CN-122022447-A - City building civil engineering project construction progress delay risk early warning method

CN122022447ACN 122022447 ACN122022447 ACN 122022447ACN-122022447-A

Abstract

The invention provides a delay risk early warning method for construction progress of urban building civil engineering projects, which comprises the steps of multidimensional data cleaning alignment, automatic construction risk knowledge graph construction and topology optimization, dynamic causal path identification and expert rule calibration, reverse reasoning and anti-fact intervention simulation of a structural equation, visual sequencing of sensitive risk factors and the like.

Inventors

LUO HUIFENG
FENG XINXING
LIANG JIAJUN
HUANG XIAOBIN
CHEN MUYI
ZHANG FANGYAN

Assignees

广州市白云城市建设投资有限公司

Dates

Publication Date: 20260512
Application Date: 20251219

Claims (10)

1. The urban building civil engineering project construction progress delay risk early warning method is characterized by comprising the following steps of: S1, acquiring plan progress, actual progress, resource investment, weather conditions, supply chain states and on-site event log data of urban building civil engineering projects, and executing data cleaning and time stamp alignment operations to generate a structured time sequence data set; S2, constructing an initial risk knowledge graph based on the structured time sequence data set, and calculating and generating a node edge weight attribute value through statistical association degree; S3, constructing a dynamic causal discovery algorithm model based on the multi-source heterogeneous data characteristics aiming at the initial risk knowledge graph, executing non-stationary causal dependency detection among variables by adopting a sliding time window mechanism, and outputting a causal intensity matrix; s4, performing pattern matching on the causal strength matrix and a preset construction management expert rule base, executing causal path weight attenuation processing violating engineering common sense, and generating a calibrated dynamic causal map; S5, constructing a counterfactual intervention analysis model based on the dynamic causal map, and simulating an influence propagation path of key variable intervention on a construction period by adopting a structural equation modeling method to generate multidimensional intervention strategy response curved surface data; S6, performing critical path sensitivity sequencing calculation according to the multidimensional intervention strategy response curved surface data, and generating a visual early warning report containing dominant causal paths, risk driving factors and inverse facts sensitivity indexes thereof; And S7, performing association mapping on the visual early warning report and the real-time progress data, constructing a dynamic risk scoring function based on causal reasoning, and outputting a delay risk quantification value with a traceable logic chain.
2. The method for early warning of delay risk of construction progress of civil engineering projects of urban building according to claim 1, wherein the step S1 specifically comprises: acquiring the plan starting time, the plan ending time and the expected construction period of a construction task item based on the construction plan data of the urban building civil engineering project management system, and forming a project plan progress benchmark data set; The on-site construction progress monitoring equipment and the manual filling system are abutted, and the actual starting time, the actual completion state and the current progress percentage of each construction task item are collected to form an item actual progress dynamic data stream; Acquiring resource allocation data required by construction based on a project resource management system to form a resource input state data set; The method comprises the steps of accessing a meteorological data interface and a field environment monitoring sensor, acquiring weather condition data during construction, and constructing an environment influence factor time sequence; Docking the supply chain management system with the logistics tracking platform, acquiring a material purchasing plan, the delivery time of a supplier, the transportation state and the on-site receiving record, and generating a supply chain state real-time updating data set; collecting a construction site event log to form a project management behavior event sequence; Performing data cleaning operation on the project plan progress reference data set, the project actual progress dynamic data stream, the resource input state data set, the environment influence factor time sequence, the supply chain state real-time update data set and the project management behavior event sequence to obtain cleaned multi-source data; Performing timestamp alignment operation on the cleaned multi-source data based on a unified time reference to generate a structured data table with unified time granularity; and organizing the structured data table according to construction task items, resource allocation, environmental factors and management behavior entity dimensions to generate a standardized structured time sequence data set.
3. The method for early warning of delay risk of construction progress of civil engineering projects of urban building according to claim 2, wherein the resource input state data set is pulled in batches through an project resource management system API to obtain resource configuration data of manpower, machinery and materials required by construction, and the resource configuration data comprises resource types, approach time, use states and consumption amounts and is converted into a resource input state time schedule through field integrity verification and standardization.
4. The method for early warning of delay risk of construction progress of civil engineering projects of urban building according to claim 1, wherein the step S2 specifically comprises: Performing entity identification and type division processing on the standardized structured time sequence data set, extracting four construction risk related entities of task items, resource allocation, environmental factors and management behaviors, and generating a standardized entity identifier set; modeling entity interaction relations in multi-source data by adopting an entity co-occurrence frequency statistical method based on the standardized entity identifier set, and generating a potential association relation set between entities as a candidate edge set of an initial knowledge graph; Performing joint calculation of the Pierson correlation coefficient and the Spilot scale correlation coefficient on the candidate edge set to quantify the linear and nonlinear statistical correlation strength between entities and generate an initial correlation weight matrix; pruning weak correlation edges by adopting a threshold filtering strategy based on the initial correlation weight matrix, and reserving significant correlation paths to form a filtered correlation edge set; Performing directionality inference processing on the screened associated edge sets, and distributing preliminary direction attributes for each edge based on causal sequence features in time sequence data and a grange causal inspection method to generate an initial risk knowledge graph; and constructing a graph structure storage model by adopting a graph database technology based on the initial risk knowledge graph to form a graph storage structure.
5. The method for early warning of delay risk of construction progress of civil engineering projects of urban building according to claim 1, wherein the step S3 specifically comprises: Based on the output initial risk knowledge graph, extracting time sequence attribute data of multiple types of nodes, and denoising the node attribute sequence by using a time sequence smoothing filter algorithm to obtain denoised node attribute time sequence data; constructing a causal structure learning model for the denoising node attribute time sequence data based on a dynamic causal discovery algorithm in a non-stationary environment, setting a significance threshold value of causal relation inference among variables and a conditional independence test criterion, and identifying potential causal paths in different time windows; based on the causal structure learning model, carrying out sectional processing on the full life cycle data of the construction project by adopting a sliding time window mechanism, executing causal graph structure learning and parameter estimation in each window, and outputting causal graph structures of each time window; carrying out causal edge intensity quantization calculation on the causal graph structures of each time window, and generating a causal intensity matrix based on the statistical frequency and the directional consistency of causal influence propagation paths; and mapping and aligning the causal strength matrix with the statistical association edges in the initial risk knowledge graph, and identifying and marking a difference region between a causal path and an association path.
6. The method for early warning of delay risk of construction progress of civil engineering projects of urban building according to claim 1, wherein the step S4 specifically comprises: based on the output causal strength matrix and a preset construction management expert rule base, executing causal path pattern matching processing, and identifying an abnormal causal relation contrary to the common engineering knowledge; carrying out correction processing on the identified abnormal causal path by adopting a causal edge weight attenuation algorithm based on a rule confidence score, and generating a corrected causal intensity matrix; constructing a causal structure confidence map based on the modified causal intensity matrix; performing graph pruning operation on the causal structure confidence map to obtain a causal structure confidence map after pruning; and generating a calibrated dynamic causal map based on the pruned causal structure confidence map.
7. The method for early warning of delay risk of construction progress of civil engineering projects of urban building according to claim 6, wherein the step S4 further comprises adopting a causal strength threshold and confidence level combined screening mechanism, and further eliminating causal edges with insufficient directional stability through path direction consistency detection.
8. The method for early warning of delay risk of construction progress of civil engineering projects of urban building according to claim 1, wherein the step S5 specifically comprises: Based on the output calibration dynamic causal map, extracting a key node variable set, and constructing a structural equation model frame by utilizing a causal map structure; Performing parameter estimation and model fitting optimization on the structural equation model framework by combining the structural time sequence data set generated in the step S1, and calibrating path coefficients by adopting a generalized moment estimation method; Based on the optimized structural equation model, setting a typical intervention variable set, executing a counterfactual intervention simulation, and setting an intervention variable value range and a step length; Performing forward propagation calculation on each set of set intervention variable values based on a structural equation model, simulating the influence of intervention on the construction period of the key path nodes, and outputting probability distribution and confidence intervals of the total construction period of the project under each intervention condition; Based on the output multiple sets of intervention response results, a construction period response surface is constructed, response surface fitting is carried out by adopting a polynomial regression and Gaussian process modeling method, and an intervention-response mapping relation surface is generated.
9. The method for early warning of delay risk of construction progress of civil engineering works of city building according to claim 8, wherein the typical intervention variable set comprises resource allocation intensity, procedure starting time and external environment disturbance factor.
10. The method for early warning of delayed risk of construction progress of civil engineering projects of urban building according to claim 1, wherein the visual early warning report comprises dominant causal paths, risk driving factors and anti-facts sensitivity indexes thereof.

Description

City building civil engineering project construction progress delay risk early warning method Technical Field The invention relates to the technical field of building engineering risk management, in particular to a delay risk early warning method for construction progress of urban building civil engineering projects. Background In the construction process of urban building civil engineering projects, early warning and management of progress delay risks are always an important technical subject for engineering digitization and intelligent transformation. The current mainstream construction risk modeling and progress prediction technology mainly adopts methods such as statistical regression, time sequence analysis, machine learning, graphic neural network and the like, and realizes automatic identification and quantitative scoring of project delay risks by modeling past historical data and real-time monitoring data. The development trend of the method pays attention to the prediction precision and the data automatic acquisition capability of the model, effectively improves the efficiency and timeliness of risk identification of the construction site, and gradually evolves towards the multi-source data fusion and intelligent decision direction; in the prior art, the method is operated in a black box mode in practical application, and a representative scheme of the method comprises a progress risk prediction system based on statistical correlation and machine learning modeling, a risk labor algorithm for carrying out association mining on construction parameters by utilizing a graph neural network, and an entity matching and risk attribution tool based on a static knowledge graph. The technologies can complete automatic monitoring and risk prompting of construction progress to a certain extent, and are suitable for risk discrimination, preliminary attribution and coping proposal of civil engineering projects of multiple cities; However, the existing solution has the following limitations and technical defects that firstly, a mainstream risk modeling method highly depends on a data-driven statistical correlation or machine learning black box model, deep description of causal relation among risk events is lacking, an inherent logic chain of risk formation is difficult to reveal, secondly, although the model can realize accurate prediction according to history and current data, a model decision process cannot be explained, actual requirements of project management personnel on risk tracing, causal attribution and scheme deduction cannot be met, early warning output is difficult to be directly trusted or used for formulating a targeted plan, and furthermore, static knowledge graph technology taking a graph neural network as a dominant cannot effectively cope with the characteristics of dynamic change of multi-source data of a construction site and causal relation time change under a non-stationary environment, and the dynamic adaptability of causal reasoning of the risk events is lacking. Some existing GANs generate enhanced paths and static map matching schemes, have natural defects in the aspects of interpretability and engineering semantic compliance, and cannot perform inverse fact reasoning and sensitivity analysis on key causal links in the project progress process. Disclosure of Invention The invention provides a delay risk early warning method for the construction progress of urban building civil engineering projects, which aims to solve the technical problems. The technical scheme of the invention is realized in such a way that the construction progress delay risk early warning method for the civil engineering project of the urban building comprises the following steps: S1, acquiring plan progress, actual progress, resource investment, weather conditions, supply chain states and on-site event log data of urban building civil engineering projects, and executing data cleaning and time stamp alignment operations to generate a structured time sequence data set; s2, constructing an initial risk knowledge graph containing task items, resource allocation, environmental factors and management behavior entities based on the time sequence data set output by the S1, and calculating and generating inter-node edge weight attribute values through statistical association; S3, constructing a dynamic causal discovery algorithm model based on the multi-source heterogeneous data characteristics aiming at the initial knowledge graph generated in the S2, executing non-stationary causal dependency detection among variables by adopting a sliding time window mechanism, and outputting a causal intensity matrix; S4, performing pattern matching on the causal strength matrix generated in the S3 and a preset construction management expert rule base, executing causal path weight attenuation processing violating engineering common sense, and generating a calibrated dynamic causal map; S5, constructing a counterfactual intervention analy