Construction progress and resource coordination prediction data processing method and related products

Abstract

The present application relates to the technical field of data processing, in particular to a kind of construction progress and resource collaborative prediction data processing method and related product, comprising: constructing BIM knowledge graph;Real-time monitoring real-time dynamic data, if it is identified that there is a risk event deviating from construction plan, generate multi-dimensional risk prediction vector;Construct multi-objective optimization model, solve and generate collaborative optimization strategy;Execute collaborative optimization strategy, update construction progress plan and resource allocation plan, and update BIM knowledge graph synchronously;The present application constructs BIM knowledge graph by fusing multi-source heterogeneous data, the present application breaks the information barrier between traditional BIM design data and field execution data, and by the chain influence propagation simulation to risk event, in a quantitative manner, the future time and cost deviation is predicted, and project management is changed from traditional "after the event remediation" mode to "before the event pre-control" mode.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a construction progress and resource coordination prediction data processing method and related products, and especially to a data processing method for construction progress dynamic prediction, chain risk quantification analysis and resource coordination optimization. BACKGROUND

[0002] The modern construction industry is rapidly developing towards digitization and informatization. Building Information Modeling (BIM) technology has been widely applied. Currently, BIM technology is mainly used in the design and construction preparation stages of a project. By creating a three-dimensional digital model, it realizes functions such as visual design, pipeline collision checking, and construction scheme simulation. By combining BIM models with time dimensions (4D) and cost dimensions (5D), it can achieve planning and control of construction progress and cost to a certain extent.

[0003] At the same time, the rise of the concept of smart construction site has also promoted the improvement of data collection capabilities on the construction site. By deploying various Internet of Things (IoT) sensors (such as environmental monitoring, equipment status monitoring), personnel real-name system based on face recognition, video monitoring system, etc. on the construction site, project managers can obtain a large amount of real-time dynamic data about people, machines, materials, methods, and environment on site.

[0004] However, in the prior art, there is a data island problem between the static, planned design data represented by BIM and the dynamic, real-time data collected by the smart construction site. The data usually belongs to different application systems and lacks deep, real-time semantic fusion and interaction.

[0005] Therefore, the current project management mode is still largely passive and lagging. Managers often learn about information after a risk event (such as: delay in transportation of key materials, sudden failure of core equipment, or absence of key personnel) has occurred, and determine its impact on subsequent engineering based on personal experience. The existing technical means cannot predict the complex chain effects on time and cost of all subsequent processes of the entire project based on the real-time occurrence of local risks. SUMMARY

[0006] The technical problem to be solved by the present application is how to deeply integrate static BIM planning models with dynamic on-site execution data, with the aim of providing a construction progress and resource coordination prediction data processing method and related products, and realizing the transition from a passive post-response management mode to an active pre-control management mode.

[0007] The present application is implemented by the following technical solutions:

[0008] A construction progress and resource coordination prediction data processing method, comprising:

[0009] Components in a building information model are taken as component nodes, processes in a construction progress plan are taken as process nodes, and a relationship edge between process nodes is established according to a preceding and following logical relationship in the plan, a resource node is associated to a process node, and a BIM knowledge graph is constructed;

[0010] Real-time dynamic data of a construction site is accessed and monitored in real time, if a risk event deviating from a construction plan is identified, influence propagation simulation is performed along a relationship edge of the BIM knowledge graph, and a multi-dimensional risk prediction vector is generated, the multi-dimensional risk prediction vector at least including: time deviation and cost deviation;

[0011] Based on the multi-dimensional risk prediction vector, a multi-objective optimization model is constructed with the goal of minimizing time deviation and cost deviation, and the multi-objective optimization model is solved under the boundary conditions of project available resources and process logic constraints, and a coordinated optimization strategy containing a specific resource allocation scheme or a process adjustment scheme is generated;

[0012] The coordinated optimization strategy is executed, the construction progress plan and the resource allocation plan are updated, and the BIM knowledge graph is synchronously updated.

[0013] Optionally, the method of constructing the BIM knowledge graph comprises:

[0014] The preceding and following logical relationship between process nodes is defined as a directed edge carrying time sequence constraint information, and the time sequence constraint information at least includes one of complete-start, start-start or complete-complete;

[0015] The resource node is associated to the process node by creating a "demand" or "consumption" type of attribute edge, and the specific requirements of the process on the quantity, specification, skill level or use time of the resource are defined on the attribute edge;

[0016] The real-time dynamic data is classified and attached to the corresponding resource node as a dynamic attribute, wherein the real-time dynamic data at least includes: state monitoring data of an Internet of Things device associated with a device type resource node, personnel real-name system attendance data associated with a human resource node, and material access information associated with a material resource node.

[0017] Optionally, the method of identifying the risk event deviating from the construction plan at least comprises:

[0018] Human resource availability identification: compare the planned quantity and skill requirement of human resource nodes associated with the upcoming construction process with the actual attendance and skill information of corresponding personnel in real-time dynamic data, and identify human resource shortage risk events when the actual attendance is less than the planned quantity or the skills do not match;

[0019] Equipment resource availability identification: obtain real-time state data of Internet of Things of equipment resource nodes associated with the upcoming construction process, and identify equipment resource unavailability risk events when the real-time state data indicates "failure", "offline" or "maintenance";

[0020] Material availability identification: query the real-time arrival status of material resource nodes required by the upcoming construction process, and identify material unavailability risk events when at least one required material is in the state of "not arrived" or "transportation delay".

[0021] Optionally, the method of performing influence propagation simulation to generate a multi-dimensional risk prediction vector comprises:

[0022] Starting from the process node or resource node directly associated with the risk event, along the pre-post logical relationship edge between the process nodes defined in the BIM knowledge graph, recursively calculate the predicted start time and predicted completion time of all affected downstream process nodes;

[0023] For each affected downstream process node, compare its predicted start time with the planned start time in the construction plan to quantitatively calculate the time deviation of the process ;

[0024] Based on the quantitative calculation of the time deviation, the cost deviation is calculated :

[0025] Identify human or equipment resources associated with downstream processes in a waiting state due to upstream process delays, and accumulate the resource idle time cost according to the unit time cost and waiting time;

[0026] If the time deviation causes the total project duration to exceed the completion date agreed in the contract, calculate the expected breach of contract cost according to the pre-set contract penalty clause.

[0027] Optionally, the method of establishing a multi-objective optimization model comprises:

[0028] Establish a double-objective function with the goal of minimizing the predicted total duration and the predicted total cost: , wherein: is the candidate collaborative optimization strategy, is the solution space composed of all valid strategies that meet the constraints, is the application strategy The subsequent function for predicting the total project duration. For application strategy The subsequent prediction function for the total project cost;

[0029] Define at least two constraints:

[0030] Resource constraints: at any given time Any kind of resource The total usage shall not exceed the total availability of the resource: ,in, For strategy Construction tasks At any moment Resources The demand, For resources Total available quantity, This refers to any point in the project timeline. , This refers to any type of resource in the project. , For at any time A collection of all ongoing construction tasks;

[0031] Logical constraints: All processes must satisfy their pre- and post-dependencies defined in the BIM knowledge graph. ,in For strategy Next preprocess The predicted completion time, For strategy Next process The predicted start time, For all processes with direct pre- or post-dependent dependencies A set of.

[0032] Alternatively, methods for solving multi-objective optimization models include:

[0033] Initialization: Initialization is performed by One candidate collaborative optimization strategy population Initialize the state-action value function. ;

[0034] Iterative solution: For each iteration Perform the following sub-steps:

[0035] a) State Recognition and Action Selection: The initial risk prediction vector that triggered this optimization is taken as the current state. And based on the value function adopting a strategy to select an action , the action corresponding to a preset mutation operator preference, the preference at least including prioritizing shortening the project duration, prioritizing reducing the project cost, or balancing optimization;

[0036] b) genetic operation: performing genetic operation on the current population adopting tournament selection and simulated binary crossover operator, and performing adaptive polynomial mutation operation according to the mutation preference corresponding to the action to generate a child population ;

[0037] c) environment evaluation and elite selection: merging the parent population and the child population , performing fast non-dominated sorting and crowding distance calculation, and selecting optimal individuals from the merged population to form a new parent population ;

[0038] reinforcement learning model updating: calculating the reward value of the current iteration according to the increment of the new generated parent population relative to the hyper-volume index of the Pareto front, and updating the value of the state-action pair in the value function using the following Q-learning update rule: wherein is the learning rate, is the discount factor, and is the state of the next iteration.

[0039] termination and output: repeating the iteration until the preset termination condition is met, and outputting a set of non-dominated solutions on the Pareto optimal front in the final population as the collaborative optimization strategy.

[0040] Optionally, the method of performing closed-loop updating comprises:

[0041] selecting a final execution strategy from the collaborative optimization strategy, and parsing the final execution strategy to obtain adjustment instructions, the adjustment instructions at least including adjustment operations on the process, human resources, or equipment resources;

[0042] directly updating the construction schedule and resource allocation plan based on the adjustment instructions, the update content at least including adjusting the planned start and end time of the affected process, or changing the allocation relationship between the process and the resource;

[0043] ​Reflect the update content in the construction progress plan and the resource allocation plan back to the BIM knowledge graph synchronously, update the time attribute of the corresponding process node in the BIM knowledge graph, and update the association relationship edge or the attribute between the process node and the resource node;

[0044] The updated BIM knowledge graph is set as a new baseline plan, and real-time monitoring is continued to identify new risk events.

[0045] Further, the data processing method further comprises a reverse reinforcement learning strategy optimization step method selected by the manager, comprising:

[0046] Constructing an implicit reward function model , wherein is a weight parameter vector representing the decision preference of the manager;

[0047] Based on the maximum entropy principle, determine the probability function of the final execution strategy selected by the manager from the collaborative optimization strategy ;

[0048] Based on the plurality of historical decision pairs recorded continuously , by solving the maximum problem of the log-likelihood function, the optimal weight of the current stage is calculated: ;

[0049] After generating a new parent population , the preference reward of the new parent population is calculated based on the optimal weight : ; ;

[0050] The preference reward and the reward value are combined to obtain a composite reward signal : , wherein is a preset hyperparameter;

[0051] The composite reward signal is used to replace the original reward value , and the value of the corresponding state-action pair in the value function is updated using the Q-learning update rule.

[0052] A construction progress and resource collaborative prediction data processing system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the construction progress and resource collaborative prediction data processing method as above when executing the computer program.​​

[0053] A computer program product comprising computer programs / instructions which, when executed by a processor, implement the construction progress and resource collaborative prediction data processing method as above.

[0054] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0055] The present application fuses building information model, construction progress plan and real-time dynamic data to construct a BIM knowledge graph, predicts the influence of a risk event on future project total duration and total cost by using the BIM knowledge graph when the risk event exists, then solves the selectable collaborative optimization strategy by a constructed multi-objective optimization algorithm under the condition of meeting the engineering reality constraints, and updates the project plan and the BIM knowledge graph, forming a closed-loop management process of "monitoring-prediction-decision-execution".

[0056] The present application breaks the information barrier between traditional BIM design data and site execution data by constructing a BIM knowledge graph fusing multi-source heterogeneous data, and predicts the future time and cost deviation in a quantitative way by simulating the chain influence propagation of the risk event, thereby changing the project management from the traditional "after-the-fact remediation" mode to the "before-the-fact pre-control" mode.

[0057] The present application generates a collaborative optimization strategy considering time and cost by using a multi-objective optimization algorithm, improves the decision-making efficiency in response to emergencies, and also makes resource allocation and plan adjustment more reasonable and efficient. BRIEF DESCRIPTION OF DRAWINGS

[0058] The accompanying drawings illustrate exemplary embodiments of the present application and together with the general description given above and the detailed description given below, serve to explain the principles of the application. These drawings are included herewith to provide a further understanding of the present application and are incorporated in and constitute a part of this specification, illustrate embodiments of the present application and, together with the description, serve to explain the principles of the present application, and are included as part of this specification to provide a further understanding of the present application.

[0059] Figure 1 is a schematic diagram of the overall process according to the construction progress and resource collaborative prediction data processing method of the present application.

[0060] Figure 2 is a detailed flowchart of the construction progress and resource collaborative prediction data processing method according to the present application.

[0061] Figure 3 is a flowchart of the multi-objective model establishment and solution according to the present application. DETAILED DESCRIPTION

[0062] In order to make the objectives, technical solutions and advantages of the present application clearer, further detailed description will be made to the present application with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the related content, but not limit the present application.

[0063] In addition, it should be further noted that, for the convenience of description, only the parts related to the present application are shown in the drawings.

[0064] The embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0065] Embodiment one

[0066] The present embodiment discloses a predictive data processing method for construction engineering management, and the main steps include: constructing a dynamic digital model, i.e. a BIM knowledge graph, when there is a sign of deviation from the plan in the real world construction site, using the digital model to deduce the chain effect of the sign on the future, and generating a coping strategy; the adopted strategy will update the digital model in reverse, thereby forming a management closed loop process from "perception-prediction-decision-execution".

[0067] As shown in Figure 1 and Figure 2 , the construction progress and resource coordination predictive data processing method for specific implementation includes the following steps:

[0068] Firstly, the components in the building information model are taken as component nodes, the processes in the construction progress plan are taken as process nodes, the relationship edges between the process nodes are established according to the preceding and following logical relationship in the plan, the resource nodes are associated to the process nodes, and a BIM knowledge graph is constructed. The knowledge graph deeply integrates and associates the following four types of core information:

[0069] Component node: each specific building component derived from BIM, such as a beam or a wall.

[0070] Process node: each specific work task derived from the construction progress plan, such as "binding reinforcement" or "pouring concrete".

[0071] Resource node: various resources required for executing the process, such as workers with specific skills, tower cranes or cement with specific specifications.

[0072] Real-time dynamic data: data stream collected from the construction site in real time, such as the actual on-site situation of workers, the working state of equipment, etc.

[0073] By connecting these nodes according to engineering logic (e.g. which component needs to be built in which process, which resources are needed, the sequence of processes), a digital twin that can fully and dynamically reflect the project plan and reality is built.

[0074] Second, access real-time dynamic data on the construction site and monitor it in real time. If a risk event deviating from the construction plan is identified, an impact propagation simulation is performed along the relationship edges of the BIM knowledge graph, generating a multi-dimensional risk prediction vector, which at least includes time deviation and cost deviation.

[0075] The system will monitor the dynamic data collected on site in real time and compare it with the planned data stored in the BIM knowledge graph to identify risk events. Risk events refer to any deviation from the predetermined plan, such as planned input of 10 workers but actual arrival of only 8.

[0076] Once a risk event is identified, the system will immediately start an impact propagation simulation, i.e. digital deduction. The system will calculate how this small deviation will be amplified and transmitted in the future construction chain along the pre-set process dependency relationship (i.e. "relationship edge") in the knowledge graph.

[0077] The final product of the simulation is a multi-dimensional risk prediction vector, which indicates potential future losses through quantitative prediction reports, for example, the predicted results may be: the total project duration will be delayed by 5 days (time deviation), and the total cost will exceed the budget by 100,000 yuan (cost deviation).

[0078] Third, based on the multi-dimensional risk prediction vector, a multi-objective optimization model is constructed to minimize time deviation and cost deviation, and the multi-objective optimization model is solved under the boundary conditions of project available resources and process logic constraints to generate a collaborative optimization strategy containing specific resource allocation schemes or process adjustment schemes.

[0079] In the face of the future risks predicted in the second step, the system constructs a multi-objective optimization model to find the best balance point between multiple conflicting objectives (i.e. "minimize time deviation" and "minimize cost deviation").

[0080] The model will be solved under the constraints of the real world (e.g. there are only these available personnel and equipment on the project, the sequence of some processes cannot be reversed). The result of the solution is a series of collaborative optimization strategies that managers can choose from, i.e. executable action plans, such as: suggest that team B be deployed from a secondary task to the current bottleneck task, suggest that team A work overtime for 2 hours to catch up with the schedule.

[0081] Fourth, execute the collaborative optimization strategy, update the construction schedule and resource allocation plan, and update the BIM knowledge graph simultaneously.

[0082] When the manager selects one from the multiple strategies generated in the third step and decides to execute it, the decision result is solidified into the system and plan, and then the decision is used to formally update the construction schedule plan and resource allocation plan of the project.

[0083] At the same time, the changes of these plans are synchronously updated to the BIM knowledge graph, so the state of the BIM knowledge graph is always consistent with the updated plans, thereby ensuring the continuous effectiveness of the entire management process, forming a complete closed loop.

[0084] In summary, the working principle of the method described in the embodiment is: by constructing a BIM knowledge graph that integrates plan information and real-time dynamic data as a digital foundation, using the BIM knowledge graph to simulate the chain effect of risk events on site to obtain a quantitative prediction of the future; based on the prediction result, a scientific response strategy is automatically generated through a multi-objective optimization model; the selected strategy is executed and synchronously updated back to the knowledge graph, thereby completing a "perception-analysis-decision-feedback" closed loop management.

[0085] Embodiment Two

[0086] This embodiment details the specific steps of the first step in Embodiment One, by defining different types of "edges" (i.e. relationships) between nodes that carry specific information, isolated engineering data (such as processes, resources, real-time state) is integrated into a knowledge graph. The method of constructing a BIM knowledge graph includes:

[0087] S11, define the pre-post logical relationship between process nodes as a directed edge carrying time constraint information, and the time constraint information at least includes one of complete-start, start-start or complete-complete;

[0088] Give the time dependency relationship between each work task, that is, the "flowchart" of the project. The pre-post relationship between process nodes is defined as a directed edge. The directed edge has a clear direction of connection, for example, from "process A" to "process B", indicating that A has an impact on B.

[0089] Standard logical rules in project management, this embodiment at least includes the following types:

[0090] Complete-start: the pre-process must be completed before the subsequent process can start. For example, after the foundation excavation is completed, the foundation pouring can start.

[0091] Start-start: the start time of the subsequent process cannot be earlier than the start time of the pre-process.

[0092] Complete-complete: the completion time of the subsequent process cannot be earlier than the completion time of the pre-process.

[0093] S12, associate resource nodes to process nodes by creating attribute edges of type "demand" or "consumption", and define the specific requirements of the process on the resource in quantity, specification, skill level or usage time on the attribute edge.

[0094] Connect resource nodes (such as manpower, materials, equipment) to process nodes that need them by creating attribute edges of specific types. Attribute edges only represent the association and explain the nature of the association, at least including:

[0095] "Demand" type: usually used for reusable resources such as equipment or specific artisans. For example, a "high-altitude welding" process is associated with a "welder" resource node and a "sling" equipment node through a "demand" edge.

[0096] "Consumption" type: usually used for consumable materials. For example, a "wall building" process is associated with a "brick" material node and a "cement" material node through a "consumption" edge.

[0097] Such as the number of resources required to complete the process (such as: 5 workers are needed), specification (such as: C30 grade concrete is needed), skill level (such as: a special welder is needed) or usage time (such as: a tower crane needs to be rented for 8 hours).

[0098] S13 classifies real-time dynamic data and attaches it as dynamic attributes to the corresponding resource nodes, where the real-time dynamic data at least includes: state monitoring data of Internet of Things devices associated with equipment class resource nodes, personnel real-name attendance data associated with manpower class resource nodes, and material access information associated with material class resource nodes.

[0099] Classify the collected real-time dynamic data and attach it as dynamic attributes to the corresponding resource nodes in the knowledge graph. At least including:

[0100] Attach the state monitoring data (such as: running, failure, offline) of Internet of Things (IoT) devices (such as tower cranes, construction elevators) to the corresponding equipment class resource nodes.

[0101] Attach the attendance data (such as: present, leave) collected by personnel real-name channel to the corresponding manpower class resource nodes.

[0102] Attach the material access information (such as: arrived, in transit, delayed) obtained by RFID or scanning code to the corresponding material class resource nodes.

[0103] In summary, the operation steps of constructing the BIM knowledge graph in this embodiment first construct the logical skeleton of the entire project by defining directed edges carrying time sequence constraints; then, fill the skeleton with the necessary resources for each link by defining "demand" or "consumption" attribute edges carrying specific requirements; finally, form a dynamic knowledge network by attaching real-time dynamic data as dynamic attributes to the corresponding resource nodes.

[0104] Embodiment Three

[0105] This embodiment details the specific process of risk prediction, which consists of the following two main parts:

[0106] S21, identify risk events deviating from the construction plan, and determine whether the key resources on the construction site are in the "ready for use" state. This embodiment includes at least the following three identification methods:

[0107] Human resource availability identification: compare the planned demand quantity and skill requirements (e.g., 5 certified welders are needed) of the human resource nodes associated with the upcoming construction process with the actual attendance check-in quantity and skill information of the corresponding personnel in real-time dynamic data. When the actual attendance quantity is less than the planned demand quantity or the skills do not match, it is identified as a human resource shortage risk event.

[0108] Equipment resource availability identification: obtain the real-time state data of the Internet of Things of the equipment resource nodes (e.g., tower crane) associated with the upcoming construction process. When the real-time state data indicates "failure", "offline" or "maintenance", it is identified as an equipment resource unavailable risk event.

[0109] Material availability identification: query the real-time arrival status of the material resource nodes (e.g., specific type of steel bar or prefabricated parts) required for the upcoming construction process. When at least one required material is in the "not arrived" or "transportation delayed" state, it is identified as a material unavailability risk event.

[0110] S22, perform impact propagation simulation to generate a multi-dimensional risk prediction vector, including the following steps:

[0111] S221, starting from the process nodes or resource nodes directly associated with the risk events, along the pre-post logical relationship edges defined in the BIM knowledge graph between the process nodes, recursively calculate the predicted start time and predicted completion time of all affected downstream process nodes.

[0112] This process starts from a node directly associated with a risk event (e.g. an unavailable equipment resource node or a delayed process node). The system will recursively propagate along the defined process dependency relationships (i.e. "precedence logic relationship edges") in the BIM knowledge graph. It will calculate how the initial delay will be propagated to all the subsequent associated processes and update their predicted start time and predicted completion time.

[0113] S222, for each affected downstream process node, compare its predicted start time with the planned start time in the construction plan to quantify the time deviation of this process .

[0114] After the simulation calculates the new predicted time, the system will compare the predicted start time of each affected downstream process with the planned start time in the original construction plan, and the difference between the two is the time deviation of the process.

[0115] S223, after quantifying the time deviation, the system will further convert it into financial impact, based on the time deviation quantification to calculate the cost deviation , at least including the following two dimensions:

[0116] (1) Identify the human or equipment resources associated with the downstream processes that are in a waiting state due to upstream process delays, and accumulate the resource idle cost based on the unit time cost and waiting time of the resources.

[0117] Identify those downstream processes that are forced to stop and wait due to upstream process delays (such as materials not arriving). Then, based on the unit time cost (e.g. hourly wage or rental fee) of the human or equipment resources already arranged for these downstream processes and the waiting time (i.e. upstream delay), accumulate the direct economic loss caused by resource idling.

[0118] (2) If the time deviation causes the total project duration to exceed the completion date agreed in the contract, calculate the expected breach of contract cost based on the pre-set contract penalty clause.

[0119] Calculate the total project duration delay caused by the cumulative time deviation. If this total delay causes the final completion date to exceed the contract stipulation, the system will automatically calculate the future possible contract breach cost according to the pre-set penalty clause in the project contract (e.g. X yuan per day of delay).

[0120] By comparing the real-time status and plan of three key resources, i.e. human, equipment and material, the occurrence of risk events is identified. Once the event is confirmed, a recursive simulation process is started to calculate and quantify the impact of the event on future schedule (time deviation) and further calculate the impact on future cost (cost deviation) based on time delay and contract terms.

[0121] Embodiment Four

[0122] This embodiment illustrates the strategy generation step in the whole method, which establishes a multi-objective optimization model and then solves the model by a technique combining reinforcement learning and genetic algorithm.

[0123] As shown in Figure 3 , the method for establishing the multi-objective optimization model comprises:

[0124] S31, in project management, time and cost are often mutually restrictive and cannot be simultaneously optimized, so a double-objective function is established to minimize the predicted total duration and the predicted total cost: wherein: is a candidate synergistic optimization strategy, is a solution space composed of all valid strategies satisfying the constraint conditions, is a predicted project total duration function after applying the strategy , and is a predicted project total cost function after applying the strategy .

[0125] A candidate synergistic optimization strategy is sought, which can make the final predicted project total duration and the final predicted project total cost reach as small as possible simultaneously in all valid solution spaces.

[0126] S32, at least two constraint conditions are defined:

[0127] In order to ensure that any generated strategy is realistic and feasible, a resource constraint is defined: the total usage of any resource at any time does not exceed the total available amount of the resource: wherein: is the demand of the construction task under the strategy for the resource at time , is the total available amount of the resource , and represents any time on the project timeline. , represents any type of resource in the project , represents the time instant at which the construction task is scheduled to be completed.

[0128] To ensure that all strategies follow the basic construction sequence, logical constraints are defined: all processes satisfy their pre- and post- dependencies defined in the BIM knowledge graph: where is the predicted completion time of the pre-process under strategy , is the predicted start time of the post-process under strategy , is the set of all process pairs with direct pre- and post- dependencies.

[0129] The method for solving the multi-objective optimization model comprises:

[0130] S33, initialization: initializing a population consisting of candidate collaborative optimization strategies ; initializing a state-action value function for storing;

[0131] S34, iterative solving: for each iteration , the following sub-steps are performed:

[0132] a) state identification and action selection: taking the initial risk prediction vector triggering this optimization as the current state , and selecting an action based on the value function using strategy, the action corresponding to the preset mutation operator preference, which at least includes prioritizing shortening the construction period, prioritizing reducing the cost, or balancing optimization;

[0133] b) genetic operation: using tournament selection and simulated binary crossover operator on the current population , and performing adaptive polynomial mutation operation according to the mutation preference corresponding to the action , to generate a child population ;

[0134] c) environment evaluation and elite selection: merging the parent population and the child population , performing fast non-dominated sorting and crowding degree calculation, and selecting The best individual forms a new parent population ;

[0135] S35, reinforcement learning model update: update the value function relative to the increment on the hyper-volume indicator on the Pareto front, calculate the reward value of this iteration , and use the following Q-learning update rule to update the value of the corresponding state-action pair in the value function : where is the learning rate, is the discount factor, is the state of the next iteration.

[0136] S36, termination and output: repeat the iteration until the preset termination condition is met (e.g., the maximum number of iterations is reached or the solution set quality is no longer significantly improved), and output a set of non-dominated solutions on the Pareto optimal front in the final population as the collaborative optimization strategy.

[0137] The system outputs all non-dominated solutions on the Pareto optimal front in the final elite population, which is not a single optimal solution, but a set of strategy combinations with different advantages (some favor time saving, and some favor cost saving).

[0138] Example Five

[0139] This embodiment details the specific method of performing closed-loop updates, including:

[0140] S41, select a final execution strategy from the collaborative optimization strategy, and parse the final execution strategy to obtain adjustment instructions, which at least include adjustment operations on processes, human resources, or equipment resources.

[0141] After the system (as described in Example Four) generates a series of collaborative optimization strategies, it will be presented to the project manager. The manager will combine his professional experience and other factors in the project that have not been modeled to select a final execution strategy from the collaborative optimization strategies.

[0142] Once the manager makes a selection, the system will parse the strategy (e.g., "add human support to process A") into specific adjustment instructions. For example, the parsed instructions may be: "establish a 'demand' association between the human resource node 'B team' and the process node 'A process' within a specific time period."

[0143] S42, directly update the construction schedule and resource allocation plan based on the adjustment instructions, and the update content at least includes adjusting the planned start and end time of the affected process, or changing the allocation relationship between the process and the resource.

[0144] After obtaining the specific adjustment instructions, the system will directly update the construction progress plan and resource allocation plan of the project based on these instructions, for example:

[0145] In the construction progress plan, modify the "planned start time" and "planned completion time" of the affected process.

[0146] In the resource allocation plan (such as the resource allocation table), remove the allocation relationship between a certain resource (such as a excavator) and the original process, and re-allocate it to a new process.

[0147] S43, the updated content in the construction progress plan and resource allocation plan is reflected back to the BIM knowledge graph, updating the time attributes of the corresponding process nodes in the BIM knowledge graph, and updating the association relationship edges or their attributes between the process nodes and the resource nodes;

[0148] After the traditional project management plan file is updated, the system must synchronously reflect these changes back to the BIM knowledge graph, at least including:

[0149] Update the time attributes of the process nodes in the knowledge graph corresponding to the adjusted process (such as planned start and end time, duration, etc.).

[0150] Update the association relationship edges or their attributes between the process nodes and the resource nodes. For example, delete an old resource allocation edge, and create a new edge according to the new plan, or modify the attributes of the edge about resource usage quantity or time.

[0151] S44, set the updated BIM knowledge graph as the new baseline plan, continue to perform real-time monitoring, and identify new risk events, that is, the system will start comparing the real-time dynamic data collected on site with this new baseline plan to identify new risk events that may occur in the future.

[0152] Embodiment six

[0153] This embodiment is based on the previous embodiments and further discloses a method for enabling the system to have self-learning and self-optimization capabilities, which enables the system to observe and analyze the historical decisions of the project manager through reverse reinforcement learning, and reversely infers the decision preferences (for example, whether to pay more attention to time or cost at the current stage of the project). Then, the system fine-tunes its internal strategy generation algorithm, so that the future recommended solutions can better meet the real needs and expert experience of the manager.

[0154] That is, the data processing method further includes a reverse reinforcement learning strategy optimization step method based on the manager's selection, including:

[0155] S51, construct an implicit reward function model wherein, is the weight parameter representing the manager's decision preference.

[0156] S52, determine the probability function of the manager choosing the final execution strategy from the co-optimization strategies based on the maximum entropy principle. .

[0157] The maximum entropy principle is to make the most uncertain and random assumption about the unknown part under the premise of satisfying the known facts, that is, the probability of the manager choosing a specific strategy from a series of alternative strategies is considered to be proportional to the evaluation value of the strategy under its implicit reward function, and can be expressed by the Boltzmann distribution (or Softmax function).

[0158] S53, based on the recorded multiple historical decision pairs , by solving the maximum problem of the log-likelihood function, the optimal weight of the current stage is calculated: .

[0159] The log-likelihood function is to measure how likely it is to observe the entire historical decision record under the assumption that the manager's preference weight is . Therefore, by solving the maximum problem of this function, the system can find a set of optimal weights .

[0160] S54, apply the learned manager preference to the solving algorithm of embodiment four to fine-tune it.

[0161] After generating a new parent population , the preference reward of the new parent population is calculated based on the optimal weight , that is, the quality of the new population is evaluated from the perspective of the manager, and .

[0162] S55, in order to consider both the objective solution set quality and the manager's subjective preference, the preference reward and the reward value are combined to obtain a composite reward signal : wherein, is a preset hyperparameter (with a value between 0 and 1).

[0163] S56, use the composite reward signal instead of the original reward value , and use the Q-learning update rule to update the value function​​ The value of each state-action pair is updated. After fine-tuning, the value function will tend to give higher values to those "tactical actions" that can produce results more in line with the manager's preferences, thus guiding the entire algorithm to generate more satisfactory collaborative optimization strategies for the manager in future iterations.

[0164] This embodiment mathematically describes the decision-making behavior of the manager, and reversely infers its internal trade-off preferences for time and cost from its historical choices through the maximum entropy principle and maximum likelihood estimation. Subsequently, this learned preference is quantified as a "preference reward" and combined with the original objective evaluation indicators of the algorithm to form a composite reward signal, and the composite reward is used to update and fine-tune the underlying reinforcement learning model, so that the decision-making recommendation capability of the entire system can converge to the manager's preferences during use.

[0165] Embodiment Seven

[0166] A construction progress and resource collaborative prediction data processing terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the construction progress and resource collaborative prediction data processing method described above when executing the computer program.

[0167] The memory can be used to store software programs and modules, and the processor can execute various functional applications and data processing of the terminal by running the software programs and modules stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one execution program required by a function, etc.

[0168] The data storage area can store data created according to the use of the terminal, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other volatile solid-state memory device.

[0169] A computer readable storage medium, the computer readable storage medium stores a computer program, the computer program is executed by a processor to implement the construction progress and resource collaborative prediction data processing method described above.

[0170] Without loss of generality, computer-readable media can include computer storage media and communication media. Computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer storage media includes, but is not limited to, RAM, ROM, EPROM, EEPROM, flash memory or other solid state memory technology, CD-ROM, DVD, or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices. It should be understood by those skilled in the art that computer storage media can not be limited to the above-mentioned several types. The system memory and mass storage device mentioned above can be collectively referred to as memory.

[0171] A computer program product includes computer programs / instructions that, when executed by a processor, implement any one of the above-described construction progress and resource coordination prediction data processing methods.

[0172] A computer program product includes computer programs or instruction sets for performing specific tasks or implementing specific functions. These programs or instructions are designed to be executed by a processor, thereby implementing a series of predefined steps or operations. The program product can be stored in various forms of computer storage media, such as memory, hard disk, solid state drive, optical disk or other forms of digital storage devices. It can exist in the form of compiled binary code, or in the form of scripts or bytecodes executable by an interpreter. The program product, through carefully designed algorithms and logical instructions, enables the processor to process data in a specific order and manner, complete various functions such as data analysis, user interaction, device control, etc.

[0173] In the description of the present specification, the description of the terms "one embodiment / way", "some embodiments / ways", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment / way or example are included in at least one embodiment / way or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment / way or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments / ways or examples. In addition, the skilled in the art can combine and combine the different embodiments / ways or examples described in the present specification and the features of the different embodiments / ways or examples without contradiction.

[0174] Furthermore, the terms "first", "second", etc. are used herein for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly pointing to the number of technical features indicated. Thus, features defined with "first", "second" etc. can explicitly or implicitly include at least one of such features. In the description of the application, the meaning of "plurality" is at least two, for example two, three, etc., unless explicitly and specifically defined otherwise.

[0175] Those skilled in the art will understand that the above-described embodiments are merely intended to clarify the present application and are not intended to limit the scope of the present application. Other changes or modifications can be made by those skilled in the art based on the above-described application, and such changes or modifications are still within the scope of the present application.

Claims

1. A method for predicting construction progress and resource coordination data processing, characterized in that, include: The components in the building information model are used as component nodes, the processes in the construction schedule are used as process nodes, and relationship edges between process nodes are established according to the preceding and following logical relationships in the plan. Resource nodes are associated with process nodes, real-time dynamic data is classified and attached to its corresponding resource nodes as dynamic attributes to construct a BIM knowledge graph. Access real-time dynamic data from the construction site and monitor the data in real time. If a risk event deviating from the construction plan is identified, conduct an impact propagation simulation along the relationship edges of the BIM knowledge graph to generate a multi-dimensional risk prediction vector. The multi-dimensional risk prediction vector includes at least: time deviation and cost deviation. Based on multidimensional risk prediction vectors, a multi-objective optimization model is constructed with the goal of minimizing time deviation and cost deviation. Under the boundary conditions of available resources and process logic constraints of the project, the multi-objective optimization model is solved to generate a collaborative optimization strategy that includes specific resource allocation schemes or process adjustment schemes. Implement collaborative optimization strategies, update construction schedules and resource allocation plans, and update the BIM knowledge graph simultaneously; The methods for solving multi-objective optimization models include: Initialization: Initialization is performed by One candidate collaborative optimization strategy population Initialize the state-action value function. ; Iterative solution: For each iteration Execute the NSGA-II algorithm to calculate the reward value for this iteration. ; Constructing an implicit reward function model ,in, The weighted parameter vector represents the decision-making preferences of managers. For application strategy The subsequent function for predicting the total project duration. For application strategy The subsequent prediction function for the total project cost; Based on the principle of maximum entropy, managers are determined to adopt collaborative optimization strategies. Select the final execution strategy The probability function, ; Multiple historical decision pairs based on continuous recording By solving the problem of maximizing the log-likelihood function, the optimal weight for the current stage can be calculated. : ; In generating new parental populations Then, based on the optimal weight Calculate the preference reward for this new population. , ; Preference reward With reward value Weighted combination to obtain composite reward signal : ,in, These are preset hyperparameters; Use composite reward signals Replace the original reward value Using Q-learning to update the rules for the value function Update the value of the corresponding state-action pair: ,in, For learning rate, As a discount factor, This is the state for the next iteration; Termination and Output: Repeat the iteration until the preset termination condition is met, and output a set of non-dominated solutions on the Pareto optimal front in the final population as a collaborative optimization strategy.

2. The method for predicting construction progress and resource coordination according to claim 1, characterized in that, Methods for constructing BIM knowledge graphs include: Define the preceding and following logical relationships between process nodes as directed edges carrying timing constraint information. The timing constraint information includes at least one of the following: finish-start, start-start, or finish-finish. By creating attribute edges of type "requirement" or "consumption", resource nodes are associated with process nodes, and specific requirements of the process for resources in terms of quantity, specifications, skill level, or usage time are defined on the attribute edges. Real-time dynamic data includes at least: status monitoring data of IoT devices associated with equipment resource nodes, real-name attendance data of personnel associated with human resource nodes, and material entry information associated with material resource nodes.

3. The method for predicting construction progress and resource coordination according to claim 1, characterized in that, Methods for identifying risk events that deviate from the construction plan include at least the following: Human resource availability identification: The planned demand and skill requirements of human resources nodes associated with the upcoming construction process are compared in real time with the actual attendance and skill information of the corresponding personnel in the real-time dynamic data. When the actual attendance is less than the planned demand or the skills do not match, it is identified as a human resource shortage risk event. Equipment resource availability identification: Obtain real-time status data of equipment resource nodes associated with the upcoming construction process from the Internet of Things. When the real-time status data indicates "fault", "offline" or "under maintenance", it is identified as a risk event of equipment resource unavailability. Material availability identification: Query the real-time arrival status of material resource nodes required for the upcoming construction process. When the status of at least one required material is "not present" or "delayed in transportation", it is identified as a material incomplete risk event.

4. The method for predicting construction progress and resource coordination according to claim 1, characterized in that, Methods for performing impact propagation simulations to generate multidimensional risk prediction vectors include: Starting from the process node or resource node directly associated with the risk event, recursively calculate the predicted start time and predicted completion time of all affected downstream process nodes along the preceding and following logical relationship edges between process nodes defined in the BIM knowledge graph. For each affected downstream process node, its predicted start time is compared with the planned start time in the construction plan to quantify the time deviation of that process. ; Cost deviation is quantified and calculated based on time deviation. : Identify the human or equipment resources associated with downstream processes that are waiting due to delays in upstream processes, and calculate the resource idle cost by summing them up based on their unit time cost and waiting time. If the time deviation causes the total project duration to exceed the contractually agreed completion date, the expected cost of breach of contract will be calculated according to the pre-set contract penalty clause.

5. The method for predicting construction progress and resource coordination according to claim 1, characterized in that, Methods for establishing multi-objective optimization models include: Establish a dual objective function that minimizes the predicted total project duration and the predicted total cost: ,in: For candidate collaborative optimization strategies, The solution space is the set of all valid strategies that satisfy the constraints. Define at least two constraints: Resource constraints: at any given time Any kind of resource The total usage shall not exceed the total availability of the resource: ,in, For strategy Construction tasks At any moment Resources The demand, For resources Total available quantity, This refers to any point in the project timeline. , This refers to any type of resource in the project. , For at any time A collection of all ongoing construction tasks; Logical constraints: All processes must satisfy their pre- and post-dependencies defined in the BIM knowledge graph. ,in For strategy Next preprocess The predicted completion time, For strategy Next process The predicted start time, For all processes with direct pre- or post-dependent dependencies A set of.

6. The method for predicting construction progress and resource coordination according to claim 1, characterized in that, The NSGA-II algorithm includes the following steps: a) State Recognition and Action Selection: The initial risk prediction vector that triggered this optimization is taken as the current state. And based on the value function use Strategy: Select an action ,action Corresponding to the preset mutation operator preference, the preference includes at least prioritizing shortening the construction period, prioritizing cost reduction, or balanced optimization; b) Genetic manipulation: on the current population Tournament selection and simulation of binary crossover operators are employed, and actions are based on... Based on the corresponding mutation preference, an adaptive polynomial mutation operation is performed to generate the offspring population. ; c) Environmental assessment and elite selection: This involves the parent population... With offspring population Merge, perform fast non-dominated sorting and crowding calculation, and select from them. The best individuals form a new parent population. ; Reinforcement learning model update: based on the newly generated parent population Compared to The reward value for this iteration is calculated based on the increment of the Pareto front hypervolume index. And using the following Q-learning update rule, the value function is updated. Update the value of the corresponding state-action pair.

7. The method for predicting construction progress and resource coordination according to claim 6, characterized in that, Methods for performing closed-loop updates include: Select a final execution strategy from the collaborative optimization strategies, and parse the final execution strategy to obtain adjustment instructions. The adjustment instructions shall at least include adjustment operations on process, human resources or equipment resources. The construction schedule and resource allocation plan are directly updated based on the adjustment instructions. The update content includes at least adjusting the planned start and end times of the affected work processes or changing the allocation relationship between work processes and resources. The updated content in the construction schedule and resource allocation plan is synchronously reflected back into the BIM knowledge graph, updating the time attributes of the corresponding process nodes in the BIM knowledge graph, as well as updating the association edges or attributes between process nodes and resource nodes. Set the updated BIM knowledge graph as the new baseline plan, continue real-time monitoring, and identify new risk events.

8. A predictive data processing system for construction progress and resource coordination, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the predictive data processing method for construction progress and resource coordination as described in any one of claims 1-7.

9. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the predictive data processing method for construction progress and resource coordination as described in any one of claims 1-7.

Images