Construction technical scheme compilation adaptive optimization method and system based on multi-source data

Abstract

The invention discloses a construction technical scheme compilation adaptive optimization method and system based on multi-source data, and relates to the field of construction management.The method comprises the steps that all kinds of data in the whole construction period are collected through a distributed terminal, and a dynamic data resource pool is constructed after standardization processing; cleaning data by using a classification preprocessing model, and extracting a construction scheme associated feature set through cosine similarity and deep learning; generating an initial scheme by combining case reasoning and rule reasoning, and performing multi-target dynamic optimization by applying reinforcement learning; and verifying the feasibility of the scheme through virtual simulation and trial construction, finally establishing a real-time monitoring feedback mechanism, and automatically triggering scheme updating when actual data deviates from a threshold value. The method has the advantages that multi-source data full-process driving is adopted, a scheme is generated through preprocessing, feature extraction and hybrid reasoning, reinforcement learning optimization, virtual simulation verification and a real-time deviation updating mechanism are combined, accurate adaptation and dynamic optimization of a construction technical scheme are achieved, and engineering comprehensive benefits are efficiently improved.

Description

Technical Field

[0001] This invention relates to the field of construction management, and in particular to an adaptive optimization method and system for developing construction technology solutions based on multi-source data. Background Technology

[0002] As modern engineering projects become increasingly complex and large-scale, the uncertainty of the construction environment increases. Fixed solutions struggle to dynamically respond to changes on-site, easily leading to inefficiency or cost overruns. Meanwhile, the maturity of technologies such as BIM, IoT, and AI provides the conditions for real-time data collection and intelligent analysis. Adaptive optimization methods integrate multi-source information and utilize algorithms to dynamically adjust construction techniques, schedules, and resource allocation, achieving continuous optimization of the solution during implementation.

[0003] Current methods for developing construction technical solutions struggle to achieve accurate and efficient adaptation across the entire process. Most methods still rely on manual experience, with fragmented and limited data collection dimensions. They lack a comprehensive distributed data collection network and standardized processing mechanisms, resulting in inconsistent data quality and failing to provide reliable support for solution development. Solution generation often depends on single cases or rule-based reasoning, failing to dynamically mine core correlation features from multiple data sources, leading to insufficient targeting and scientific rigor in initial solutions. Optimization is mostly static, one-off adjustments, lacking dynamic adaptive mechanisms based on reinforcement learning. This makes it difficult to balance multiple objectives such as quality, schedule, and cost, and candidate solution selection lacks a systematic and scientific evaluation framework. Furthermore, the absence of a closed-loop update mechanism for virtual simulation verification and real-time monitoring feedback results in a significant disconnect between the solution and actual site conditions. This leads to delayed responses to changes or deviations in the construction environment, easily causing rework, cost overruns, and safety risks, making it unsuitable for the dynamic management needs of complex projects. Summary of the Invention

[0004] To improve existing methods and systems, this paper provides an adaptive optimization method and system for construction technology scheme development based on multi-source data. This method is driven by multi-source data throughout the entire process. Through intelligent preprocessing, feature extraction, and hybrid inference to generate schemes, combined with reinforcement learning optimization, virtual simulation verification, and real-time deviation update mechanism, it achieves accurate adaptation and dynamic optimization of construction technology schemes, and effectively improves the overall benefits of the project.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An adaptive optimization method for construction technology solutions based on multi-source data is developed, including: Multi-source heterogeneous data throughout the entire lifecycle of a construction project are collected through distributed data acquisition terminals, including basic environmental data, engineering design data, construction resource data, historical engineering data, and real-time construction monitoring data. After the collected raw data is standardized and packaged, it is stored in a dynamic data resource pool. A classification preprocessing model is built based on data type differences to perform hierarchical processing on data in the dynamic data resource pool and obtain a high-quality standardized dataset. The correlation between different data in the standardized dataset is calculated by cosine similarity, the weight coefficients of each data source are obtained, the core correlation features of construction plan preparation are extracted by deep learning model, and after fusion processing, a construction plan correlation feature set is formed. A construction scheme generation model is constructed based on a hybrid reasoning mechanism that combines case-based reasoning and rule-based reasoning. The model takes the feature set associated with the construction scheme as input and combines it with the core requirements of the construction project to generate an initial construction technical scheme. An adaptive optimization model for construction technology schemes is constructed based on reinforcement learning theory. A multi-objective optimization function is established, and real-time updated data and associated feature sets are input. By dynamically adjusting the optimization variables, multiple sets of optimization candidate schemes are generated, and the optimal construction technology scheme is selected using the analytic hierarchy process. A virtual construction simulation platform was built, the optimal construction technology solution was imported, and a simulation model was established based on the actual environment and design data to simulate the entire construction process. By carrying out small-scale trial construction, the real-time collected data and simulation results were compared to evaluate the suitability of the solution. In actual implementation, real-time monitoring data is compared with preset parameters in the optimization plan. When the deviation between the monitoring data and the preset parameters exceeds the threshold, the plan update mechanism is triggered to generate an updated construction technology plan.

[0006] Preferably, the step of collecting multi-source heterogeneous data throughout the entire lifecycle of a construction project via a distributed data acquisition terminal, including basic environmental data, engineering design data, construction resource data, historical project data, and real-time construction monitoring data, and then storing the collected raw data in a dynamic data resource pool after standardizing and encapsulating the data, specifically includes: Multiple types of data acquisition terminals are deployed at key nodes, equipment operating surfaces, and surrounding environments in the construction area to build a distributed data acquisition network covering the entire area and collect data synchronously. Collect multi-source heterogeneous data in a categorized manner, including basic environmental data, engineering design data, construction resource data, historical engineering data, and real-time monitoring data; All types of raw data are packaged in a unified format, stored in a dynamic resource pool, and a real-time update trigger mechanism is set.

[0007] Preferably, the step of constructing a classification preprocessing model based on data type differences to perform hierarchical processing on the data in the dynamic data resource pool and obtain a high-quality standardized dataset specifically includes: Construct a classification preprocessing model, divide the processing levels based on data types, and establish a mapping relationship between data types and processing strategies; A sliding window smoothing algorithm is used to eliminate random noise in the basic environmental data, and missing data is filled in by trend fitting. Perform syntax validation and logical consistency checks on engineering design data, remove redundant data, and correct error terms; An outlier detection algorithm is used to identify and remove outlier data from construction resource data, and normalization processing is used to unify the dimensions. Natural language processing technology is used to extract keywords and perform structured transformation on historical engineering data; Data downsampling and time-series alignment are applied to real-time construction monitoring data to obtain a high-quality standardized dataset.

[0008] Preferably, the step of calculating the correlation between different data in the standardized dataset using cosine similarity, obtaining the weight coefficients of each data source, extracting the core correlation features of the construction plan compilation using a deep learning model, and forming a construction plan correlation feature set after fusion processing specifically includes: The standardized dataset is classified and labeled with data sources, and the association strength between different data sources is calculated using cosine similarity. The weights are assigned based on the correlation strength between historical engineering data and current project cases. The weights of real-time monitoring data are adjusted in a stepwise manner based on the data update frequency. Static data is assigned a fixed baseline weight, forming a dynamic weight allocation scheme. By fusing data from different dimensions at the feature level, and extracting the core related features for construction plan preparation through a deep learning model; Redundancy removal is performed on the fused feature data to form a structured set of construction scheme related features.

[0009] Preferably, the construction scheme generation model, which uses a hybrid reasoning mechanism combining case-based reasoning and rule-based reasoning, inputs a set of associated features of the construction scheme and, in conjunction with the core requirements of the construction project, generates an initial construction technical scheme, specifically including: A construction scheme generation model is constructed, and a hybrid reasoning mechanism combining case reasoning and rule reasoning is adopted. Based on the core correlation features, the most similar case schemes are retrieved from historical engineering data, and the key construction technology, process arrangement and parameter settings in the case are extracted. By combining current construction specifications, industry standards, and project-specific requirements, a reasoning rule base is constructed. The retrieved case solutions are adaptively adjusted to determine the construction process, construction methods for each process, equipment selection scheme, material usage plan, and personnel division of labor details, thereby generating an initial construction technical solution that meets the basic requirements of the project.

[0010] Preferably, the adaptive optimization model for construction technology solutions based on reinforcement learning theory, the establishment of a multi-objective optimization function, the input of real-time updated data and associated feature sets, the generation of multiple sets of optimization candidate solutions through dynamic adjustment of optimization variables, and the use of the analytic hierarchy process (AHP) to screen the optimal construction technology solution specifically include: An adaptive optimization model for construction technology solutions is constructed based on reinforcement learning theory, and a multi-objective optimization function is built with construction quality compliance rate, construction period completion efficiency, cost control accuracy and safety risk level as optimization objectives. The construction process parameters, resource allocation ratio, and process connection time in the initial construction technical plan are used as optimization variables. Based on the model-based intelligent decision-making mechanism, variables are dynamically adjusted and optimized in combination with real-time data feedback to generate multiple sets of differentiated candidate solutions; The analytic hierarchy process (AHP) is used to comprehensively evaluate multiple candidate solutions and select the optimal solution. If the optimal solution does not meet the preset optimization target threshold, the model optimization strategy is adjusted and the optimization process is repeated iteratively.

[0011] Preferably, the construction of the virtual construction simulation platform, the import of the optimal construction technology solution, the establishment of a simulation model based on the actual environment and design data, the simulation of the entire construction process, and the evaluation of the suitability of the solution by comparing the real-time collected data with the simulation results through small-scale trial construction specifically include: A virtual construction simulation platform is built, which, based on the collected basic environment and engineering design data, constructs a simulation model that is consistent with the actual construction scenario. The optimized construction technology solution is imported into the simulation model to simulate the entire construction process, and the simulation results include the construction period simulation results, quality achievement nodes, and resource consumption curves. Select a segment of the project process as the trial construction area, collect construction data in real time, and compare it with the simulation results dimension by dimension to obtain the causes of deviations; An evaluation system is constructed based on technical feasibility, economic rationality, and safety controllability to generate feasibility assessment results. If there are infeasible items, the system is iterated and optimized again.

[0012] Preferably, during actual execution, the real-time monitoring data is compared with the preset parameters in the optimization scheme. When the deviation between the monitoring data and the preset parameters exceeds a threshold, a scheme update mechanism is triggered to generate an updated construction technical scheme. This specifically includes: During the actual implementation of the optimized construction technology plan, data on the progress of process completion, component construction quality inspection, material and equipment consumption, and changes in the on-site environment are collected simultaneously. The real-time monitoring data is compared with the preset parameters of the optimization scheme dimension by dimension. By setting a deviation threshold, when the deviation value between the monitoring data and the preset parameters exceeds the threshold, the update process is triggered. The latest data in the dynamic data resource pool is called, the feature fusion model is reused, the bias-related features are re-extracted, and an updated feature set is generated. The updated feature set is input into the adaptive optimization model, and local parameter corrections or global process optimizations are performed based on the deviation type to generate an updated construction technology solution.

[0013] Furthermore, an adaptive optimization system for construction technology solutions based on multi-source data is proposed, including: Data acquisition and management module: Real-time acquisition of multi-source heterogeneous data throughout the entire construction cycle, which is then standardized, packaged, and stored in a dynamic data resource pool; Data preprocessing module: Based on a classification model, heterogeneous data is processed in a hierarchical manner to output a standardized dataset; Feature fusion and weight calculation module: Dynamically allocates data source weights using cosine similarity, extracts core related features of construction schemes through deep learning, and generates a structured feature set; Solution generation module: Generates initial construction solutions by combining case-based reasoning and rule-based reasoning; Reinforcement learning optimization module: Constructs a multi-objective function with quality, schedule, cost, and safety as objectives, dynamically adjusts parameters to generate candidate solutions, and selects the optimal solution through the analytic hierarchy process; Virtual simulation module: Simulates the entire construction process, evaluates the feasibility of the plan by comparing trial construction data, and provides feedback on optimization needs; Update and Execution Module: Monitors construction deviations in real time, triggers a scheme update mechanism when the deviation exceeds a threshold, and generates a corrected scheme based on the latest data; Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

[0014] Compared with the prior art, the advantages of the present invention are: This system enables data-driven and dynamically adapted construction plans from development to execution. Distributed data acquisition and hierarchical preprocessing ensure the comprehensiveness and high quality of multi-source heterogeneous data, laying a reliable data foundation for plan development. Cosine similarity weighting and deep learning feature extraction accurately capture core related elements, and a hybrid reasoning mechanism enhances the relevance and scientific rigor of the initial plan. Relying on reinforcement learning multi-objective optimization and hierarchical analysis screening, it generates the optimal plan while considering multiple dimensions such as quality and schedule. Virtual simulation and trial construction verification further ensure the feasibility of the plan. Furthermore, real-time monitoring and deviation-triggered update mechanisms enable dynamic optimization throughout the plan's lifecycle, effectively improving construction adaptability, decision-making efficiency, and overall project benefits while reducing construction risks. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the method proposed in this invention; Figure 2 This is a schematic diagram of the full-domain acquisition of construction data proposed in this invention; Figure 3 This is a schematic diagram of the data preprocessing proposed in this invention; Figure 4 This is a schematic diagram of the data fusion and feature extraction proposed in this invention; Figure 5 This is a schematic diagram of the initial construction technology solution proposed in this invention; Figure 6 This is a schematic diagram illustrating the construction of the adaptive optimization model and the iterative optimization of the scheme proposed in this invention; Figure 7 This is a schematic diagram illustrating the verification of the optimization scheme proposed in this invention; Figure 8 This is a schematic diagram of the construction process monitoring proposed in this invention. Detailed Implementation

[0016] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0017] An adaptive optimization system for construction technology solutions based on multi-source data is developed, including: Data acquisition and management module: Real-time acquisition of multi-source heterogeneous data throughout the entire construction cycle, which is then standardized, packaged, and stored in a dynamic data resource pool; Data preprocessing module: Based on a classification model, heterogeneous data is processed in a hierarchical manner to output a standardized dataset; Feature fusion and weight calculation module: Dynamically allocates data source weights using cosine similarity, extracts core related features of construction schemes through deep learning, and generates a structured feature set; Solution generation module: Generates initial construction solutions by combining case-based reasoning and rule-based reasoning; Reinforcement learning optimization module: Constructs a multi-objective function with quality, schedule, cost, and safety as objectives, dynamically adjusts parameters to generate candidate solutions, and selects the optimal solution through the analytic hierarchy process; Virtual simulation module: Simulates the entire construction process, evaluates the feasibility of the plan by comparing trial construction data, and provides feedback on optimization needs; Update and Execution Module: Monitors construction deviations in real time, triggers a scheme update mechanism when the deviation exceeds a threshold, and generates a corrected scheme based on the latest data; Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

[0018] See Figure 1 As shown, an adaptive optimization method for construction technology solutions based on multi-source data is developed, including: Step 1: Collect multi-source heterogeneous data throughout the entire lifecycle of the construction project through distributed data acquisition terminals, including basic environmental data, engineering design data, construction resource data, historical engineering data, and real-time construction monitoring data. After standardizing and encapsulating the collected raw data, store it in a dynamic data resource pool. Step 2: Construct a classification preprocessing model based on data type differences to perform hierarchical processing on the data in the dynamic data resource pool and obtain a high-quality standardized dataset; Step 3: Calculate the correlation between different data in the standardized dataset using cosine similarity, obtain the weight coefficients of each data source, extract the core correlation features of the construction plan preparation using a deep learning model, and form a construction plan correlation feature set after fusion processing; Step 4: Construct a construction scheme generation model based on a hybrid reasoning mechanism that combines case-based reasoning and rule-based reasoning. Input the feature set associated with the construction scheme and combine it with the core requirements of the construction project to generate an initial construction technical scheme. Step 5: Construct an adaptive optimization model for construction technology solutions based on reinforcement learning theory, establish a multi-objective optimization function, input real-time updated data and associated feature sets, generate multiple sets of optimization candidate solutions by dynamically adjusting the optimization variables, and use the analytic hierarchy process (AHP) to screen the optimal construction technology solution. Step Six: Construct a virtual construction simulation platform, import the optimal construction technology solution, establish a simulation model based on the actual environment and design data, simulate the entire construction process, and compare the real-time collected data with the simulation results by carrying out small-scale trial construction to evaluate the suitability of the solution. Step 7: During actual implementation, the real-time monitoring data is compared with the preset parameters in the optimization plan. When the deviation between the monitoring data and the preset parameters exceeds the threshold, the plan update mechanism is triggered to generate an updated construction technology plan.

[0019] See Figure 2 As shown, multi-source heterogeneous data throughout the entire lifecycle of a construction project is collected via distributed data acquisition terminals. This includes basic environmental data, engineering design data, construction resource data, historical project data, and real-time construction monitoring data. After standardizing and encapsulating the collected raw data, it is stored in a dynamic data resource pool. Specifically, this includes: Multiple types of data acquisition terminals are deployed at key nodes, equipment operating surfaces, and surrounding environments in the construction area to build a distributed data acquisition network covering the entire area and collect data synchronously. Collect multi-source heterogeneous data in a categorized manner, including basic environmental data, engineering design data, construction resource data, historical engineering data, and real-time monitoring data; All types of raw data are packaged in a unified format, stored in a dynamic resource pool, and a real-time update trigger mechanism is set.

[0020] Specifically, the raw data collected from each terminal is classified and packaged. Structural monitoring data is constructed into structured data frames in the format of "component number-monitoring time-indicator type-value-collection terminal ID". Equipment operation data is associated with the unique code of the equipment and the operating condition label. Environmental data is supplemented with the coordinate information of the monitoring point. The data is output in JSON format. Invalid data is removed by data cleaning script. Missing data is marked and stored in a dynamic data resource pool. The resource pool adopts a distributed database architecture and manages the data according to a three-level storage strategy of "real-time data-short-term storage-long-term archiving". Real-time monitoring data is retained for 72 hours of high-priority storage, and historical data is archived and compressed daily.

[0021] See Figure 3 As shown, a classification preprocessing model is constructed based on data type differences to perform hierarchical processing on the data in the dynamic data resource pool and obtain a high-quality standardized dataset. Specifically, this includes: Construct a classification preprocessing model, divide the processing levels based on data types, and establish a mapping relationship between data types and processing strategies; A sliding window smoothing algorithm is used to eliminate random noise in the basic environmental data, and missing data is filled in by trend fitting. Perform syntax validation and logical consistency checks on engineering design data, remove redundant data, and correct error terms; An outlier detection algorithm is used to identify and remove outlier data from construction resource data, and normalization processing is used to unify the dimensions. Natural language processing technology is used to extract keywords and perform structured transformation on historical engineering data; Data downsampling and time-series alignment are applied to real-time construction monitoring data to obtain a high-quality standardized dataset.

[0022] Specifically, a three-level classification system is constructed based on the types of construction monitoring data. The first level is divided into major categories: "structural monitoring data, equipment operation data, environmental monitoring data, and progress monitoring data." The second level is further subdivided according to specific monitoring indicators, such as settlement data and stress data under structural monitoring data. The third level is based on the data acquisition terminal ID and monitoring point location. Unstructured data is converted into a standardized structured format of "time-location-indicator-value," and field naming rules and data precision are unified. Invalid data entries with missing fields or incorrect formats are eliminated. Differentiated processing strategies are adopted for the noise characteristics of different types of data. High-frequency random noise in structural monitoring data is smoothed using a sliding window method, with the window size dynamically adapted according to the data sampling frequency to filter out instantaneous fluctuation interference. Pulse-type outliers in equipment operation data are eliminated by setting reasonable threshold ranges and judging the trends of adjacent data points. Instantaneous peak values ​​of dust and noise in environmental monitoring data are eliminated by median filtering to remove false data caused by environmental interference. Personnel positioning drift data in progress monitoring data are eliminated by verifying trajectory continuity and removing abnormal point data that exceed the reasonable movement speed range.

[0023] See Figure 4 As shown, the correlation between different data in the standardized dataset is calculated using cosine similarity to obtain the weight coefficients of each data source. The core correlation features for construction plan preparation are extracted using a deep learning model. After fusion processing, the construction plan correlation feature set is formed, which specifically includes: The standardized dataset is classified and labeled with data sources, and the association strength between different data sources is calculated using cosine similarity. The weights are assigned based on the correlation strength between historical engineering data and current project cases. The weights of real-time monitoring data are adjusted in a stepwise manner based on the data update frequency. Static data is assigned a fixed baseline weight, forming a dynamic weight allocation scheme. By fusing data from different dimensions at the feature level, and extracting the core related features for construction plan preparation through a deep learning model; Redundancy removal is performed on the fused feature data to form a structured set of construction scheme related features.

[0024] Specifically, the preprocessed standardized time-series dataset is first classified based on its source, grouping data from the same monitoring object and the same monitoring dimension into one category. Abnormal data from the same source is then removed through data consistency verification. Weights are initialized based on the impact of each data source on the optimization of the construction plan, with real-time structural monitoring data weighted at 0.35, large equipment operation data at 0.25, environmental monitoring data at 0.2, and construction progress monitoring data at 0.2. The weight allocation is dynamically adjusted according to the construction stage; for example, the weight of structural monitoring data is increased to 0.45 during the deep foundation pit excavation stage, and the weight of equipment operation data is increased to 0.3 during the equipment hoisting stage. A hierarchical fusion strategy of "homogeneous data aggregation—cross-source data association—global data fusion" is adopted. The first level aggregates homogeneous data by splicing and removing redundancy to form a complete data chain for a single monitoring object. For example, stress sensor data at different depths of a support pile are aggregated into a full-depth stress time-series curve for that pile. The second level conducts cross-source data association by constructing associated data pairs based on timestamp alignment. For example, the load data of tower crane operation at the same time is associated with the weight data of hoisted components and the environmental wind speed data to uncover the inherent correlation patterns between different data sources. The third level performs global data fusion by identifying key data correlations through association rule mining and integrating multi-dimensional data into a structured global monitoring data matrix to achieve data complementarity and value-added.

[0025] See Figure 5 As shown, a construction scheme generation model is constructed based on a hybrid reasoning mechanism that combines case-based reasoning and rule-based reasoning. The model inputs a set of associated features of the construction scheme and, combined with the core requirements of the construction project, generates an initial construction technical scheme, specifically including: A construction scheme generation model is constructed, and a hybrid reasoning mechanism combining case reasoning and rule reasoning is adopted. Based on the core correlation features, the most similar case schemes are retrieved from historical engineering data, and the key construction technology, process arrangement and parameter settings in the case are extracted. By combining current construction specifications, industry standards, and project-specific requirements, a reasoning rule base is constructed. The retrieved case solutions are adaptively adjusted to determine the construction process, construction methods for each process, equipment selection scheme, material usage plan, and personnel division of labor details, thereby generating an initial construction technical solution that meets the basic requirements of the project.

[0026] Specifically, a feature-demand matching mapping model is constructed to accurately match the generated dynamic monitoring feature set of the construction process with the quantified construction objectives. Safety features correspond to safety control objectives, such as associating the feature of "frequency of stress exceeding the limit in support structure" with the objective of "settlement limit of deep foundation pit", and the feature of "abnormal equipment operation duration" with the objective of "safe operation requirements of construction equipment". Quality features correspond to quality standard objectives, such as associating the feature of "deviation in component construction accuracy" with the objective of "installation accuracy tolerance". Efficiency features correspond to schedule and cost objectives, such as associating the feature of "deviation in process completion time" with the objective of "completion time limit of key nodes". Through the mapping relationship, key fusion features that have a significant impact on the core objectives of the current construction stage are selected to form an objective-oriented feature subset. An initial construction technical solution is generated using a hybrid reasoning mechanism combining "case-based reasoning" and "rule-based reasoning." In the case-based reasoning stage, based on a goal-oriented feature subset, the system retrieves the most similar construction cases from a historical engineering case database, extracting core construction techniques, equipment selection parameters, process connections, and resource allocation schemes. In the rule-based reasoning stage, a reasoning rule base is constructed, incorporating current construction specifications, industry technical standards, and project-specific requirements. This base is then used to adaptively adjust the retrieved case solutions, such as adjusting the embedment depth of the support structure based on the geological conditions of the current construction area and optimizing the tower crane hoisting operation process based on wind speed parameters from real-time environmental monitoring data. Finally, the adjusted results are integrated to generate an initial construction technical solution that includes detailed construction process rules, process parameters for each procedure, equipment and personnel allocation plans, and key points for quality and safety control, ensuring that the solution accurately matches the actual construction conditions and core requirements.

[0027] See Figure 6 As shown, an adaptive optimization model for construction technology schemes is constructed based on reinforcement learning theory. A multi-objective optimization function is established, and real-time updated data and associated feature sets are input. By dynamically adjusting the optimization variables, multiple sets of candidate optimization schemes are generated. The analytic hierarchy process (AHP) is used to screen the optimal construction technology scheme. Specifically, the following steps are taken: An adaptive optimization model for construction technology solutions is constructed based on reinforcement learning theory, and a multi-objective optimization function is built with construction quality compliance rate, construction period completion efficiency, cost control accuracy and safety risk level as optimization objectives. The construction process parameters, resource allocation ratio, and process connection time in the initial construction technical plan are used as optimization variables. Based on the model-based intelligent decision-making mechanism, variables are dynamically adjusted and optimized in combination with real-time data feedback to generate multiple sets of differentiated candidate solutions; The analytic hierarchy process (AHP) is used to comprehensively evaluate multiple candidate solutions and select the optimal solution. If the optimal solution does not meet the preset optimization target threshold, the model optimization strategy is adjusted and the optimization process is repeated iteratively.

[0028] Specifically, based on the generated initial construction technology plan, core optimization variables are extracted, covering construction process parameters, resource allocation ratios, and process connection parameters; an adaptive optimization model based on reinforcement learning is deployed, using the extracted dynamic monitoring and correlation feature set of the construction process as the model environment input, the optimization variables as the model decision variables, and the evaluation results of the multi-objective optimization system as the model feedback signal, thus constructing a closed-loop optimization logic of "input-decision-feedback-adjustment"; An adaptive optimization model is initiated for multiple rounds of iterative optimization. Based on the initial values ​​of the optimization variables in the initial scheme and combined with real-time monitoring data, multiple sets of differentiated optimization candidate schemes are generated. Each set of schemes corresponds to different combinations of process parameters, resource allocation methods, and process connection plans. The analytic hierarchy process (AHP) is used to comprehensively evaluate the candidate schemes, scoring them one by one from four dimensions: safety adaptability, quality assurance capability, schedule feasibility, and cost economy. The comprehensive score of each scheme is calculated. The scheme with the highest comprehensive score is selected as the optimal scheme for this round and compared with the preset optimization target threshold. If the threshold is not reached, the evaluation results and the latest monitoring data are used as feedback signals to adjust the model optimization strategy and repeat the iterative process until the optimal construction technology scheme that meets all target thresholds is selected. The optimal solution selected in the final screening is dynamically verified in combination with the dynamic monitoring data of the construction process. The key verification is the adaptability of the optimized variables in the solution to the real-time working conditions, such as whether the adjusted excavation thickness matches the current geological stress monitoring data and whether the optimized equipment shift allocation meets the real-time construction progress requirements. After the verification is passed, the core process parameters, resource allocation plan and process connection requirements in the optimal solution are locked to form standardized optimization solution execution rules.

[0029] See Figure 7 As shown, a virtual construction simulation platform is constructed, the optimal construction technology solution is imported, and a simulation model is established based on the actual environment and design data to simulate the entire construction process. Through small-scale trial construction, real-time collected data is compared with simulation results to evaluate the suitability of the solution. Specifically, this includes: A virtual construction simulation platform is built, which, based on the collected basic environment and engineering design data, constructs a simulation model that is consistent with the actual construction scenario. The optimized construction technology solution is imported into the simulation model to simulate the entire construction process, and the simulation results include the construction period simulation results, quality achievement nodes, and resource consumption curves. Select a segment of the project process as the trial construction area, collect construction data in real time, and compare it with the simulation results dimension by dimension to obtain the causes of deviations; An evaluation system is constructed based on technical feasibility, economic rationality, and safety controllability to generate feasibility assessment results. If there are infeasible items, the system is iterated and optimized again.

[0030] Specifically, based on the actual working conditions of the construction project, a dual-dimensional verification scenario of "virtual simulation + on-site trial construction" is built. In the virtual scenario construction phase, the full-element model of the construction area is restored by relying on BIM technology, and the collected basic environmental data such as topography, surrounding buildings and underground pipelines, as well as core data such as structural component parameters and construction node coordinates in the engineering design are imported. At the same time, the generated dynamic monitoring associated feature set of the construction process is embedded to build a virtual construction environment that matches the actual construction scenario 1:1. In the on-site trial construction scenario, the core and key process areas are selected, obstacles in the construction area are cleared, and dynamic monitoring terminals consistent with the formal construction are deployed. A tiered verification indicator system was established, focusing on four core objectives: construction safety, quality, efficiency, and cost. Safety verification indicators include peak structural stress, component settlement and displacement, frequency of equipment malfunctions, and response time to safety risk warnings. Quality verification indicators cover component construction precision deviations, the rate of compliance with quality standards for process connections, and the matching degree of material usage. Efficiency verification indicators include process completion time, resource input-output ratio, and equipment utilization rate. Cost verification indicators involve resource consumption costs during the trial construction phase, potential costs of schedule delays, and costs of quality rectification. During virtual simulation and on-site trial construction, dynamic monitoring terminals were deployed to collect data on each verification indicator in real time. A data comparison and analysis model is constructed to compare the various indicator data output from the virtual simulation verification with the real-time monitoring data from the on-site trial construction in each dimension. The model focuses on analyzing the consistency of core data such as the trend of structural stress change, the development law of component displacement, the fluctuation of equipment operating parameters, and the efficiency of process completion. It identifies the sources of deviation between the two, such as deviations caused by factors not considered in the virtual scenario, such as on-site wind disturbance and slight equipment wear. At the same time, it combines the construction specification requirements and the core needs of the project to determine whether each verification indicator meets the preset threshold. For the deviations found in the comparison, the model conducts an in-depth analysis of the rationality of core contents such as construction process parameters, resource allocation ratio, and process connection time in the optimization plan, and forms the data comparison and analysis results. The formula for the consistency coefficient of simulation-measured deviation is:

[0031] in, This is the deviation consistency coefficient, with a value ranging from [−1, 1]. The closer it is to 1, the higher the consistency. This is a sequence of indicator data output from the virtual simulation. This is a sequence of index data measured during on-site trial construction. for The covariance reflects the degree of linear correlation between the two. The variance of the simulation data and the measured data reflects the degree of data dispersion.

[0032] See Figure 8As shown, during actual implementation, real-time monitoring data is compared with preset parameters in the optimization plan. When the deviation between the monitoring data and the preset parameters exceeds a threshold, the plan update mechanism is triggered, generating an updated construction technical plan, which specifically includes: During the actual implementation of the optimized construction technology plan, data on the progress of process completion, component construction quality inspection, material and equipment consumption, and changes in the on-site environment are collected simultaneously. The real-time monitoring data is compared with the preset parameters of the optimization scheme dimension by dimension. By setting a deviation threshold, when the deviation value between the monitoring data and the preset parameters exceeds the threshold, the update process is triggered. The latest data in the dynamic data resource pool is called, the feature fusion model is reused, the bias-related features are re-extracted, and an updated feature set is generated. The updated feature set is input into the adaptive optimization model, and local parameter corrections or global process optimizations are performed based on the deviation type to generate an updated construction technology solution.

[0033] Specifically, a deviation comparison model of "monitoring data - scheme parameters" is constructed to compare the real-time collected monitoring data with the preset parameters in the optimization scheme in real time, dimension by dimension; a graded deviation threshold is set, with the first level deviation being a slight deviation, which is only recorded and does not trigger an update; the second level deviation is a moderate deviation, which triggers an early warning; and the third level deviation is a serious deviation or a safety risk warning signal, which triggers the scheme update mechanism. After the update mechanism is triggered, the newly collected dynamic monitoring data is quickly preprocessed to remove invalid data, fill in missing data, and combine it with the updated historical data and resource data in the data resource pool. The feature fusion model is then invoked to re-extract related features, forming an updated target-oriented feature subset. Based on this feature subset, an adaptive optimization model is launched to determine the adjustment direction based on the source of the deviation: if the structural stress exceeds the limit, the construction process parameters are adjusted; if the construction period is delayed, the resource allocation is optimized; if the equipment is malfunctioning, the equipment usage parameters are corrected or the operation sequence is adjusted. Multiple candidate schemes for local adjustment or global optimization are generated, and the optimal adjustment scheme is quickly selected through the analytic hierarchy process.

[0034] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0035] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0036] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An adaptive optimization method for developing construction technology solutions based on multi-source data, characterized in that, include: Multi-source heterogeneous data throughout the entire lifecycle of a construction project are collected through distributed data acquisition terminals, including basic environmental data, engineering design data, construction resource data, historical engineering data, and real-time construction monitoring data. After the collected raw data is standardized and packaged, it is stored in a dynamic data resource pool. A classification preprocessing model is built based on data type differences to perform hierarchical processing on data in the dynamic data resource pool and obtain a high-quality standardized dataset. The correlation between different data in the standardized dataset is calculated by cosine similarity, the weight coefficients of each data source are obtained, the core correlation features of construction plan preparation are extracted by deep learning model, and after fusion processing, a construction plan correlation feature set is formed. A construction scheme generation model is constructed based on a hybrid reasoning mechanism that combines case-based reasoning and rule-based reasoning. The model takes the feature set associated with the construction scheme as input and combines it with the core requirements of the construction project to generate an initial construction technical scheme. An adaptive optimization model for construction technology schemes is constructed based on reinforcement learning theory. A multi-objective optimization function is established, and real-time updated data and associated feature sets are input. By dynamically adjusting the optimization variables, multiple sets of optimization candidate schemes are generated, and the optimal construction technology scheme is selected using the analytic hierarchy process. A virtual construction simulation platform was built, the optimal construction technology solution was imported, and a simulation model was established based on the actual environment and design data to simulate the entire construction process. By carrying out small-scale trial construction, the real-time collected data and simulation results were compared to evaluate the suitability of the solution. In actual implementation, real-time monitoring data is compared with preset parameters in the optimization plan. When the deviation between the monitoring data and the preset parameters exceeds the threshold, the plan update mechanism is triggered to generate an updated construction technology plan.

2. The adaptive optimization method for construction technology scheme preparation based on multi-source data according to claim 1, characterized in that, The process of collecting multi-source heterogeneous data throughout the entire lifecycle of a construction project via distributed data acquisition terminals, including basic environmental data, engineering design data, construction resource data, historical project data, and real-time construction monitoring data, and then standardizing and encapsulating the collected raw data before storing it in a dynamic data resource pool, specifically includes: Multiple types of data acquisition terminals are deployed at key nodes, equipment operating surfaces, and surrounding environments in the construction area to build a distributed data acquisition network covering the entire area and collect data synchronously. Collect multi-source heterogeneous data in a categorized manner, including basic environmental data, engineering design data, construction resource data, historical engineering data, and real-time monitoring data; All types of raw data are packaged in a unified format, stored in a dynamic resource pool, and a real-time update trigger mechanism is set.

3. The adaptive optimization method for construction technology scheme preparation based on multi-source data according to claim 1, characterized in that, The step of constructing a classification preprocessing model based on data type differences to perform hierarchical processing on data in the dynamic data resource pool and obtain a high-quality standardized dataset specifically includes: Construct a classification preprocessing model, divide the processing levels based on data types, and establish a mapping relationship between data types and processing strategies; A sliding window smoothing algorithm is used to eliminate random noise in the basic environmental data, and missing data is filled in by trend fitting. Perform syntax validation and logical consistency checks on engineering design data, remove redundant data, and correct error terms; An outlier detection algorithm is used to identify and remove outlier data from construction resource data, and normalization processing is used to unify the dimensions. Natural language processing technology is used to extract keywords and perform structured transformation on historical engineering data; Data downsampling and time-series alignment are applied to real-time construction monitoring data to obtain a high-quality standardized dataset.

4. The adaptive optimization method for construction technology scheme preparation based on multi-source data according to claim 1, characterized in that, The process of calculating the correlation between different data in the standardized dataset using cosine similarity, obtaining the weight coefficients of each data source, extracting the core correlation features of the construction plan compilation using a deep learning model, and then fusing them to form a construction plan correlation feature set specifically includes: The standardized dataset is classified and labeled with data sources, and the association strength between different data sources is calculated using cosine similarity. The weights are assigned based on the correlation strength between historical engineering data and current project cases. The weights of real-time monitoring data are adjusted in a stepwise manner based on the data update frequency. Static data is assigned a fixed baseline weight, forming a dynamic weight allocation scheme. By fusing data from different dimensions at the feature level, and extracting the core related features for construction plan preparation through a deep learning model; Redundancy removal is performed on the fused feature data to form a structured set of construction scheme related features.

5. The adaptive optimization method for construction technology scheme preparation based on multi-source data according to claim 1, characterized in that, The construction scheme generation model, constructed using a hybrid reasoning mechanism combining case-based reasoning and rule-based reasoning, inputs a set of associated features of the construction scheme and, in conjunction with the core requirements of the construction project, generates an initial construction technical scheme, specifically including: A construction scheme generation model is constructed, and a hybrid reasoning mechanism combining case reasoning and rule reasoning is adopted. Based on the core correlation features, the most similar case schemes are retrieved from historical engineering data, and the key construction technology, process arrangement and parameter settings in the case are extracted. By combining current construction specifications, industry standards, and project-specific requirements, a reasoning rule base is constructed. The retrieved case solutions are adaptively adjusted to determine the construction process, construction methods for each process, equipment selection scheme, material usage plan, and personnel division of labor details, thereby generating an initial construction technical solution that meets the basic requirements of the project.

6. The adaptive optimization method for construction technology scheme preparation based on multi-source data according to claim 1, characterized in that, The adaptive optimization model for construction technology schemes based on reinforcement learning theory is constructed, a multi-objective optimization function is established, real-time updated data and associated feature sets are input, and multiple sets of optimization candidate schemes are generated through dynamic adjustment of optimization variables. The optimal construction technology scheme is selected using the analytic hierarchy process (AHP). Specifically, this includes: An adaptive optimization model for construction technology solutions is constructed based on reinforcement learning theory, and a multi-objective optimization function is built with construction quality compliance rate, construction period completion efficiency, cost control accuracy and safety risk level as optimization objectives. The construction process parameters, resource allocation ratio, and process connection time in the initial construction technical plan are used as optimization variables. Based on the model-based intelligent decision-making mechanism, variables are dynamically adjusted and optimized in combination with real-time data feedback to generate multiple sets of differentiated candidate solutions; The analytic hierarchy process (AHP) is used to comprehensively evaluate multiple candidate solutions and select the optimal solution. If the optimal solution does not meet the preset optimization target threshold, the model optimization strategy is adjusted and the optimization process is repeated iteratively.

7. The adaptive optimization method for compiling construction technology solutions based on multi-source data according to claim 1, characterized in that, The construction of a virtual construction simulation platform, the import of optimal construction technology solutions, the establishment of simulation models based on actual environment and design data, the simulation of the entire construction process, and the evaluation of the suitability of the solution by comparing real-time collected data with simulation results through small-scale trial construction include: A virtual construction simulation platform is built, which, based on the collected basic environment and engineering design data, constructs a simulation model that is consistent with the actual construction scenario. The optimized construction technology solution is imported into the simulation model to simulate the entire construction process, and the simulation results include the construction period simulation results, quality achievement nodes, and resource consumption curves. Select a segment of the project process as the trial construction area, collect construction data in real time, and compare it with the simulation results dimension by dimension to obtain the causes of deviations; An evaluation system is constructed based on technical feasibility, economic rationality, and safety controllability to generate feasibility assessment results. If there are infeasible items, the system is iterated and optimized again.

8. The adaptive optimization method for compiling construction technology solutions based on multi-source data according to claim 1, characterized in that, In actual implementation, the real-time monitoring data is compared with the preset parameters in the optimization plan. When the deviation between the monitoring data and the preset parameters exceeds a threshold, the plan update mechanism is triggered to generate an updated construction technical plan. Specifically, this includes: During the actual implementation of the optimized construction technology plan, data on the progress of process completion, component construction quality inspection, material and equipment consumption, and changes in the on-site environment are collected simultaneously. The real-time monitoring data is compared with the preset parameters of the optimization scheme dimension by dimension. By setting a deviation threshold, when the deviation value between the monitoring data and the preset parameters exceeds the threshold, the update process is triggered. The latest data in the dynamic data resource pool is called, the feature fusion model is reused, the bias-related features are re-extracted, and an updated feature set is generated. The updated feature set is input into the adaptive optimization model, and local parameter corrections or global process optimizations are performed based on the deviation type to generate an updated construction technology solution.

9. An adaptive optimization system for compiling construction technology schemes based on multi-source data, used to implement the adaptive optimization method for compiling construction technology schemes based on multi-source data as described in any one of claims 1-8, characterized in that, include: Data acquisition and management module: Real-time acquisition of multi-source heterogeneous data throughout the entire construction cycle, which is then standardized, packaged, and stored in a dynamic data resource pool; Data preprocessing module: Based on a classification model, heterogeneous data is processed in a hierarchical manner to output a standardized dataset; Feature fusion and weight calculation module: Dynamically allocates data source weights using cosine similarity, extracts core related features of construction schemes through deep learning, and generates a structured feature set; Solution generation module: Generates initial construction solutions by combining case-based reasoning and rule-based reasoning; Reinforcement learning optimization module: Constructs a multi-objective function with quality, schedule, cost, and safety as objectives, dynamically adjusts parameters to generate candidate solutions, and selects the optimal solution through the analytic hierarchy process; Virtual simulation module: Simulates the entire construction process, evaluates the feasibility of the plan by comparing trial construction data, and provides feedback on optimization needs; Update and Execution Module: Monitors construction deviations in real time, triggers a scheme update mechanism when the deviation exceeds a threshold, and generates a corrected scheme based on the latest data; Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

Images