A scientific research project whole life cycle management method and system

Abstract

The application discloses a kind of scientific research project whole life cycle management method and system, it is related to project supervision technical field.The application obtains the input-output dependency between the task decomposition structure of project under research and each subtask, each subtask is abstracted as the progress puzzle unit with boundary constraint, each progress puzzle unit is spliced to obtain project parallel progress puzzle network, risk node is established by pre-setting risk event category, the project causal dependency between risk node and transmission probability is constructed, and the cross-project risk causal network is obtained.The deviation of actual progress and planned progress is quantified as puzzle matching deviation degree;When the puzzle matching deviation degree of puzzle unit exists and exceeds the preset threshold value, locate the secondary risk spread arc region in the cross-project risk causal network, according to the risk propagation direction and the predicted impact range of secondary risk spread arc region, retrieve the verified coping strategy with the highest matching degree of current risk propagation path in historical coping strategy library and execute.

Description

Technical Field

[0001] This invention relates to the field of project management technology, specifically to a method and system for managing the entire lifecycle of scientific research projects. Background Technology

[0002] Scientific research project management is a crucial link in ensuring the smooth implementation of innovation activities. Traditional management methods mostly rely on static tools such as Gantt charts, critical path methods, and work breakdown structures to track and control projects by setting milestone nodes and linear schedules. However, scientific research projects are highly exploratory, uncertain, and dynamic, with complex technical interface dependencies, resource competition relationships, and data interaction constraints among subtasks. Traditional methods are difficult to effectively model this task network where "parallel dependencies and resource coupling" coexist.

[0003] Meanwhile, existing management tools typically handle risk events using a separate "risk register" model, which involves pre-identifying a list of risks and developing contingency plans. However, this approach overlooks the causal transmission and cascading amplification effects between risks. For example, the departure of key personnel can directly lead to obstruction of a technological path, which in turn can cause supply chain disruptions and budget overruns. This chain-like risk propagation is particularly prominent in research projects, and traditional methods lack the ability to quantitatively extrapolate risk propagation paths. Furthermore, most current project management software is based on static plans, requiring manual adjustments once deviations occur. It cannot dynamically update task boundaries, time windows, and risk propagation probabilities based on actual execution feedback, and it is even more difficult to achieve adaptive scheduling and closed-loop optimization throughout the entire lifecycle.

[0004] Therefore, there is an urgent need for a research project lifecycle management technology that can integrate task-dependent modeling, risk-causal networks, and dynamic feedback learning. To this end, we provide a research project lifecycle management method and system. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for managing the entire life cycle of scientific research projects, in order to address the shortcomings in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for managing the entire lifecycle of scientific research projects includes the following steps: Step S1: Obtain the task decomposition structure of the project under research and the input-output dependencies between each subtask. Abstract each subtask into a schedule puzzle unit with boundary constraints. Extract the technical interface parameters, resource requirement attributes and time window labels of each schedule puzzle unit. Establish a puzzle adjacency matrix. Based on the puzzle adjacency matrix, stitch together each schedule puzzle unit to obtain the project parallel schedule puzzle network. Step S2: Preset risk event types to establish risk nodes. Construct project causal dependencies and transmission probabilities between risk nodes by corresponding historical project risk events for various risk events, and obtain a cross-project risk causal network. Associate and bind the cross-project risk causal network with the jigsaw puzzle adjacency matrix so that each progress jigsaw puzzle unit is loaded with the corresponding risk triggering conditions and transmission weights, and complete the risk perception initialization of the project parallel progress jigsaw puzzle network. Step S3: During the execution of the scientific research project, periodically collect the progress completion rate and risk trigger indicators of each progress puzzle unit, and quantify the deviation between the actual progress and the planned progress as the puzzle matching deviation degree; when the puzzle matching deviation degree of a puzzle unit exceeds the preset threshold, locate the risk node corresponding to the progress puzzle unit in the cross-project risk causal network as the disturbance source, and deduce the affected adjacent progress puzzle units step by step along the causal dependency edge. Generate the secondary risk spread arc area according to the transmission probability and the puzzle adjacency relationship, and evaluate the risk propagation direction and expected impact range of the secondary risk spread arc area. Step S4: Based on the risk propagation direction and expected impact range of the secondary risk spread arc, retrieve the verified response strategy with the highest matching degree with the current risk propagation path from the historical response strategy library. Then, update the cross-project transmission probability based on the execution feedback of the verified response strategy, adjust the boundary constraints and time windows of the affected schedule puzzle units in the parallel schedule puzzle network, and recalculate the global stitching path of the project's parallel schedule puzzle network until the puzzle matching deviation of all schedule puzzle units converges to the tolerance range.

[0007] Furthermore, the process of obtaining the task decomposition structure of the ongoing project and the input-output dependencies between each subtask, and abstracting each subtask into a schedule puzzle unit with boundary constraints, includes: Obtain the project proposal, task contract or implementation plan of the target scientific research project, and parse out the task decomposition structure of the project level. The task decomposition structure breaks down the overall work of the project into several sub-tasks. The task decomposition structure is analyzed in depth to identify the input-output dependencies between subtasks. These dependencies are recorded as directed edges, where the direction of the edge represents the direction from the preceding subtask to the subsequent subtask, and the weight of the edge can be assigned according to the tightness of the dependency. For each subtask in the task decomposition structure, it is abstracted into a progress puzzle unit; the progress puzzle unit represents the smallest indivisible logical block of project progress management, which encapsulates all the attributes and boundary constraints required to complete the subtask. The boundary constraints specifically include technical interface parameter boundaries, resource requirement attribute boundaries, and time window labels.

[0008] Furthermore, the process of extracting the technical interface parameters, resource requirement attributes, and time window labels of each schedule puzzle unit, establishing a puzzle adjacency matrix, and then piecing together the various schedule puzzle units based on the puzzle adjacency matrix to obtain the project parallel schedule puzzle network includes: Traverse all progress puzzle units, extract the attributes of each dimension to form an attribute vector set, named the puzzle adjacency matrix, and construct the initial project parallel progress puzzle network with each progress puzzle unit as a node and the directed edges in the puzzle adjacency matrix as connecting lines. During the construction of the parallel schedule puzzle network for the project: scan the puzzle adjacency matrix to identify all schedule puzzle unit pairs that do not have direct or indirect dependencies, and mark the independent schedule puzzle unit pairs as units that can be executed in parallel.

[0009] Furthermore, the process of establishing risk nodes by pre-setting risk event types, and constructing project causal dependencies and transmission probabilities between risk nodes by corresponding to various risk events and historical project risk events, to obtain a cross-project risk causal network includes: Establish a standardized ontology of risk event types, which covers all types of risks that may be encountered throughout the entire life cycle of a research project and classifies them by level. For each type of risk event, a risk node is created, and historical project risk event records of the target research institution or field are retrieved. These historical project risk event records record the occurrence time, development process, countermeasures, and final consequences of the risk events in the corresponding research projects. Causal chain analysis is performed on each historical project risk event record. Based on the above historical causal chain, a risk causal graph is constructed. The directed edges in the graph represent that the occurrence of one risk node will directly lead to the occurrence of another risk node. The propagation probability is calculated for each directed edge. All risk nodes, their causal directed edges, and propagation probabilities are integrated to obtain a cross-project risk causal network.

[0010] Furthermore, the cross-project risk causal network is linked and bound to the puzzle adjacency matrix, so that each schedule puzzle unit is attached with the corresponding risk triggering conditions and propagation weights. The process of initializing the risk perception of the project parallel schedule puzzle network includes: Traverse each schedule puzzle unit in the project's parallel schedule puzzle network and perform the following binding operations, including attaching risk trigger conditions and attaching propagation weights; The transmission weight indicates that only when there is a strong task dependency between two puzzle units can the risk on one progress puzzle unit be more easily transmitted to the other progress puzzle unit. If there is no dependency between two progress puzzle units, even if the risk could be transmitted in the cross-project risk causal network, it will not be transmitted in the actual project. Thus, after all bindings are completed, the risk perception initialization of the project parallel progress puzzle network is completed.

[0011] Furthermore, the process of periodically collecting progress completion rates and risk trigger indicators for each progress puzzle unit during the execution of scientific research projects, and quantifying the deviation between actual progress and planned progress as puzzle matching deviation, includes: Set up a project health monitoring cycle, and automatically or manually collect the progress completion rate and risk trigger indicators of all active progress puzzle units when each project health monitoring cycle arrives. The puzzle matching deviation is calculated for each progress puzzle unit based on the progress completion rate and risk trigger indicators. The puzzle matching deviation represents the degree of deviation between the progress lag and the risk trigger of the corresponding progress puzzle unit.

[0012] Furthermore, when the puzzle matching deviation of a puzzle unit exceeds a preset threshold, the process of locating the risk node corresponding to the schedule puzzle unit in the cross-project risk causal network as the disturbance source, and deducing the affected adjacent schedule puzzle units step by step along the causal dependency edges includes: When the matching deviation of a progress puzzle unit exceeds a preset deviation threshold, the corresponding progress puzzle unit is marked as a disturbance source unit; otherwise, it is not marked. In the cross-project risk causal network, the risk node with the highest trigger degree among the risk nodes attached to the schedule puzzle unit is located and denoted as the starting risk node. Starting from the starting risk node, along the directed edges of the cross-project risk causal network, and taking into account the dependencies in the puzzle adjacency matrix, a restricted breadth-first risk deduction is performed. Based on the risk deduction results, the risk radiates outward layer by layer with the disturbance source unit as the center. Mark the position of all schedule puzzle units covered by risk transmission paths in the project parallel schedule puzzle network, and then extract the secondary risk spread arc area from the project parallel schedule puzzle network by forming a sub-graph composed of the marked schedule puzzle units and the risk transmission paths between them. Within the secondary risk spread arc zone, the cumulative effective intensity of each risk transmission path is obtained, and the risk transmission path with the largest cumulative effective intensity is traversed to assess the expected impact range of the secondary risk spread arc zone. The expected impact range is quantified by the set of affected units and the comprehensive impact index.

[0013] Furthermore, based on the risk propagation direction and expected impact range of the secondary risk contagion arc, the process of retrieving the most validated response strategy with the highest matching degree to the current risk propagation path from the historical response strategy database includes: Before a research project is launched, collect and structure the historical project records of risk response for typical research projects in the unit and field. Each historical project record is a verified response strategy, which includes the risk propagation model, response strategy description and implementation feedback effect. Once a secondary risk spread arc zone is generated, the risk propagation path and the type of the affected progress puzzle unit are extracted, and the matching degree is calculated with the risk propagation pattern of the verified response strategy. The top three verified response strategies with the highest matching degree are selected as the candidate response strategy set. If the highest matching degree is lower than the preset threshold, the project manager will manually intervene to formulate a new verified response strategy.

[0014] Furthermore, the process of updating the cross-project propagation probability based on the execution feedback of the validated response strategy, adjusting the boundary constraints and time windows of the affected schedule puzzle units in the parallel schedule puzzle network, and recalculating the global stitching path of the project's parallel schedule puzzle network until the puzzle matching deviation of all schedule puzzle units converges to the tolerance range includes: The verified response strategy with the highest matching degree is selected from the candidate response strategy set and executed. During and after execution, the execution feedback data is continuously collected. The execution feedback data is the number of times risk transmission actually occurred and the number of times transmission did not occur within the original secondary risk spread arc area. The relevant transmission probabilities in the cross-project risk causal network are dynamically corrected. Based on the selected response strategy and the updated transmission probabilities, the boundary constraints and time windows of the project parallel schedule puzzle network are adjusted. The adjustment targets all affected schedule puzzle units located in the secondary risk spread arc zone. After completing the boundary constraints and time window adjustments, scan all affected schedule puzzle units, re-evaluate the puzzle matching deviation, and determine whether the deviation of all schedule puzzle units has been reduced to the preset tolerance range. Based on the judgment result, determine whether the current verified response strategy is insufficient or whether the adjusted project parallel schedule puzzle network has generated new disturbance sources. Restart monitoring under the new project parallel schedule puzzle network state.

[0015] A scientific research project lifecycle management system includes a project risk management module, an ongoing project dynamic monitoring module, and a risk control module; The project risk management module is used to obtain the task decomposition structure of the project under development and the input-output dependencies between each subtask. Each subtask is abstracted into a schedule puzzle unit with boundary constraints. The technical interface parameters, resource requirement attributes and time window labels of each schedule puzzle unit are extracted to establish a puzzle adjacency matrix. Based on the puzzle adjacency matrix, each schedule puzzle unit is pieced together to obtain a parallel schedule puzzle network for the project. Risk nodes are established by pre-setting risk event types. The project causal dependencies and transmission probabilities between risk nodes are constructed by corresponding historical project risk events for various risk events to obtain a cross-project risk causal network. The cross-project risk causal network is associated and bound to the puzzle adjacency matrix so that each schedule puzzle unit is equipped with corresponding risk triggering conditions and transmission weights. The dynamic monitoring module for ongoing projects is used to periodically collect the progress completion rate and risk trigger indicators of each progress puzzle unit during the execution of scientific research projects. The deviation between the actual progress and the planned progress is quantified as the puzzle matching deviation. When the puzzle matching deviation of a puzzle unit exceeds a preset threshold, the risk node corresponding to the progress puzzle unit is located in the cross-project risk causal network as a disturbance source. The affected adjacent progress puzzle units are deduced step by step along the causal dependency edge. A secondary risk spread arc is generated based on the transmission probability and the puzzle adjacency relationship. The risk propagation direction and expected impact range of the secondary risk spread arc are evaluated. The risk control module is used to retrieve the most verified response strategy with the highest matching degree with the current risk propagation path from the historical response strategy library based on the risk propagation direction and expected impact range of the secondary risk spread arc zone. Then, based on the execution feedback of the verified response strategy, the module updates the cross-project transmission probability, adjusts the boundary constraints and time windows of the affected schedule puzzle units in the parallel schedule puzzle network, and recalculates the global stitching path of the project's parallel schedule puzzle network until the puzzle matching deviation of all schedule puzzle units converges to the tolerance range.

[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows: 1. This invention abstracts subtasks into progress puzzle units with technical interfaces, resource requirements, and time windows, and links them with cross-project risk causal networks. It can dynamically deduce the cascading propagation path and impact range of risks along the task dependency edges, realizing the transformation from passive response to proactive early warning, and effectively improving the foresight and accuracy of risk response in scientific research projects.

[0017] 2. After each risk intervention, this invention corrects the risk transmission probability and automatically adjusts the boundary constraints and time windows of the affected puzzle units. By iteratively recalculating the global puzzle path, the progress deviation is brought to within the tolerance range, forming a closed-loop adaptive learning and optimization capability, which effectively reduces the risk of project delays and failures. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a flowchart of the method of the present invention.

[0020] Figure 2 This is a system block diagram of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figure 1 As shown, a method for managing the entire lifecycle of a scientific research project includes the following steps: Step S1: Obtain the task decomposition structure of the project under research and the input-output dependencies between each subtask. Abstract each subtask into a schedule puzzle unit with boundary constraints. Extract the technical interface parameters, resource requirement attributes and time window labels of each schedule puzzle unit. Establish a puzzle adjacency matrix. Based on the puzzle adjacency matrix, stitch together each schedule puzzle unit to obtain the project parallel schedule puzzle network. Step S2: Preset risk event types to establish risk nodes. Construct project causal dependencies and transmission probabilities between risk nodes by corresponding historical project risk events for various risk events, and obtain a cross-project risk causal network. Associate and bind the cross-project risk causal network with the jigsaw puzzle adjacency matrix so that each progress jigsaw puzzle unit is loaded with the corresponding risk triggering conditions and transmission weights, and complete the risk perception initialization of the project parallel progress jigsaw puzzle network. Step S3: During the execution of the scientific research project, periodically collect the progress completion rate and risk trigger indicators of each progress puzzle unit, and quantify the deviation between the actual progress and the planned progress as the puzzle matching deviation degree; when the puzzle matching deviation degree of a puzzle unit exceeds the preset threshold, locate the risk node corresponding to the progress puzzle unit in the cross-project risk causal network as the disturbance source, and deduce the affected adjacent progress puzzle units step by step along the causal dependency edge. Generate the secondary risk spread arc area according to the transmission probability and the puzzle adjacency relationship, and evaluate the risk propagation direction and expected impact range of the secondary risk spread arc area. Step S4: Based on the risk propagation direction and expected impact range of the secondary risk spread arc, retrieve the verified response strategy with the highest matching degree with the current risk propagation path from the historical response strategy library. Then, update the cross-project transmission probability based on the execution feedback of the verified response strategy, adjust the boundary constraints and time windows of the affected schedule puzzle units in the parallel schedule puzzle network, and recalculate the global stitching path of the project's parallel schedule puzzle network until the puzzle matching deviation of all schedule puzzle units converges to the tolerance range.

[0023] Furthermore, step S1 is implemented through the following process: Step S101: Obtain the task decomposition structure of the project under development and the input-output dependencies between each subtask. Abstract each subtask into a progress puzzle unit with boundary constraints. The specific process includes: Obtain the project application, task contract or implementation plan of the target scientific research project, and parse out the task decomposition structure at the project level. The task decomposition structure breaks down the overall work of the project into several sub-tasks, such as literature review, theoretical modeling, algorithm design, software development, hardware-in-the-loop testing, field experiment, data analysis, report writing, etc. Each sub-task has a clear start and end time node, deliverable form and acceptance criteria. The task decomposition structure is analyzed in depth to identify the input-output dependencies between each subtask. The dependencies include, but are not limited to: sequence dependencies, parallel dependencies, data dependencies, and resource dependencies; Dependencies are recorded as directed edges, where the direction of the edge represents the direction from the preceding subtask to the subsequent subtask, and the weight of the edge can be assigned according to the tightness of the dependency. Each subtask in the task decomposition structure is abstracted as a schedule puzzle unit; the schedule puzzle unit represents the smallest indivisible logical block of project schedule management, and it encapsulates all the attributes and boundary constraints required to complete the subtask. The boundary constraints specifically include the following dimensions: Technical interface parameter boundaries: Define the technical specifications for the interaction between this progress puzzle unit and other progress puzzle units. For example, the input interface parameters of an "algorithm module development" unit include "data sampling frequency (must be between 100Hz and 1000Hz)" and "output result format (JSON or CSV)". Its output interface parameters include "algorithm processing latency (not exceeding 50ms)" and "computing resource consumption (CPU utilization rate less than 30%)". If any upstream or downstream puzzle unit cannot meet these interface parameter boundaries, the puzzle operation cannot be performed. Resource requirement attribute boundaries: Define the upper and lower limits of various resources required to complete this progress puzzle unit. Resource types include, but are not limited to, human resources, equipment resources, financial resources, and external collaboration resources. Resource requirement attribute boundaries are the basis for subsequent resource conflict detection and scheduling optimization. Time window label: Assign an executable time window to the progress puzzle unit. The time window consists of four parameters: earliest start time, latest start time, planned duration, and latest end time. The initial value of the time window label is directly derived from the planned time nodes in the task breakdown structure. For example, the earliest start time of a "field test" unit is the 90th day after the project starts, the latest start time is the 100th day, the planned duration is 5 days, and therefore its latest end time is the 105th day. This time window will be dynamically adjusted in subsequent steps as risk propagation and response strategies are implemented.

[0024] Step S102: Extract the technical interface parameters, resource requirement attributes, and time window labels of each progress mosaic unit, establish a mosaic adjacency matrix, and stitch the various progress mosaic units together based on the mosaic adjacency matrix to obtain the project parallel progress mosaic network. The specific process includes: Traverse all progress puzzle units, extract the attributes of each dimension to form an attribute vector set, and construct an N×N two-dimensional matrix with an initial value of 0, named the puzzle adjacency matrix, where N is the total number of progress puzzle units. The rows and columns of the puzzle adjacency matrix are arranged according to the progress puzzle unit numbers (P1, P2, ..., P...). N Indexing; For any two progress puzzle units P i and P j Based on the input-output dependency relationship extracted in step S101, determine whether there is a direct correlation between the two. If P i The output is P j If the input is given, then the element in the i-th row and j-th column of the puzzle adjacency matrix is ​​set to 1, indicating that there exists a path from P. i Point to P j A directed dependency edge, if P j The output is P i If the input is , then the element in the j-th row and i-th column is set to 1. If there is bidirectional data interaction, both directions are set to 1. If there is no dependency, the corresponding matrix element remains 0. Using each progress puzzle unit as a node and the directed edges in the puzzle adjacency matrix as connecting lines, an initial parallel project progress puzzle network is constructed. The parallel project progress puzzle network is a directed acyclic graph, and its topology reflects the strict temporal and logical order between all subtasks. During the construction of the parallel schedule puzzle network for the project: scan the puzzle adjacency matrix to identify all schedule puzzle unit pairs that do not have direct or indirect dependencies (i.e., no paths connect them), and mark the independent schedule puzzle unit pairs as units that can be executed in parallel. For example, the "theoretical modeling" unit and the "procure experimental consumables" unit usually do not have dependencies, so they can be executed in parallel as long as the time window allows.

[0025] Furthermore, step S2 is implemented through the following process: Step S201: Preset risk event types and establish risk nodes. Construct project causal dependencies and transmission probabilities between risk nodes by matching various risk events with historical project risk events, thus obtaining a cross-project risk causal network. The specific process includes: A standardized ontology of risk event types is established, covering all types of risks that may be encountered throughout the entire lifecycle of a research project. These risks are categorized hierarchically, with top-level categories including, but not limited to, user-specified categories: technical route obstruction, departure of key personnel, supply chain disruption, and deviation from budget execution. Each top-level category is further subdivided into second-level subcategories. For example, "technical route obstruction" can be further subdivided into "key algorithm non-convergence and experimental reproducibility failure" and "design performance indicators exceeding tolerance"; "supply chain disruption" can be further subdivided into "delayed delivery of customized chips," "out of stock of standard reagents," and "insufficient capacity of external processing plants." For each risk event (including top-level and second-level subcategories), a risk node is created. The risk node is a data structure whose attributes include: risk node ID (risk name, type, historical occurrence frequency (based on publicly available project archives of this unit or field, in times / hundred project years), inherent impact level (level 1-5, level 5 is catastrophic impact, such as directly causing project termination), and observable precursor indicators (for example, for the risk of "key algorithm not converging", its precursor indicators include "the number of iteration rounds exceeds twice the historical average" and "the loss function value does not decrease for 10 consecutive rounds", etc.). Retrieve historical project risk event records from the past 5-10 years within the target research institution or field. These records document the occurrence time, development process, response measures, and ultimate consequences of the risk events in the corresponding research projects. Perform causal chain analysis on each historical project risk event record. The causal chain analysis process, for example, discovers from the records that: "Supply chain disruption" (risk node A) occurred 2 weeks later, which led to "technical route obstruction" (risk node B) because the performance of the alternative chip did not meet the algorithm requirements. At the same time, "departure of key personnel" (risk node C) also directly led to "technical route obstruction" (risk node B) because the person was the only engineer who had mastered the core code. Based on the aforementioned historical causal chain, a directed risk causal graph is constructed. The nodes in the risk causal graph are the risk nodes defined in step S201. The directed edges in the graph represent that the occurrence of one risk node directly leads to the occurrence of another risk node. The propagation probability P is calculated for each directed edge. The formula for calculating the propagation probability P is: in This represents the total number of projects in which risk node A occurred in the historical project risk event record. The number of projects that, after risk node A occurs, subsequently experience risk node B within a certain time window (e.g., 3 months). Indicates risk node A The transmission probability between risk nodes B; integrating all risk nodes and their causal directed edges and transmission probabilities to obtain a cross-project risk causal network, which is used to reflect the cascading evolution law of various risk events under natural conditions without external intervention.

[0026] Step S202: Associate and bind the cross-project risk causal network with the puzzle adjacency matrix, so that each schedule puzzle unit is attached with the corresponding risk triggering conditions and transmission weights, completing the risk perception initialization of the project parallel schedule puzzle network. The specific process includes: Iterate through each schedule puzzle unit in the project's parallel schedule puzzle network and perform the following binding operations: First layer of binding: Risk triggering condition attachment, analysis progress puzzle unit P x The task content, technical interface parameters, and resource requirement attributes are used to filter out those related to P from the cross-project risk causal network. x The most relevant risk nodes, such as the "algorithm design" puzzle unit, are attached to the "key algorithm does not converge" sub-node in the "technical route obstruction" risk node. The trigger condition is set as follows: when the algorithm accuracy index in the technical interface parameters of the progress puzzle unit improves by less than 0.5% for three consecutive iterations, the risk is triggered. Each progress puzzle unit can be attached to one or more risk trigger conditions to form the risk feature vector of the unit. Second binding: Weight propagation, for weights mounted to P x Each risk node R on a Analyze R a In all outgoing edges (i.e., R) of the cross-project risk causal network a Other risk nodes that can be directly triggered), the transmission probability corresponding to the outgoing edge is weighted twice based on the dependence strength of Px in the puzzle adjacency matrix on other progress puzzle units to obtain the transmission weight. The specific formula is as follows: Where W is the number of puzzle units P x Risk node R on aTransmitted to puzzle unit P y Risk node R on b The weight, Indicates risk node R a Risk node R b The transmission probability between them P represents the adjacency matrix of the puzzle. x With P y The strength of the dependency between them (values ​​range from 0 to 1, or 0 if there is no dependency), where a and b are unequal natural numbers that are both greater than 0, and x and y are positive integers less than or equal to N that are unequal; The transmission weight indicates that only when there is a strong task dependency between two puzzle units can the risk on one progress puzzle unit be more easily transmitted to the other progress puzzle unit. If there is no dependency between two progress puzzle units, even if the risk could be transmitted in the cross-project risk causal network, it will not be transmitted in the actual project. Thus, after all bindings are completed, the risk perception initialization of the project parallel progress puzzle network is completed.

[0027] Furthermore, step S3 is implemented through the following process: Step S301: During the execution of the research project, periodically collect the progress completion rate and risk trigger indicators of each progress puzzle unit, and quantify the deviation between the actual progress and the planned progress as the puzzle matching deviation. The specific process includes: Set up a project health monitoring cycle. At the end of each project health monitoring cycle, automatically or manually collect the progress completion rate and risk trigger indicators of all active progress puzzle units: Schedule Completion Rate: Through the API interface of integrated project management software (such as Jira, MS Project), the actual completion percentage (0%-100%) of each schedule puzzle unit is read. Simultaneously, based on the planned duration and current elapsed time in the time window label of the schedule puzzle unit, its planned completion percentage is calculated using linear interpolation: assuming the time elapsed from the current time point to the earliest start time of the unit is t, and the total planned duration is T, then the planned completion percentage = (t / T) × 100%; Risk Trigger Indicators: The actual parameter values ​​corresponding to the risk trigger conditions attached to each schedule puzzle unit are collected. For example, monitoring the accuracy improvement value after each iteration of the "Algorithm Design" unit; monitoring the supplier delivery status and logistics tracking information of the "Procurement" unit; Based on the progress completion rate and risk trigger indicators, a puzzle matching deviation degree Y is calculated for each progress puzzle unit. The puzzle matching deviation degree Y represents the degree of deviation between the progress lag and the risk trigger of the corresponding progress puzzle unit. The calculation formula is as follows:

[0028] and These are preset weighting coefficients, all defaulting to 0.5. They can be adjusted according to project type (e.g., exploratory research focuses on risk, while engineering projects focus on schedule). It is the maximum value among the propagation weights of all triggered risk nodes on the progress puzzle unit that point to other progress puzzle units.

[0029] Step S302: When the puzzle matching deviation of a puzzle unit exceeds a preset threshold, locate the risk node corresponding to the schedule puzzle unit in the cross-project risk causal network as the disturbance source, and deduce the affected adjacent schedule puzzle units step by step along the causal dependency edge. The specific process includes: When the puzzle matching deviation Y of a progress puzzle unit exceeds the preset deviation threshold (e.g., 0.3, with a value range of 0-1, set by the project management based on the project's risk tolerance), the corresponding progress puzzle unit will be marked as a disturbance source unit; otherwise, it will not be marked. When the disturbance source unit is marked (e.g., progress puzzle unit P) k (k is a natural number greater than 0). First, in the cross-project risk causal network, locate the risk node with the highest trigger degree among the risk nodes attached to the schedule puzzle unit, and denote it as the starting risk node R. sta ; From R sta Starting from the beginning, along the directed edges of the cross-project risk causal network, and taking into account the dependencies in the jigsaw puzzle adjacency matrix, a restricted breadth-first risk deduction is performed. The deduction rules are as follows: From R sta Begin by examining the next risk node R that all its outgoing edges point to. next The propagation probability is P(R). next |R sta ); In the project parallel schedule jigsaw puzzle network, find all those related to P. k Adjacent progress puzzle units that have a direct dependency (i.e., the values ​​of elements in the puzzle adjacency matrix are non-zero) are denoted as P. nei ; Determine R next Is it mounted on P? nei If so, then the current transmission is considered valid, and a transmission from P is generated. k To P nei The risk transmission path, the effective strength of the transmission = the transmission probability P(R) next |R sta × In the adjacency matrix of the puzzle, P k With P nei The strength of dependence; If Rnext Not mounted on any site related to P k P with direct dependence nei Up, but R next Through multi-step causal chains (such as R) sta →R mid →R next It can reach a certain mount on P nei The risk node, and the intermediate risk node R mid If there is no corresponding progress puzzle unit, whether to perform virtual transmission depends on the specific project type. Under the default rule, virtual transmission is not performed. Repeat the above process to disturb the source unit P. k Centered on the core, the simulation radiates outwards layer by layer until any of the following termination conditions are met: 1. The simulation depth exceeds the preset 3 layers; 2. The effective intensity of risk transmission is less than the minimum transmission threshold (e.g., 0.05); 3. No new adjacent progress puzzle units on the radiation boundary can be affected. The positions of all schedule puzzle units covered by risk transmission paths in the project parallel schedule puzzle network are marked, and the subgraphs formed by the marked schedule puzzle units and the risk transmission paths between them are then extracted from the project parallel schedule puzzle network to separate the secondary risk spread arc area.

[0030] Step S303: Generate secondary risk spread arc zones based on transmission probability and puzzle adjacency relationships, and assess the risk propagation direction and expected impact range of the secondary risk spread arc zones. The specific process includes: For the generated secondary risk spread arc, the maximum path method is used to assess its propagation direction and impact range; Risk propagation direction assessment: Within the secondary risk spread arc region, calculate the cumulative effective strength of each risk propagation path. The cumulative effective strength is the product of the effective strengths of all edges on the risk propagation path. Traverse the risk propagation path with the largest cumulative effective strength, and the direction of this risk propagation path is the current highest priority risk propagation direction. The risk propagation direction points from the disturbance source unit to one or more end progress mosaic units at the edge of the arc region, thereby assessing the expected impact range of the secondary risk spread arc region. The expected scope of impact is quantified using the set of affected units and a comprehensive impact index: Affected Unit Set: The set of all schedule puzzle units within the secondary risk contagion arc; records the earliest point in time when the schedule puzzle unit is expected to be affected (estimated based on the length of the risk propagation path and the planned duration of each unit); Comprehensive Impact Index: For all affected schedule puzzle units within the secondary risk spread arc, the puzzle matching deviation is weighted and summed with the maximum effective intensity received on the corresponding schedule puzzle unit. The formula is: ; The magnitude of the comprehensive impact index represents the overall degree of harm in the secondary risk contagion arc zone. The higher the value, the more widespread and serious the current risk cascade has become to the project. The assessment results (risk propagation direction, list of affected units, expected earliest impact time, and comprehensive impact index) are packaged into a risk status report. After each project health monitoring cycle, steps S301 to S303 are re-executed to update the status of the secondary risk propagation arc zone until the puzzle matching deviation Y of the disturbance source unit falls below the threshold or the current research project is terminated.

[0031] Furthermore, step S4 is implemented through the following process: Step S401: Based on the risk propagation direction and expected impact range of the secondary risk spread arc, retrieve the verified response strategy with the highest matching degree to the current risk propagation path from the historical response strategy database. The specific process includes: Before a research project is launched, collect and structure-store historical project records of risk response for typical research projects within the unit and field. Each historical project record represents a validated response strategy, including: Risk propagation pattern: An abstract representation describing the risk propagation path that occurred at that time, in the format of [risk node sequence] > [affected puzzle unit type sequence]; Response strategy description: A detailed action plan document, including the person in charge, execution steps, and required resources; Execution feedback results: After the strategy is implemented, quantitative indicators such as the percentage reduction in puzzle matching deviation, the updated value of risk transmission probability, and the number of days of project delay are included. After step S303 generates a new secondary risk propagation arc, the risk propagation path (risk node sequence) and the type of the affected progress puzzle unit are extracted, and the matching degree is calculated with the risk propagation pattern of the verified response strategy. The matching degree calculation uses a combination of edit distance and semantic similarity. First, the risk node sequence in the current risk propagation path is compared with the [risk node sequence] using string edit distance calculation; the smaller the distance, the more similar the structure. Second, ontology semantic matching is performed on the types of affected puzzle units. For example, the "algorithm design" unit and the "model training" unit are relatively close in the ontology. The final matching degree = 0.6 × (1 - normalized edit distance) + 0.4 × semantic similarity; The top three verified response strategies with the highest matching degree are selected as the candidate response strategy set. If the highest matching degree is lower than the preset threshold (e.g., 0.6), the project manager will manually intervene to formulate a new verified response strategy.

[0032] Step S402: Update the cross-project propagation probability based on the execution feedback of the verified response strategy, adjust the boundary constraints and time windows of the affected schedule puzzle units in the parallel schedule puzzle network, and recalculate the global stitching path of the project's parallel schedule puzzle network until the puzzle matching deviation of all schedule puzzle units converges to the tolerance range. The specific process includes: Select the verified response strategy with the highest matching degree from the candidate response strategy set and execute it (if the matching degree is equal, one is randomly selected). During and after execution, continuously collect execution feedback data. The execution feedback data is the number of times risk transmission actually occurred and the number of times transmission did not occur within the original secondary risk spread arc area. Dynamically adjust the relevant transmission probabilities in the cross-project risk causal network for each pair of risk nodes (R) on the current risk propagation path. a →R b ), statistics show that during the implementation of the verified response strategy, R a After R occurred b The updated posterior propagation probability is the number of actual conditions that occur within the preset time window: , Where λ is the Laplace smoothing parameter (usually taken as 1) to prevent zero probability problems, and V is... The number of possible outcomes; Based on the selected response strategy and the updated propagation probability, boundary constraints and time windows are adjusted for the project parallel schedule puzzle network. The adjustment targets all affected schedule puzzle units located within the secondary risk spread arc. The boundary constraints include resource requirement attribute boundaries and technical interface parameter boundaries. The time window adjustment recalculates the earliest start time, latest start time, duration, and latest end time of the affected units based on the expected impact range of the secondary risk spread arc zone, thereby increasing relaxation time for severely affected progress puzzle units. After completing the boundary constraints and time window adjustments, scan all affected progress puzzle units and re-evaluate the puzzle matching deviation (based on the newly collected actual progress and the adjusted new plan) to determine whether the deviation of all progress puzzle units has been reduced to the preset tolerance range (e.g., deviation ≤ 0.1). If so, the risk intervention was successful, the project has returned to a healthy state, and the next monitoring cycle will continue. If not, it is determined that the current verified response strategy is insufficient, or that the adjusted project parallel schedule puzzle network has generated new disturbance sources. Then, return to step S301 and start monitoring again in the new project parallel schedule puzzle network state, entering a new round of closed-loop iterative optimization process. This iterative process will continue until the puzzle matching deviation of all schedule puzzle units converges to the tolerance range, or the project management decides to terminate the project.

[0033] Please see Figure 2 As shown, a scientific research project lifecycle management system includes a project risk management module, an ongoing project dynamic monitoring module, and a risk control module. The project risk management module is used to obtain the task decomposition structure of the project under development and the input-output dependencies between each subtask. Each subtask is abstracted into a schedule puzzle unit with boundary constraints. The technical interface parameters, resource requirement attributes and time window labels of each schedule puzzle unit are extracted to establish a puzzle adjacency matrix. Based on the puzzle adjacency matrix, each schedule puzzle unit is pieced together to obtain a parallel schedule puzzle network for the project. Risk nodes are established by pre-setting risk event types. The project causal dependencies and transmission probabilities between risk nodes are constructed by corresponding historical project risk events for various risk events to obtain a cross-project risk causal network. The cross-project risk causal network is associated and bound to the puzzle adjacency matrix so that each schedule puzzle unit is equipped with corresponding risk triggering conditions and transmission weights. The dynamic monitoring module for ongoing projects is used to periodically collect the progress completion rate and risk trigger indicators of each progress puzzle unit during the execution of scientific research projects. The deviation between the actual progress and the planned progress is quantified as the puzzle matching deviation. When the puzzle matching deviation of a puzzle unit exceeds a preset threshold, the risk node corresponding to the progress puzzle unit is located in the cross-project risk causal network as a disturbance source. The affected adjacent progress puzzle units are deduced step by step along the causal dependency edge. A secondary risk spread arc is generated based on the transmission probability and the puzzle adjacency relationship. The risk propagation direction and expected impact range of the secondary risk spread arc are evaluated. The risk control module is used to retrieve the most verified response strategy with the highest matching degree with the current risk propagation path from the historical response strategy library based on the risk propagation direction and expected impact range of the secondary risk spread arc zone. Then, based on the execution feedback of the verified response strategy, the module updates the cross-project transmission probability, adjusts the boundary constraints and time windows of the affected schedule puzzle units in the parallel schedule puzzle network, and recalculates the global stitching path of the project's parallel schedule puzzle network until the puzzle matching deviation of all schedule puzzle units converges to the tolerance range.

[0034] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for managing the entire lifecycle of scientific research projects, characterized in that, Includes the following steps: Step S1: Obtain the task decomposition structure of the project under research and the input-output dependencies between each subtask. Abstract each subtask into a schedule puzzle unit with boundary constraints. Extract the technical interface parameters, resource requirement attributes and time window labels of each schedule puzzle unit. Establish a puzzle adjacency matrix. Based on the puzzle adjacency matrix, stitch together each schedule puzzle unit to obtain the project parallel schedule puzzle network. Step S2: Preset risk event types to establish risk nodes. Construct project causal dependencies and transmission probabilities between risk nodes by corresponding historical project risk events for various risk events, and obtain a cross-project risk causal network. Associate and bind the cross-project risk causal network with the jigsaw puzzle adjacency matrix so that each progress jigsaw puzzle unit is loaded with the corresponding risk triggering conditions and transmission weights, and complete the risk perception initialization of the project parallel progress jigsaw puzzle network. Step S3: During the execution of the scientific research project, periodically collect the progress completion rate and risk trigger indicators of each progress puzzle unit, and quantify the deviation between the actual progress and the planned progress as the puzzle matching deviation degree; when the puzzle matching deviation degree of a puzzle unit exceeds the preset threshold, locate the risk node corresponding to the progress puzzle unit in the cross-project risk causal network as the disturbance source, and deduce the affected adjacent progress puzzle units step by step along the causal dependency edge. Generate the secondary risk spread arc area according to the transmission probability and the puzzle adjacency relationship, and evaluate the risk propagation direction and expected impact range of the secondary risk spread arc area. Step S4: Based on the risk propagation direction and expected impact range of the secondary risk spread arc, retrieve the verified response strategy with the highest matching degree with the current risk propagation path from the historical response strategy library. Then, update the cross-project transmission probability based on the execution feedback of the verified response strategy, adjust the boundary constraints and time windows of the affected schedule puzzle units in the parallel schedule puzzle network, and recalculate the global stitching path of the project's parallel schedule puzzle network until the puzzle matching deviation of all schedule puzzle units converges to the tolerance range.

2. The method for managing the entire lifecycle of scientific research projects according to claim 1, characterized in that, The process of obtaining the task decomposition structure of the ongoing project and the input-output dependencies between each subtask, and abstracting each subtask into a schedule puzzle unit with boundary constraints, includes: Obtain the project proposal, task contract or implementation plan of the target scientific research project, and parse out the task decomposition structure of the project level. The task decomposition structure breaks down the overall work of the project into several sub-tasks. The task decomposition structure is analyzed in depth to identify the input-output dependencies between subtasks. These dependencies are recorded as directed edges, where the direction of the edge represents the direction from the preceding subtask to the subsequent subtask, and the weight of the edge can be assigned according to the tightness of the dependency. For each subtask in the task decomposition structure, it is abstracted into a progress puzzle unit; the progress puzzle unit represents the smallest indivisible logical block of project progress management, which encapsulates all the attributes and boundary constraints required to complete the subtask. The boundary constraints specifically include technical interface parameter boundaries, resource requirement attribute boundaries, and time window labels.

3. The method for managing the entire lifecycle of scientific research projects according to claim 2, characterized in that, The process of extracting the technical interface parameters, resource requirement attributes, and time window labels of each schedule puzzle unit, establishing a puzzle adjacency matrix, and then piecing together the various schedule puzzle units based on the puzzle adjacency matrix to obtain the project parallel schedule puzzle network includes: Traverse all progress puzzle units, extract the attributes of each dimension to form an attribute vector set, named the puzzle adjacency matrix, and construct the initial project parallel progress puzzle network with each progress puzzle unit as a node and the directed edges in the puzzle adjacency matrix as connecting lines. During the construction of the parallel schedule puzzle network for the project: scan the puzzle adjacency matrix to identify all schedule puzzle unit pairs that do not have direct or indirect dependencies, and mark the independent schedule puzzle unit pairs as parallel execution units.

4. The method for managing the entire lifecycle of scientific research projects according to claim 3, characterized in that, The process of establishing risk nodes by pre-setting risk event types, and constructing project causal dependencies and transmission probabilities between risk nodes by corresponding historical project risk events for various risk events, to obtain a cross-project risk causal network includes: Establish a standardized ontology of risk event types, which covers all types of risks that may be encountered throughout the entire life cycle of a research project and classifies them by level. For each type of risk event, a risk node is created, and historical project risk event records of the target research institution or field are retrieved. These historical project risk event records record the occurrence time, development process, countermeasures, and final consequences of the risk events in the corresponding research projects. Causal chain analysis is performed on each historical project risk event record. Based on the above historical causal chain, a risk causal graph is constructed. The directed edges in the graph represent that the occurrence of one risk node will directly lead to the occurrence of another risk node. The propagation probability is calculated for each directed edge. All risk nodes, their causal directed edges, and propagation probabilities are integrated to obtain a cross-project risk causal network.

5. The method for managing the entire life cycle of scientific research projects according to claim 4, characterized in that, The process of associating and binding the cross-project risk causal network with the jigsaw puzzle adjacency matrix, so that each schedule jigsaw puzzle unit is attached with the corresponding risk triggering conditions and propagation weights, and completing the risk perception initialization of the project parallel schedule jigsaw puzzle network includes: Traverse each schedule puzzle unit in the project's parallel schedule puzzle network and perform the following binding operations, including attaching risk trigger conditions and attaching propagation weights; The transmission weight indicates that only when there is a strong task dependency between two puzzle units can the risk on one progress puzzle unit be more easily transmitted to the other progress puzzle unit. If there is no dependency between two progress puzzle units, even if the risk could be transmitted in the cross-project risk causal network, it will not be transmitted in the actual project. Thus, after all bindings are completed, the risk perception initialization of the project parallel progress puzzle network is completed.

6. The method for managing the entire life cycle of scientific research projects according to claim 5, characterized in that, The process of periodically collecting progress completion rates and risk trigger indicators for each progress puzzle unit during the execution of scientific research projects, and quantifying the deviation between actual progress and planned progress as puzzle matching deviation includes: Set up a project health monitoring cycle, and automatically or manually collect the progress completion rate and risk trigger indicators of all active progress puzzle units when each project health monitoring cycle arrives. The puzzle matching deviation is calculated for each progress puzzle unit based on the progress completion rate and risk trigger indicators. The puzzle matching deviation represents the degree of deviation between the progress lag and the risk trigger of the corresponding progress puzzle unit.

7. A method for managing the entire lifecycle of scientific research projects according to claim 6, characterized in that, When the mismatch of a puzzle piece exceeds a preset threshold, the process of locating the risk node corresponding to the schedule puzzle piece in the cross-project risk causal network as the disturbance source, and deducing the affected adjacent schedule puzzle pieces step by step along the causal dependency edge includes: When the matching deviation of a progress puzzle unit exceeds a preset deviation threshold, the corresponding progress puzzle unit is marked as a disturbance source unit; otherwise, it is not marked. In the cross-project risk causal network, the risk node with the highest trigger degree among the risk nodes attached to the schedule puzzle unit is located and denoted as the starting risk node. Starting from the starting risk node, along the directed edges of the cross-project risk causal network, and taking into account the dependencies in the puzzle adjacency matrix, a restricted breadth-first risk deduction is performed. Based on the risk deduction results, the risk radiates outward layer by layer with the disturbance source unit as the center. Mark the position of all schedule puzzle units covered by risk transmission paths in the project parallel schedule puzzle network, and then extract the secondary risk spread arc area from the project parallel schedule puzzle network by forming a sub-graph composed of the marked schedule puzzle units and the risk transmission paths between them. Within the secondary risk spread arc zone, the cumulative effective intensity of each risk transmission path is obtained, and the risk transmission path with the largest cumulative effective intensity is traversed to assess the expected impact range of the secondary risk spread arc zone. The expected impact range is quantified by the set of affected units and the comprehensive impact index.

8. A method for managing the entire lifecycle of scientific research projects according to claim 7, characterized in that, The process of retrieving the most validated response strategy from the historical response strategy database, based on the risk propagation direction and expected impact range of the secondary risk contagion arc, includes: Before a research project is launched, collect and structure the historical project records of risk response for typical research projects in the unit and field. Each historical project record is a verified response strategy, which includes the risk propagation model, response strategy description and implementation feedback effect. Once a secondary risk spread arc zone is generated, the risk propagation path and the type of the affected progress puzzle unit are extracted, and the matching degree is calculated with the risk propagation pattern of the verified response strategy. The top three verified response strategies with the highest matching degree are selected as the candidate response strategy set. If the highest matching degree is lower than the preset threshold, the project manager will manually intervene to formulate a new verified response strategy.

9. A method for managing the entire lifecycle of scientific research projects according to claim 8, characterized in that, The process of updating the cross-project propagation probability based on the execution feedback of the validated response strategy, adjusting the boundary constraints and time windows of the affected schedule puzzle units in the parallel schedule puzzle network, and recalculating the global stitching path of the project's parallel schedule puzzle network until the puzzle matching deviation of all schedule puzzle units converges to the tolerance range includes: The verified response strategy with the highest matching degree is selected from the candidate response strategy set and executed. During and after execution, the execution feedback data is continuously collected. The execution feedback data is the number of times risk transmission actually occurred and the number of times transmission did not occur within the original secondary risk spread arc area. The relevant transmission probabilities in the cross-project risk causal network are dynamically corrected. Based on the selected response strategy and the updated transmission probabilities, the boundary constraints and time windows of the project parallel schedule puzzle network are adjusted. The adjustment targets all affected schedule puzzle units located in the secondary risk spread arc zone. After completing the boundary constraints and time window adjustments, scan all affected schedule puzzle units, re-evaluate the puzzle matching deviation, and determine whether the deviation of all schedule puzzle units has been reduced to the preset tolerance range. Based on the judgment result, determine whether the current verified response strategy is insufficient or whether the adjusted project parallel schedule puzzle network has generated new disturbance sources. Restart monitoring under the new project parallel schedule puzzle network state.

10. A research project lifecycle management system, used to implement the research project lifecycle management method according to any one of claims 1-9, characterized in that, It includes a project risk management module, a dynamic monitoring module for projects under development, and a risk control module; The project risk management module is used to obtain the task decomposition structure of the project under development and the input-output dependencies between each subtask. Each subtask is abstracted into a schedule puzzle unit with boundary constraints. The technical interface parameters, resource requirement attributes and time window labels of each schedule puzzle unit are extracted to establish a puzzle adjacency matrix. Based on the puzzle adjacency matrix, each schedule puzzle unit is pieced together to obtain a parallel schedule puzzle network for the project. Risk nodes are established by pre-setting risk event types. The project causal dependencies and transmission probabilities between risk nodes are constructed by corresponding historical project risk events for various risk events to obtain a cross-project risk causal network. The cross-project risk causal network is associated and bound to the puzzle adjacency matrix so that each schedule puzzle unit is equipped with corresponding risk triggering conditions and transmission weights. The dynamic monitoring module for ongoing projects is used to periodically collect the progress completion rate and risk trigger indicators of each progress puzzle unit during the execution of scientific research projects. The deviation between the actual progress and the planned progress is quantified as the puzzle matching deviation. When the puzzle matching deviation of a puzzle unit exceeds a preset threshold, the risk node corresponding to the progress puzzle unit is located in the cross-project risk causal network as a disturbance source. The affected adjacent progress puzzle units are deduced step by step along the causal dependency edge. A secondary risk spread arc is generated based on the transmission probability and the puzzle adjacency relationship. The risk propagation direction and expected impact range of the secondary risk spread arc are evaluated. The risk control module is used to retrieve the most verified response strategy with the highest matching degree with the current risk propagation path from the historical response strategy library based on the risk propagation direction and expected impact range of the secondary risk spread arc zone. Then, based on the execution feedback of the verified response strategy, the module updates the cross-project transmission probability, adjusts the boundary constraints and time windows of the affected schedule puzzle units in the parallel schedule puzzle network, and recalculates the global stitching path of the project's parallel schedule puzzle network until the puzzle matching deviation of all schedule puzzle units converges to the tolerance range.

Images