Construction interruption risk reasoning and plan generation method and device

Abstract

The invention provides a construction interruption risk reasoning and plan generation method and device. The method comprises the following steps: acquiring multiple types of risk factors influencing construction interruption and a historical data set; taking a Bayesian information criterion as a scoring function, and constructing a directed acyclic graph structure based on the scoring function, a historical data set and a constraint set determined by domain knowledge; the constraint set is used for limiting whether edges between nodes exist or not in the construction process of the directed acyclic graph structure; based on the directed acyclic graph structure and the historical data set, a conditional probability table of the directed acyclic graph structure is determined, and the directed acyclic graph structure and the conditional probability table form a Bayesian network model; determining the risk probability of the target construction interruption event based on a Bayesian network model; and optimizing a preset construction plan combination based on the risk probability, and determining an optimal plan combination. According to the directed acyclic graph structure, objective data and domain knowledge are considered, and the accuracy of subsequent risk reasoning and plan generation is improved.

Description

Technical Field

[0001] This invention relates to the field of construction management technology, specifically to a method and apparatus for construction interruption risk reasoning and contingency plan generation. Background Technology

[0002] During the construction of large-scale engineering projects, due to the complex environment, numerous participants, and close coordination of processes, various risk factors (such as severe weather, equipment failure, supply chain delays, and management oversights) intertwine and evolve dynamically, easily leading to construction interruptions, causing schedule delays, cost overruns, and even safety accidents. Therefore, conducting forward-looking and precise reasoning and simulation of construction interruption risks, and generating scientific and operable emergency plans based on this, has become a core requirement for the refined and intelligent management of modern engineering projects.

[0003] Currently, Bayesian network-based risk assessment is the mainstream technology in this field. However, existing technologies often take two disparate paths when constructing Bayesian network structures: one is a purely data-driven method that relies entirely on historical data (such as scoring search algorithms), which is easily affected by data noise and insufficient sample size, resulting in false associations that violate common sense in engineering; the other is a qualitative description that relies entirely on expert experience, which is highly subjective, makes it difficult to handle the implicit cross-influences between multiple factors in complex projects, and results in significant differences in the structures constructed by different experts. In other words, existing Bayesian network structures either ignore the explicit causal rules in domain knowledge or fail to utilize data to correct the limitations of expert experience, resulting in network structures that cannot balance objectivity and engineering practicality.

[0004] Therefore, there is an urgent need to provide a method and device for construction interruption risk reasoning and contingency plan generation, which can organically integrate data-driven objective scoring with the subjective constraints of expert knowledge, taking into account both objectivity and engineering applicability, and improving the accuracy of risk reasoning and contingency plan generation. Summary of the Invention

[0005] In view of this, it is necessary to provide a method and apparatus for construction interruption risk reasoning and contingency plan generation, in order to solve the technical problem that the existing Bayesian network structure does not establish a hybrid learning mechanism of data-driven and expert-constrained learning, which leads to the network structure being unable to balance objectivity and engineering applicability, and thus resulting in low accuracy of risk reasoning and contingency plan generation.

[0006] To address the aforementioned technical problems, in a first aspect, the present invention provides a method for construction interruption risk reasoning and contingency plan generation, comprising: Obtain various risk factors and historical datasets that affect construction interruption, and use these risk factors as nodes in a Bayesian network; Using Bayesian information criteria as a scoring function, a directed acyclic graph structure is constructed based on the scoring function, the historical dataset, and a constraint set determined by domain knowledge; the constraint set is used to restrict the existence of edges between nodes during the construction of the directed acyclic graph structure. Based on the directed acyclic graph structure and historical dataset, a conditional probability table for the directed acyclic graph structure is determined, and the directed acyclic graph structure and the conditional probability table constitute a Bayesian network model. The risk probability of the target construction interruption event is determined based on the Bayesian network model. Based on the aforementioned risk probabilities, the preset combination of construction plans is optimized to determine the optimal combination of plans.

[0007] In one possible implementation, constructing the directed acyclic graph structure based on the scoring function, the historical dataset, and the constraint set determined by domain knowledge includes: Step 1: Construct the initial directed acyclic graph structure; Step 2: Adjust the edges of the initial directed acyclic graph structure to obtain multiple candidate directed acyclic graph structures; Step 3: Based on the constraint set, modify the multiple candidate directed acyclic graph structures to obtain multiple valid directed acyclic graph structures; Step 4: Determine the scoring function value for each of the effective directed acyclic graph structures based on the scoring function and the historical dataset; Step 5: Take the effective directed acyclic graph structure with the highest scoring function value as the initial directed acyclic graph structure, and return to Step 2. Repeat Step 2-Step 5 until the change in the scoring function value of the effective directed acyclic graph structure is less than a preset value, and take the effective directed acyclic graph structure as the directed acyclic graph structure.

[0008] In one possible implementation, the scoring function is:

[0009] In the formula, D represents the score function value; F represents the historical dataset; and F represents the directed acyclic graph structure. The log-likelihood function represents the degree of data fit. d represents the parameter set under maximum likelihood estimation; d represents the number of independent parameters in the directed acyclic graph structure; and N represents the amount of data in the historical dataset.

[0010] In one possible implementation, the constraint set includes a set of mandatory edges and a set of prohibited edges. Edges in the set of mandatory edges must exist in the directed acyclic graph structure, while edges in the set of prohibited edges must not exist in the directed acyclic graph structure.

[0011] In one possible implementation, determining the conditional probability table of the directed acyclic graph structure based on the directed acyclic graph structure and the historical dataset includes: Based on the historical dataset, determine the conditional probability of each node in the directed acyclic graph structure; The virtual conditional probabilities of each node in the directed acyclic graph structure are determined based on expert knowledge. The conditional probabilities of the data and the virtual conditional probabilities are fused to obtain the conditional probabilities of each node, and the conditional probabilities of all nodes constitute the conditional probability table.

[0012] In one possible implementation, the conditional probabilities of each node are:

[0013]

[0014]

[0015] In the formula, Given that the parent node is in state j, the conditional probability that the i-th node is in state k. Given that the parent node is in state j in the historical dataset, the number of nodes with state k at the i-th node. This represents the total number of samples in the historical dataset where the parent node's state is j. Given that the parent node is in state j in the expert knowledge, the confidence level is given that the state of the i-th node is k. This represents the total confidence level of the parent node with state j in the expert knowledge.

[0016] In one possible implementation, determining the risk probability of the target construction interruption event based on the Bayesian network model includes: Construct a joint tree corresponding to the directed acyclic graph structure; the joint tree contains multiple connected clique nodes; The potential function of each clique node is determined according to the conditional probability table; The target construction interruption event is allocated as evidence information in the joint tree, and the evidence information is transmitted according to the connection relationship between the nodes of each group; The potential function of each cluster node is updated based on the evidence information, and the risk probability of the target construction interruption event is determined based on the updated potential function.

[0017] In one possible implementation, optimizing the preset combination of construction plans based on the risk probability to determine the optimal combination of plans includes: The objective function is constructed with the aim of maximizing the net return of the proposed plan combination; With the constraint that the total cost of the proposed plan combination is less than or equal to the maximum budget, the objective function is solved based on a genetic algorithm to obtain the optimal proposed plan combination.

[0018] In one possible implementation, the objective function is:

[0019] The constraints are as follows:

[0020] In the formula, Construction interruption event for target Baseline probability of occurrence when no contingency plan is in place; Construction interruption event for target The probability of occurrence after adopting contingency plan combination X; Construction interruption event for target Potential economic losses upon occurrence; This is the i-th contingency plan; Let m be the cost of the i-th contingency plan; m is the total number of combined contingency plans. For the maximum budget.

[0021] Secondly, the present invention also provides a construction interruption risk reasoning and contingency plan generation device, comprising: The data acquisition unit is used to acquire multiple risk factors and historical datasets that affect construction interruption, and to use the risk factors as nodes in a Bayesian network. A directed acyclic graph (DAG) structure building unit is used to construct a DAG structure based on the Bayesian information criterion as a scoring function, the historical dataset, and a constraint set determined by domain knowledge; the constraint set is used to restrict the existence of edges between nodes during the construction of the DAG structure. A Bayesian network model building unit is used to determine the conditional probability table of the directed acyclic graph structure based on the directed acyclic graph structure and the historical dataset. The directed acyclic graph structure and the conditional probability table constitute a Bayesian network model. The risk probability determination unit is used to determine the risk probability of the target construction interruption event based on the Bayesian network model. The contingency plan combination determination unit is used to optimize the preset construction contingency plan combination based on the risk probability and determine the optimal contingency plan combination.

[0022] The beneficial effects of this invention are as follows: The construction interruption risk reasoning and contingency plan generation method provided by this invention considers that when constructing a directed acyclic graph structure based solely on a purely data-driven scoring function, the problem arises because pure data-driven approaches maximize the interpretation of historical data, potentially leading to false associations or omissions of potential but unrealized causal relationships. By constructing a directed acyclic graph structure based on a scoring function and a constraint set determined by domain knowledge, the method forces the absence of edges with false associations and ensures the existence of edges representing potential but unrealized relationships. In other words, invalid edges are eliminated in advance based on expert constraints, and edges that are guaranteed to exist are forced to guide the search direction. This ensures that the constructed directed acyclic graph structure incorporates both objective data and domain knowledge, guaranteeing its accuracy and thus improving the accuracy of the risk reasoning and contingency plan generation based on the directed acyclic graph structure. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A schematic flowchart of an embodiment of the construction interruption risk reasoning and contingency plan generation method provided by the present invention; Figure 2 A schematic diagram of an embodiment of the constraint set provided by the present invention; Figure 3 For the present invention Figure 1 A schematic diagram of an embodiment of S102; Figure 4 For the present invention Figure 1 A schematic diagram of an embodiment of S103; Figure 5 For the present invention Figure 1 A schematic diagram of an embodiment of S104; Figure 6 For the present invention Figure 1 A schematic diagram of an embodiment of S105; Figure 7 This is a schematic flowchart of an embodiment of the present invention for solving the objective function based on a genetic algorithm; Figure 8 This is a schematic diagram of an embodiment of the construction interruption risk reasoning and contingency plan generation device provided by the present invention. Detailed Implementation

[0025] 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 a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0026] It should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this invention illustrate operations implemented according to some embodiments of the invention. It should be understood that the operations in the flowcharts may be implemented out of order, and steps without logical contextual relationships may be reversed or performed simultaneously. Furthermore, those skilled in the art, guided by the content of this invention, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0027] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0028] This invention provides a method and apparatus for construction interruption risk reasoning and contingency plan generation, which will be described below.

[0029] Figure 1 This is a schematic flowchart of an embodiment of the construction interruption risk reasoning and contingency plan generation method provided by the present invention, as shown below. Figure 1 As shown, the methods for construction interruption risk reasoning and contingency plan generation include: S101. Obtain various risk factors and historical datasets that affect construction interruption, and use these risk factors as nodes in a Bayesian network.

[0030] Risk factors can be identified through methods such as expert interviews and historical data analysis. Risk factors can be qualitative, such as equipment failure or severe weather, or they can be quantitative variables that have been discretized, such as the degree of delay.

[0031] Historical datasets can be determined by collecting historical project data and expert experience data. To ensure the data meets the requirements of Bayesian modeling, continuous data is discretized, and missing values ​​are imputed and outliers are removed.

[0032] S102. Using the Bayesian information criterion as the scoring function, construct a directed acyclic graph structure based on the scoring function, historical dataset, and constraint set determined by domain knowledge; the constraint set is used to restrict the existence of edges between nodes during the construction of the directed acyclic graph structure.

[0033] In a specific embodiment of the present invention, the constraint set includes a set of mandatory edges and a set of prohibited edges. Edges in the set of mandatory edges must exist in the directed acyclic graph structure, while edges in the set of prohibited edges must not exist in the directed acyclic graph structure.

[0034] In specific embodiments of the present invention, such as Figure 2 As shown, risk factors are divided into three categories: environmental factors, resource factors, and management factors. Environmental factors include rainstorms and typhoons, resource factors include equipment failures, and management factors include design changes and personnel scheduling errors. These risk factors can lead to transportation disruptions or work process stoppages, which in turn can cause construction interruptions.

[0035] by Figure 2 For example, the edges between heavy rain and transportation interruption, process stagnation, typhoon and transportation interruption, process stagnation, equipment failure and transportation interruption, process stagnation, design change and process stagnation, and personnel scheduling error and process stagnation are the set of forced edges, while the edges between design change, personnel scheduling error and transportation interruption are the set of prohibited edges.

[0036] S103. Based on the directed acyclic graph structure and historical dataset, determine the conditional probability table of the directed acyclic graph structure. The directed acyclic graph structure and the conditional probability table form a Bayesian network model. S104. Determine the risk probability of the target construction interruption event based on the Bayesian network model; S105. Optimize the preset combination of construction plans based on risk probability to determine the optimal combination of plans.

[0037] It should be understood that the construction interruption risk reasoning and contingency plan generation method in this embodiment of the invention can be implemented in any device based on the construction interruption risk reasoning and contingency plan generation method, such as electronic devices such as decision-making devices or management devices based on risk probabilities and contingency plans. Specifically, the construction interruption risk reasoning and contingency plan generation method is stored in the aforementioned device as a pre-programmed program. When the device is started, the program is invoked, and the construction interruption risk reasoning and contingency plan generation method is implemented.

[0038] Compared with existing technologies, the construction interruption risk reasoning and contingency plan generation method provided in this invention addresses the issue that constructing a directed acyclic graph (DAG) structure based solely on a data-driven scoring function can lead to false associations or omissions of potential but unrealized causal relationships, as data-driven approaches maximize the interpretation of historical data. This is achieved by constructing a DAG structure based on a scoring function and a constraint set determined by domain knowledge. This forces the absence of false associations and ensures the existence of edges representing potential but unrealized relationships. In other words, invalid edges are preemptively eliminated based on expert constraints, while edges that are guaranteed to exist are forced to guide the search direction. This ensures that the constructed DAG structure incorporates both objective data and domain knowledge, guaranteeing its accuracy and thus improving the accuracy of the risk reasoning and contingency plan generation based on the DAG structure.

[0039] In some embodiments of the present invention, the construction process of the directed acyclic graph structure is based on a hill-climbing search strategy, specifically, as follows: Figure 3 As shown, step S102 includes: S301. Construct the initial directed acyclic graph structure.

[0040] The initial directed acyclic graph structure can be a random network structure that conforms to the set of constraints, or it can be an empty network structure.

[0041] S302. Adjust the edges of the initial directed acyclic graph structure to obtain multiple candidate directed acyclic graph structures.

[0042] Specifically, the methods for adjusting edges include, but are not limited to, adding edges, subtracting edges, and rotating edges.

[0043] S303. Based on the constraint set, multiple candidate directed acyclic graph structures are modified to obtain multiple effective directed acyclic graph structures.

[0044] Specifically, edges in the forbidden edge set are removed, and edges in the mandatory edge set are added.

[0045] S304. Determine the score function value for each valid directed acyclic graph structure based on the score function and historical dataset; S305. Take the effective directed acyclic graph structure with the highest scoring function value as the initial directed acyclic graph structure, and return to S302. Repeat S302-S305 until the change in the scoring function value of the effective directed acyclic graph structure is less than the preset value, and take the effective directed acyclic graph structure as the directed acyclic graph structure.

[0046] It should be noted that: the change in the score function value of an effective directed acyclic graph structure being less than the preset value means that during a preset number of iterations, the increase in the score function value is less than the preset increase. In other words, when no operation can further improve the score function value, the search stops, and a directed acyclic graph structure is obtained.

[0047] The preset number of times and preset increment can be set or adjusted according to the actual application scenario, which will not be elaborated here.

[0048] In a specific embodiment of the present invention, the scoring function is:

[0049] In the formula, D represents the score function value; F represents the historical dataset; and F represents the directed acyclic graph structure. The log-likelihood function represents the degree of data fit. d represents the parameter set under maximum likelihood estimation; d represents the number of independent parameters in the directed acyclic graph structure; and N represents the amount of data in the historical dataset.

[0050] Existing technologies, when determining conditional probability tables for directed acyclic graph structures, rely on maximum likelihood estimation based on observed data. However, these methods fail to consider the problems of small sample sizes and high missing rates in construction risk data. For example, when data is missing, maximum likelihood estimation cannot obtain conditional probabilities; when data is extremely limited, overfitting due to isolated random samples can occur, resulting in instability. To address these technical problems, in some embodiments of this invention, such as... Figure 4 As shown, step S103 includes: S401. Determine the conditional probability of each node in a directed acyclic graph structure based on historical datasets. S402. Determine the virtual conditional probability of each node in a directed acyclic graph structure based on expert knowledge. S403. The conditional probabilities of the data and the virtual conditional probabilities are fused to obtain the conditional probabilities of each node. The conditional probabilities of all nodes constitute a conditional probability table.

[0051] In view of the problem that the maximum likelihood method relying solely on data has poor accuracy in zero-sample and few-sample scenarios, this embodiment of the invention adds a virtual conditional probability determined based on expert knowledge. This allows the experience-driven approach to compensate for the shortcomings of the data-driven approach, and the virtual conditional probability to compensate for the shortcomings in zero-sample and few-sample scenarios, thereby improving the accuracy of conditional probability estimation in zero-sample and few-sample scenarios.

[0052] In a specific embodiment of the present invention, the conditional probability of each node is:

[0053]

[0054]

[0055] In the formula, Given that the parent node is in state j, the conditional probability that the i-th node is in state k. Given that the parent node is in state j in the historical dataset, the number of nodes with state k at the i-th node. This represents the total number of samples in the historical dataset where the parent node's state is j. Given that the parent node is in state j in the expert knowledge, the confidence level is given that the state of the i-th node is k. This represents the total confidence level of the parent node with state j in the expert knowledge.

[0056] It should be noted that: can be adjusted The size allows for flexible control over the weighting ratio between data-driven and experience-driven approaches.

[0057] Existing technologies for determining the risk probability of target construction interruption events based on Bayesian network models commonly employ Monte Carlo sampling, an approximate inference algorithm that sacrifices inference accuracy for increased efficiency. However, it cannot accurately quantify the contribution of each risk factor to construction interruption, nor can it precisely identify the intensity of key risk transmission paths, such as the risk transmission path of heavy rain-road closure-material delay-construction interruption. Furthermore, errors accumulate during risk transmission, leading to a significant deviation between the final interruption probability estimate and the actual situation, thus failing to provide accurate basis for contingency planning. To address this technical problem, in some embodiments of the present invention, such as... Figure 5 As shown, step S104 includes: S501. Construct a joint tree corresponding to the directed acyclic graph structure; the joint tree contains multiple connected clique nodes.

[0058] Among them, the joint tree is an undirected tree structure that transforms the complex dependencies of Bayesian networks into a hierarchical structure of cliques, and efficiently calculates probabilities through a message passing algorithm.

[0059] The specific process of step S501 is as follows: convert the directed edges of the directed acyclic graph structure into undirected edges to obtain a moral graph with all edges being undirected, and determine the target cycle in the moral graph whose length is greater than the preset length. Then, triangulate the moral graph to obtain a chord graph. The chord is used to connect any two non-adjacent nodes in the target cycle with an edge. Then, identify the maximum clique from the chord graph and construct a joint tree with the maximum clique as the unification point.

[0060] Specifically, a maximal clique refers to a chord graph where no more nodes can be added while maintaining full connectivity.

[0061] S502. Determine the potential function of each clique node based on the conditional probability table; S503. The target construction interruption event is allocated as evidence information in the joint tree, and the evidence information is transmitted according to the connection relationship between the nodes of each group.

[0062] Among these, the evidence information may be whether a certain risk factor in the target construction interruption event has occurred or not occurred.

[0063] Evidence information allocation in the union tree is divided into two stages: evidence collection and evidence distribution. During the evidence collection phase, each node sends a message to its neighboring nodes. The message content is the product of the node's potential function and the received messages from other neighboring nodes. Then, all variables except the variables of the received message are summed. For example, if the current node has two neighboring nodes, during the evidence collection phase, the current node first receives messages from its neighboring nodes and, based on its potential function, passes a message based on the neighboring node information and the current node's message to its current parent node. The messages are passed sequentially along the union tree from the leaf nodes to the root node. Each node calculates and passes a new message based on the received message and its own potential function.

[0064] During the evidence distribution phase, each node calculates and sends messages to its neighboring nodes, and the calculation method is basically the same as that during the evidence collection phase.

[0065] S504. Update the potential function of each cluster node based on the evidence information, and determine the risk probability of the target construction interruption event based on the updated potential function.

[0066] Compared to existing technologies that determine risk probabilities through approximation and estimation, this invention utilizes the numerical calculation process of a joint tree to eliminate uncertainty and improve the accuracy of risk probabilities. Specifically, performing precise global calculations directly on a directed acyclic graph (DAG) structure is extremely complex. This invention transforms the DAG structure into an undirected chord graph, removing structural obstacles for subsequent precise information transmission. Simultaneously, the transmission of evidence information within the joint tree follows strict sum-product formulas, constituting pure mathematical calculation without any random steps. This completely eliminates the random errors inherent in approximate reasoning algorithms, enabling the calculation of the risk probability of a target construction interruption event and accurately identifying key risk sources and transmission paths. This elevates risk decision-making from fuzzy estimation to precise calculation.

[0067] To address the technical problems of existing technologies where most contingency plans are qualitative suggestions, unable to quantify their effects, and lacking the ability to optimize multiple contingency plan combinations under budget constraints, in some embodiments of the present invention, such as... Figure 6 As shown, step S105 includes: S601. Construct an objective function with the aim of maximizing the net return of the proposed combination.

[0068] It should be noted that in the process of constructing the objective function, the implementation of each contingency plan is modeled as a modification of the state (evidence) of one or more nodes in the Bayesian network model or its conditional probability table, so that the effect of the contingency plan implementation can be simulated and calculated.

[0069] S602. With the constraint that the total cost of the proposed plan combination is less than or equal to the maximum budget, the objective function is solved based on the genetic algorithm to obtain the optimal proposed plan combination.

[0070] In a specific embodiment of the present invention, the objective function is:

[0071] The constraints are:

[0072] In the formula, Construction interruption event for target Baseline probability of occurrence when no contingency plan is in place; Construction interruption event for target The probability of occurrence after adopting contingency plan combination X; Construction interruption event for target Potential economic losses upon occurrence; This is the i-th contingency plan; Let m be the cost of the i-th contingency plan; m is the total number of combined contingency plans. For the maximum budget.

[0073] In specific embodiments of the present invention, such as Figure 7 As shown, the specific process of solving the objective function based on the genetic algorithm is as follows: 1) Encoding: Each individual (chromosome) is a binary string of length m, representing a pre-plan combination.

[0074] 2) Initialization: Randomly generate an initial population in which each individual satisfies the constraints.

[0075] 3) Fitness assessment: For each individual X in the population, the Bayesian network inference algorithm from step 4 is called to calculate... This is used as the fitness value.

[0076] 4) Selection operation: Based on fitness values, selection operators (such as roulette wheel selection, tournament selection) are used to select parent individuals.

[0077] 5) Crossover operation: Perform a crossover operation on the selected parent individuals to generate offspring individuals.

[0078] 6) Mutation operation: Randomly change the gene loci of offspring individuals to introduce diversity.

[0079] 7) Constraint handling: For individuals that do not meet the constraints after crossover and mutation, a penalty mechanism (reducing their fitness) or a repair mechanism can be adopted.

[0080] 8) Iteration and Termination: Repeat the selection, crossover, and mutation process until the preset maximum number of iterations is reached or the fitness converges, and output the individual with the highest fitness as the optimal combination.

[0081] This invention, through optimization aimed at net benefit (reduction in risk loss minus cost), can automatically find the optimal combination of risk control measures or the most cost-effective solutions within a limited budget. This transforms the proposed solutions from qualitative suggestions into quantitative optimization plans, greatly improving their scientific rigor, operability, and economic efficiency.

[0082] To provide more intuitive data for subsequent project managers, some embodiments of the present invention use charts to visually display risk probabilities, optimal contingency plan combinations, and the cost-benefit ratios corresponding to the optimal contingency plan combinations, thereby assisting project managers in making decisions.

[0083] To avoid the technical problem that a Bayesian network model with fixed parameters cannot adapt to newly added data, leading to a decline in the effectiveness of subsequent risk assessment and contingency plan optimization, and making it difficult to meet the dynamic management and control needs of the entire construction cycle, in some embodiments of the present invention, the newly collected data and new expert experience are periodically integrated into the historical dataset, and steps S102-S103 are repeated to update the structure and parameters of the Bayesian network model, forming a continuously improving risk management closed loop.

[0084] In summary, the construction interruption risk reasoning and contingency plan generation method proposed in this invention has the following advantages: First, it employs a structure learning algorithm that combines scoring search and expert constraints, taking into account both objective data and domain knowledge to ensure the rationality of the Bayesian network structure; second, it integrates expert knowledge and historical datasets to determine the conditional probability table, improving the accuracy of parameter estimation in scenarios with small samples and missing data; third, it achieves accurate reasoning based on the joint tree algorithm, precisely quantifying the risk transmission path and interruption probability; and fourth, it utilizes a genetic algorithm to optimize the combination of multiple contingency plans under budget constraints, quantifying the effectiveness of the plans and balancing risk control and cost savings.

[0085] On the other hand, embodiments of the present invention also provide a construction interruption risk reasoning and contingency plan generation device, such as... Figure 8 As shown, the construction interruption risk reasoning and contingency plan generation device 800 includes: The data acquisition unit 801 is used to acquire various risk factors and historical datasets that affect construction interruption, and uses the risk factors as nodes in a Bayesian network. The directed acyclic graph (DAG) structure building unit 802 is used to construct a DAG structure based on the Bayesian information criterion as the scoring function, the historical dataset, and the constraint set determined by domain knowledge. The constraint set is used to restrict the existence of edges between nodes during the construction of the DAG structure. The Bayesian network model building unit 803 is used to determine the conditional probability table of the directed acyclic graph structure based on the directed acyclic graph structure and historical dataset. The directed acyclic graph structure and the conditional probability table constitute the Bayesian network model. Risk probability determination unit 804 is used to determine the risk probability of the target construction interruption event based on a Bayesian network model; The contingency plan combination determination unit 805 is used to optimize the preset construction contingency plan combination based on risk probability and determine the optimal contingency plan combination.

[0086] The construction interruption risk reasoning and contingency plan generation device 800 provided in the above embodiments can realize the technical solutions described in the above embodiments of the construction interruption risk reasoning and contingency plan generation method. The specific implementation principles of each module or unit can be found in the corresponding content in the above embodiments of the construction interruption risk reasoning and contingency plan generation method, which will not be repeated here.

[0087] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.), and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0088] The above provides a detailed description of the construction interruption risk reasoning and contingency plan generation method and apparatus provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for construction interruption risk reasoning and contingency plan generation, characterized in that, include: Obtain various risk factors and historical datasets that affect construction interruption, and use these risk factors as nodes in a Bayesian network; Using Bayesian information criteria as the scoring function, a directed acyclic graph structure is constructed based on the scoring function, the historical dataset, and the constraint set determined by domain knowledge. The constraint set is used to restrict the existence of edges between nodes during the construction of the directed acyclic graph structure; Based on the directed acyclic graph structure and historical dataset, a conditional probability table for the directed acyclic graph structure is determined, and the directed acyclic graph structure and the conditional probability table constitute a Bayesian network model. The risk probability of the target construction interruption event is determined based on the Bayesian network model. Based on the aforementioned risk probabilities, the preset combination of construction plans is optimized to determine the optimal combination of plans.

2. The method for construction interruption risk reasoning and contingency plan generation according to claim 1, characterized in that, The construction of the directed acyclic graph structure based on the scoring function, the historical dataset, and the constraint set determined by domain knowledge includes: Step 1: Construct the initial directed acyclic graph structure; Step 2: Adjust the edges of the initial directed acyclic graph structure to obtain multiple candidate directed acyclic graph structures; Step 3: Based on the constraint set, modify the multiple candidate directed acyclic graph structures to obtain multiple valid directed acyclic graph structures; Step 4: Determine the scoring function value for each of the effective directed acyclic graph structures based on the scoring function and the historical dataset; Step 5: Take the effective directed acyclic graph structure with the highest scoring function value as the initial directed acyclic graph structure, and return to Step 2. Repeat Step 2-Step 5 until the change in the scoring function value of the effective directed acyclic graph structure is less than a preset value, and take the effective directed acyclic graph structure as the directed acyclic graph structure.

3. The method for construction interruption risk reasoning and contingency plan generation according to claim 1, characterized in that, The scoring function is: In the formula, D represents the score function value; F represents the historical dataset; and F represents the directed acyclic graph structure. The log-likelihood function represents the degree of data fit. is the parameter set under maximum likelihood estimation; d is the number of independent parameters in the directed acyclic graph structure; N represents the amount of data in the historical dataset.

4. The method for construction interruption risk reasoning and contingency plan generation according to claim 1 or 2, characterized in that, The constraint set includes a set of mandatory edges and a set of prohibited edges. Edges in the set of mandatory edges must exist in the directed acyclic graph structure, while edges in the set of prohibited edges must not exist in the directed acyclic graph structure.

5. The method for construction interruption risk reasoning and contingency plan generation according to claim 1, characterized in that, The step of determining the conditional probability table of the directed acyclic graph structure based on the directed acyclic graph structure and historical dataset includes: Based on the historical dataset, determine the conditional probability of each node in the directed acyclic graph structure; The virtual conditional probabilities of each node in the directed acyclic graph structure are determined based on expert knowledge. The conditional probabilities of the data and the virtual conditional probabilities are fused to obtain the conditional probabilities of each node, and the conditional probabilities of all nodes constitute the conditional probability table.

6. The method for construction interruption risk reasoning and contingency plan generation according to claim 5, characterized in that, The conditional probabilities of each node are: In the formula, Given that the parent node is in state j, the conditional probability that the i-th node is in state k. Given that the parent node is in state j in the historical dataset, the number of nodes with state k at the i-th node. This represents the total number of samples in the historical dataset where the parent node's state is j. Given that the parent node is in state j in the expert knowledge, the confidence level is given that the state of the i-th node is k. This represents the total confidence level of the parent node with state j in the expert knowledge.

7. The method for construction interruption risk reasoning and contingency plan generation according to claim 1, characterized in that, The determination of the risk probability of the target construction interruption event based on the Bayesian network model includes: Construct a joint tree corresponding to the directed acyclic graph structure; the joint tree contains multiple connected clique nodes; The potential function of each clique node is determined according to the conditional probability table; The target construction interruption event is allocated as evidence information in the joint tree, and the evidence information is transmitted according to the connection relationship between the nodes of each group; The potential function of each cluster node is updated based on the evidence information, and the risk probability of the target construction interruption event is determined based on the updated potential function.

8. The method for construction interruption risk reasoning and contingency plan generation according to claim 1, characterized in that, The optimization of the preset combination of construction plans based on the risk probability to determine the optimal combination of plans includes: The objective function is constructed with the aim of maximizing the net return of the proposed plan combination; With the constraint that the total cost of the proposed plan combination is less than or equal to the maximum budget, the objective function is solved based on a genetic algorithm to obtain the optimal proposed plan combination.

9. The method for construction interruption risk reasoning and contingency plan generation according to claim 8, characterized in that, The objective function is: The constraints are as follows: In the formula, Construction interruption event for target Baseline probability of occurrence when no contingency plan is in place; Construction interruption event for target The probability of occurrence after adopting contingency plan combination X; Construction interruption event for target Potential economic losses upon occurrence; This is the i-th contingency plan; Let m be the cost of the i-th contingency plan; m is the total number of combined contingency plans. For the maximum budget.

10. A device for construction interruption risk reasoning and contingency plan generation, characterized in that, include: The data acquisition unit is used to acquire multiple risk factors and historical datasets that affect construction interruption, and to use the risk factors as nodes in a Bayesian network. A directed acyclic graph (DAG) structure building unit is used to construct a DAG structure based on the Bayesian information criterion as a scoring function, the historical dataset, and a constraint set determined by domain knowledge; the constraint set is used to restrict the existence of edges between nodes during the construction of the DAG structure. A Bayesian network model building unit is used to determine the conditional probability table of the directed acyclic graph structure based on the directed acyclic graph structure and the historical dataset. The directed acyclic graph structure and the conditional probability table constitute a Bayesian network model. The risk probability determination unit is used to determine the risk probability of the target construction interruption event based on the Bayesian network model. The contingency plan combination determination unit is used to optimize the preset construction contingency plan combination based on the risk probability and determine the optimal contingency plan combination.

Images