CN-121981688-A - Building engineering management system and method based on big data

CN121981688ACN 121981688 ACN121981688 ACN 121981688ACN-121981688-A

Abstract

The invention relates to the technical field of large data management of constructional engineering, in particular to a large data-based constructional engineering management system and a large data-based constructional engineering management method, comprising the following steps: and receiving full-period engineering data of a target building engineering project, performing space-time alignment and feature fusion operation on the data, integrating the design feature flow, the resource feature flow, the process feature flow and the quality feature flow which are mutually related, and constructing a multidimensional engineering state tensor representing the coupling relation of the multi-working-element elements. Inputting tensors into a deep learning engineering state reasoning network, iteratively identifying potential risk paths and simulating state evolution to generate an engineering state evolution directed graph, deducing a multi-strip-shaped track from a current construction stage to a future target node according to the engineering state evolution directed graph, and generating an engineering management decision analysis report through track analysis. The method realizes the associative integration of engineering data and the intelligent evolution deduction of engineering states, and optimizes the data analysis and decision support form of building engineering management.

Inventors

LIU ZHENFAN
Lin Zhangji
Fan Haizheng
DONG QIULIN
Zhu Tengjie

Assignees

福建水利电力职业技术学院

Dates

Publication Date: 20260505
Application Date: 20260409

Claims (10)

1. The construction engineering management method based on big data is characterized by comprising the following steps: Receiving full-period engineering data of a target building engineering project; Performing space-time alignment and feature fusion operation on the full-period engineering data, and integrating a design feature stream, a resource feature stream, a process feature stream and a quality feature stream which are mutually related; Constructing a multi-dimensional engineering state tensor based on the integrated design feature flow, the integrated resource feature flow, the integrated process feature flow and the integrated quality feature flow, wherein the multi-dimensional engineering state tensor is used for representing the coupling relation among engineering entities, construction activities, resource consumption and quality indexes; inputting the multidimensional engineering state tensor into an engineering state reasoning network based on deep learning, and generating an engineering state evolution directed graph by iteratively identifying potential risk paths and simulating state evolution by the engineering state reasoning network based on the deep learning; Forward pushing a multi-strip track from a current construction stage to a future target node according to the engineering state evolution directed graph; And analyzing the multi-strip-shaped track to generate an engineering management decision analysis report.
2. The big data based construction engineering management method of claim 1, wherein performing a space-time alignment and feature fusion operation on the full-period engineering data integrates a design feature stream, a resource feature stream, a process feature stream and a quality feature stream which are associated with each other, comprising: The whole period engineering data comprise design drawing information, material purchase records, construction progress logs, personnel attendance information, equipment operation parameters and quality detection reports; Carrying out component identification and space relation analysis on the design drawing information, extracting geometric features, topological relations and design parameter sequences of the design drawing information, and forming the design feature flow; Normalizing the material purchase record, the personnel attendance information and the equipment operation parameters, and mapping the normalized material purchase record, the personnel attendance information and the equipment operation parameters to uniform space positions and time coordinates to form the resource feature stream; performing process decomposition and activity coding on the construction progress log, constructing a construction activity sequence map, and extracting a time sequence dependency mode from the construction activity sequence map to form the process characteristic stream; Performing defect identification and index quantification on the quality detection report, and converting qualitative description into quantitative quality state vectors to form the quality characteristic stream; Establishing a space-time alignment layer, and calibrating the design feature stream, the resource feature stream, the process feature stream and the quality feature stream according to uniform space position coordinates and time stamps; And carrying out deep fusion on the design feature stream, the resource feature stream, the process feature stream and the quality feature stream which are subjected to space-time alignment by adopting a feature level fusion strategy, and generating a fused engineering feature representation as the multidimensional engineering state tensor.
3. The big data based construction project management method of claim 2, wherein inputting the multi-dimensional project state tensor into a deep learning based project state inference network, the deep learning based project state inference network generating a project state evolution directed graph by iteratively identifying potential risk paths and modeling state evolution, comprising: Initializing a space-time diagram convolution-based encoder network, and encoding the multi-dimensional engineering state tensor to obtain an initial hidden representation of the engineering state; connecting a serialized state transition prediction network, wherein the serialized state transition prediction network uses engineering states at the previous moment as inputs, and predicts various engineering states and transition probabilities thereof at the next moment; In the prediction process, a state transition path with high probability is explored through a Monte Carlo tree search algorithm, and a path with low probability or high risk is pruned; and integrating the explored state transition paths and the probabilities thereof to form a weighted directed graph structure, namely the engineering state evolution directed graph, wherein nodes in the graph represent discretized engineering states, and the edge weights represent the probabilities of state transition.
4. A method of building engineering management based on big data according to claim 3, wherein forward-pushing the multi-strip trajectory from the current construction stage to the future target node according to the engineering state evolution directed graph, comprises: marking a node representing a final acceptance condition of the project as a target node in the engineering state evolution directed graph; using a node representing the current actual construction state as a starting point, and traversing the engineering state evolution directed graph along the forward direction of the directed edge by adopting a heuristic graph searching algorithm; In the traversal process, recording a complete node sequence and an edge sequence which are passed by the progress from a starting node to the target node, wherein each complete search path forms a state track; and checking whether the contained working procedure logic of each state track accords with the standard construction process flow of the building engineering, and eliminating paths with logic conflict.
5. The method of claim 4, wherein analyzing the multi-strip trajectory to generate an engineering management decision analysis report comprises: calculating a comprehensive cost index and a time feasibility score of each state track, and sorting and screening all the state tracks according to the comprehensive cost index and the time feasibility score; Integrating all the state tracks after sequencing and screening to form a multi-level engineering state evolution network, wherein each node in the multi-level engineering state evolution network represents a specific engineering state combination, and each side represents a state transition relation; Community division is carried out on the multi-level engineering state evolution network, a complex network structure is decomposed into a plurality of construction stage clusters, and each construction stage cluster focuses on the construction process of one core; Performing cross-process coupling influence analysis for each of the construction stage clusters, quantifying a sensitivity coefficient of each of the engineering state variables to a final project objective; Based on the sensitivity coefficient of each engineering state variable, a structured engineering management decision analysis report is generated in combination with a building engineering specification knowledge base.
6. The method of claim 5, wherein calculating a comprehensive cost index and a time feasibility score for each of the status tracks comprises: For one state track, extracting estimated resource consumption, risk cost and time consumption data corresponding to all state transition edges on the state track; The estimated resource consumption, the risk cost and the time consumption data are weighted and summed according to a preset weight coefficient to obtain the comprehensive cost index; analyzing the accumulated value of the time required by each construction activity on the state track to obtain the total estimated construction period of the state track; Comparing the total estimated construction period with the project planning construction period, and calculating the satisfaction degree of the project planning construction period by considering the forced time interval constraint among the working procedures to be used as the time feasibility score; Integrating all the state tracks after sequencing and screening to form a multi-level engineering state evolution network, which specifically comprises the following steps: Combining all state tracks meeting a preset comprehensive cost index threshold and a time feasibility score threshold; In the merging process, identifying key engineering state nodes shared among different state tracks; Weaving the multi-strip state track into a directed acyclic mesh topology by taking the key engineering state nodes as hinges; According to the engineering stage corresponding to the nodes, the nodes in the mesh topology are divided into a basic construction layer, a main construction layer, a decoration construction layer and an acceptance layer, so that the multi-level engineering state evolution network with a definite time level is formed.
7. The big data based construction engineering management method of claim 6, wherein the community division is performed on the multi-level engineering state evolution network, and the complex network structure is decomposed into a plurality of construction stage clusters, comprising: Calculating the state transition similarity among all nodes in the multi-level engineering state evolution network; Dividing the nodes into different communities according to the state transition similarity by adopting a modularity optimization algorithm, wherein the transition relationship among the nodes in each community is tight, and the transition relationship among the different communities is sparse; extracting each community and the edges connected with the communities inside the communities to form an independent sub-network, wherein the independent sub-network is the construction stage cluster; and marking the corresponding main construction subsection project names for each construction stage cluster.
8. The big data based construction project management method of claim 7, wherein said performing cross-process coupling impact analysis for each of said construction stage clusters quantifies a sensitivity coefficient of each of the project state variables to a final project objective, comprising: defining key performance index nodes directly related to a final project target as influence target nodes in one construction stage cluster; calculating the gradient of the numerical variation of each engineering state variable node in the construction stage cluster on the numerical variation of the affected target node by using a back propagation algorithm based on the gradient; Normalizing the calculated gradient value, and taking the absolute value as an initial influence measure of the engineering state variable node on the influence target node; And correcting the initial influence force by combining the fluctuation range of the engineering state variable in the historical data to obtain the final sensitivity coefficient.
9. The method of claim 8, wherein said generating a structured report of engineering management decision analysis based on said sensitivity coefficients for each engineering state variable in combination with a knowledge base of engineering specifications comprises: marking engineering state variables of which the sensitivity coefficients exceed a preset threshold as key control variables; retrieving construction specification terms, quality control criteria and recommended regulatory measures associated with each key control variable from the construction specification knowledge base; Arranging the key control variables in descending order of the sensitivity coefficient; For each key control variable, detail listing the construction stage cluster to which the key control variable belongs, the current monitoring value, the sensitivity coefficient, the associated specification clause and the recommended regulation measure parameter; and carrying out formatting filling on all the information according to a preset management report template to generate a final structured engineering management decision analysis report.
10. A big data based construction project management system comprising a memory, a processor and a computer program stored in the memory and running on the processor, characterized in that the processor, when executing the computer program, realizes the steps of a big data based construction project management method according to any of the preceding claims 1 to 9.

Description

Building engineering management system and method based on big data Technical Field The invention relates to the technical field of large data management of constructional engineering, in particular to a large data-based constructional engineering management system and method. Background The existing construction engineering management field processes project full-period engineering data by adopting a mode of independent collection and single dimension statistics by a separate module, respectively stores and sorts design information, resource allocation, construction process, quality detection and other data, and completes basic information verification by means of a conventional data processing mode without executing unified time-space alignment and feature fusion processing on multiple types of engineering data. The correlation matching mechanism is lacking among various engineering data, the design, resource, process and quality related data cannot form a characteristic flow system which is correlated with each other, a multidimensional engineering state tensor which can represent the coupling relation among engineering entities, construction activities, resource consumption and quality indexes cannot be constructed, and the overall representation of the engineering state is lacking. In the prior art, the deep learning network is not used for carrying out engineering state reasoning related work, potential risk paths in engineering construction links cannot be identified through iterative operation, the evolution process of an engineering state cannot be simulated, an engineering state evolution directed graph is difficult to generate, and forward deduction of multiple-strip-shaped tracks from the current construction stage to a future target node cannot be realized. The engineering management related analysis only stays at the static data carding level, and cannot form decision analysis basis for dynamic development of the fitting engineering. Feature flow integration and multidimensional tensor construction are needed to be carried out on the full-period engineering data, and meanwhile risk path identification, state evolution simulation and multi-track deduction are completed through a deep learning network, so that a corresponding engineering management decision analysis report is generated. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides a building engineering management system and method based on big data. In order to achieve the purpose, the invention adopts the following technical scheme that the building engineering management method based on big data comprises the following steps: Receiving full-period engineering data of a target building engineering project; Performing space-time alignment and feature fusion operation on the full-period engineering data, and integrating a design feature stream, a resource feature stream, a process feature stream and a quality feature stream which are mutually related; Constructing a multi-dimensional engineering state tensor based on the integrated design feature flow, the integrated resource feature flow, the integrated process feature flow and the integrated quality feature flow, wherein the multi-dimensional engineering state tensor is used for representing the coupling relation among engineering entities, construction activities, resource consumption and quality indexes; inputting the multidimensional engineering state tensor into an engineering state reasoning network based on deep learning, and generating an engineering state evolution directed graph by iteratively identifying potential risk paths and simulating state evolution by the engineering state reasoning network based on the deep learning; Forward pushing a multi-strip track from a current construction stage to a future target node according to the engineering state evolution directed graph; And analyzing the multi-strip-shaped track to generate an engineering management decision analysis report. As a further aspect of the present invention, performing a space-time alignment and feature fusion operation on the full-period engineering data, integrating a design feature stream, a resource feature stream, a process feature stream, and a quality feature stream that are associated with each other, including: The whole period engineering data comprise design drawing information, material purchase records, construction progress logs, personnel attendance information, equipment operation parameters and quality detection reports; Carrying out component identification and space relation analysis on the design drawing information, extracting geometric features, topological relations and design parameter sequences of the design drawing information, and forming the design feature flow; Normalizing the material purchase record, the personnel attendance information and the equipment operation parameters, and mapping the normalized material purchase record, the per