch degree and the complexity adaptation mismatch degree; when the first cognitive safety index exceeds a preset dynamic threshold, the work order is disassembled in combination with the first cognitive safety index to obtain a work order set; the operator is pushed with the work order set, and the operator terminal application program obtains a work order management index vector of the operator completing the work order set, wherein the work order management index vector includes real-time execution data and a second cognitive safety index, and the second cognitive safety index is calculated through the corresponding values of operation step sequence deviation and operation time deviation. According to the work order management index vector and the plan reference vector, a global risk influence probability distribution is obtained by using an enhanced management entropy vector fusion module and a double-layer association fusion graph model, the enhanced management entropy vector fusion module being a calculation unit for receiving and processing the work order management index vector and the plan reference vector to identify and quantify the deviation between actual execution and planning; Project-level risk early warning information and planning strategies are generated according to the global risk influence probability distribution.

2. The method of claim 1, wherein, The assignment relationship is determined based on the task dependency graph and the personnel organization relationship graph, and a double-layer association fusion graph model is constructed in the graph database based on the assignment relationship between tasks and personnel, including: For any task node, a candidate person set corresponding to the task node is obtained in the graph database according to the basic skills; For each personnel node in the candidate person set, a matching score between the personnel node and the task node is calculated by considering the cognitive load, dynamic effective performance, planned working hours, and budget cost; A preliminary assignment scheme list is determined according to the matching score and a preset target constraint condition; Each preliminary assignment scheme list is traversed, and for each task node and each personnel node, an assignment relationship is created in the graph database, and an assignment attribute is added to each assignment relationship; The double-layer association fusion graph model is updated by updating the cognitive load of the assigned personnel node.

3. The method of claim 1, wherein, The complexity factor of each piece of information is extracted according to the operation instruction of each piece of information, and the complexity value of each piece of information is calculated by using the complexity factor, including: Obtain historical operation data; The number of sub-tasks in the operation instruction is identified by parsing the operation instruction; The number of operation nesting layers in the operation instruction is identified by parsing the conditional judgment instruction in the operation instruction; The average response time and error rate corresponding to the number of sub-tasks and the number of operation nesting layers are determined by statistical analysis of historical operation data; The historical execution difficulty coefficient is determined by combining the average response time and error rate; The complexity value of each piece of information is calculated by using the historical execution difficulty coefficient, the number of operation nesting layers, and the number of sub-tasks.

4. The method of claim 1, wherein, The work order is decomposed to obtain a work order set by combining the first cognitive safety index, including: The main contribution value of the time sequence coupling mismatch degree, the semantic consistency mismatch degree, and the complexity adaptation mismatch degree to the first cognitive safety index is calculated respectively by using a game theory algorithm to determine the main cause of the cognitive bias of the operator at present; A task dependency graph is constructed according to the complexity value of each piece of information, the expected operation target, and the sub-tasks; The sub-task decomposition trigger point of the work order is identified according to the task dependency graph; A task decomposition strategy is generated by combining the sub-task decomposition trigger point and the main cause of the cognitive bias; The work order set is obtained by intelligently decomposing the work order according to the task decomposition strategy.

5. The method of claim 4, wherein, The sub-task decomposition trigger point includes a semantic task decomposition point, a time sequence task decomposition point, and a complexity task decomposition point, and the task decomposition strategy is generated by combining the sub-task decomposition trigger point and the main cause of the cognitive bias, including: When the main contribution value of the semantic consistency mismatch degree contribution is the largest, it is determined that the main cause of the cognitive bias of the operator at present is semantic understanding bias. The work order is analyzed and recognized by a preset semantic analysis model, and a key phrase with semantic ambiguity is located; Based on the key phrase, a semantic error mode is extracted to determine the source of semantic deviation; A semantic reconstruction and refinement strategy is adopted to convert the operation instruction into an executable structured statement combined with the semantic deviation source; The work order is task-decomposed through the executable structured statement and the semantic task-decomposition point; When the main contribution value of the timing coupling mismatch degree is the largest, it is determined that the current cognitive bias of the operator is caused by timing cognitive bias; The timing dislocation node is identified through the information flow timing sequence and the operation flow timing sequence; The timing dislocation feature is determined according to the timing dislocation node, and the task timing logic is reconstructed according to the timing dislocation feature; The work order is task-decomposed in combination with the task timing logic and the timing task-decomposition point; When the main contribution value of the complexity adaptation mismatch degree is the largest, it is determined that the current cognitive bias of the operator is caused by excessive cognitive load, and the cognitive ability baseline of the operator is obtained; Based on the complexity value of the information and the cognitive ability baseline, the overload target point is located, and a targeted load reduction strategy is generated according to the overload target point; The work order is task-decomposed through the targeted load reduction strategy and the complexity task-decomposition point.

6. The method of claim 1, wherein, The global risk influence probability distribution is obtained by using the enhanced management entropy vector fusion module and the double-layer association fusion graph model according to the work order management index vector and the plan benchmark vector, including: The second cognitive safety index is converted into a forward-looking safety risk index by using a preset nonlinear mapping function; The real-time execution data is converted into a dynamic effective performance coefficient by using a preset decay function; The cognitive dimension feature vector is obtained by combining the forward-looking safety risk index and the dynamic effective performance coefficient; The execution deviation feature vector is calculated by using the real-time execution data and the plan benchmark vector; The cognitive dimension feature vector and the execution deviation feature vector are fused by using the enhanced management entropy vector fusion module to obtain a cognitive enhanced management entropy vector; The cognitive enhanced management entropy vector and the double-layer association fusion graph model are used as the input graph structure of the preset graph neural network prediction model, and the local risk probability is obtained by using the preset graph neural network prediction model; The global risk influence probability distribution is obtained according to the local risk probability.

7. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores instructions for causing a machine to execute the intelligent integrated management method for infrastructure projects according to any one of claims 1-6.

8. An intelligent integrated management platform system for infrastructure projects, characterized in that, Including: a memory configured to store instructions; and a processor configured to call the instructions from the memory and enable the intelligent integrated management method for infrastructure projects according to any one of claims 1-6 when executing the instructions.

Images