BIM-based methods and systems for dynamic management and optimization of project schedule and cost

Abstract

This invention provides a BIM-based method and system for dynamic management and optimization of project schedule and cost, belonging to the field of project management and optimization technology. It includes constructing a digital twin platform with real-time sensing capabilities based on a BIM model and multi-source sensor data; importing schedule plans and cost baseline data to establish a schedule-cost correlation model, calculating a comprehensive deviation value, and automatically generating multiple sets of optimization schemes containing resource allocation, process organization, and cost control parameters; performing parallel simulations on the digital twin platform to obtain schedule and cost prediction results, conducting multi-objective evaluation, selecting the optimal scheme, and issuing it to the site for execution; continuously collecting actual execution data and comparing it with simulation results; when the deviation exceeds a threshold, updating the schedule-cost correlation model in the digital twin platform using actual data to achieve adaptive evolution and form dynamic collaborative management and control of schedule and cost. This invention can effectively improve the accuracy and efficiency of schedule and cost management and reduce project risks.

Description

Technical Field

[0001] This invention relates to engineering management and optimization technology, and more particularly to a method and system for dynamic management and optimization of engineering progress and cost based on BIM. Background Technology

[0002] Building Information Modeling (BIM) technology is widely used in the engineering construction field, providing digital support for the full lifecycle management of engineering projects. BIM technology enables 3D visualization, information integration, and collaborative management of engineering projects. During the construction phase, BIM technology can be used to simulate the construction process, optimize construction plans, and control construction progress and costs. Traditional construction management methods mainly rely on manual experience and static data, making them difficult to adapt to the complex and ever-changing construction site environment. With the rapid development of technologies such as the Internet of Things, big data, and artificial intelligence, integrating these new technologies with BIM technology can achieve dynamic control and optimization of construction progress and costs.

[0003] However, existing BIM-based methods for project schedule and cost control still have some shortcomings and deficiencies:

[0004] Lack of real-time sensing capabilities: Traditional BIM models are usually static and cannot reflect real-time changes on the construction site. This lack of real-time awareness of construction resources, personnel, and equipment leads to inaccurate prediction and control of schedule and costs.

[0005] Weak correlation between schedule and cost: Most existing methods manage schedule and cost separately, lacking in-depth analysis of the relationship between the two. This makes it difficult to effectively coordinate the relationship between schedule and cost, resulting in the inability to maximize the overall benefits of the project.

[0006] The methods for generating and evaluating optimization solutions are not intelligent enough: traditional optimization solutions mainly rely on human experience and lack intelligent decision support. This results in low efficiency in generating optimization solutions and makes it difficult to guarantee the effectiveness and feasibility of the solutions. Summary of the Invention

[0007] This invention provides a BIM-based method and system for dynamic management and optimization of project progress and cost, which can solve the problems in the prior art.

[0008] A first aspect of the present invention,

[0009] Provides BIM-based methods for dynamic management and optimization of project schedule and cost, including:

[0010] Computer vision algorithms are used to analyze the BIM model to extract engineering spatial information and component attribute information, and an initial construction scene model is established. Visual sensors are deployed at the construction site to collect construction implementation status data, positioning sensors to collect personnel distribution data, and IoT sensors to collect mechanical equipment operation data. The construction implementation status data, personnel distribution data, and mechanical equipment operation data are weighted and combined to generate multi-source data. This multi-source data is then input into a pre-trained deep learning model for feature extraction and data fusion to generate real-time construction resource status data. Finally, the real-time construction resource status data is dynamically mapped to the initial construction scene model to construct a digital twin platform with real-time perception capabilities.

[0011] Based on the aforementioned digital twin platform, construction schedule data and cost control benchmark data are imported to establish a schedule-cost correlation model. Real-time status data of construction resources is received, and the current progress and cost consumption are calculated using the schedule-cost correlation model. A comprehensive deviation value is obtained by comparing this value with the construction schedule data and the cost control benchmark data. Multiple optimization schemes, including resource allocation parameters, process organization parameters, and cost control parameters, are automatically generated based on the comprehensive deviation value. These optimization schemes are then input into the digital twin platform for parallel simulation to obtain simulation prediction results for each scheme. The simulation prediction results include schedule prediction results and cost prediction results.

[0012] The progress prediction results and cost prediction results are evaluated against multiple objectives to select the execution plan with the best overall benefits. The optimal execution plan is then distributed to the field management terminal through the digital twin platform, and the execution monitoring module is activated to continuously collect actual execution data. The actual execution data is compared and analyzed in real time with the simulation prediction results. When the comparison deviation exceeds a preset threshold, the progress-cost correlation model in the digital twin platform is updated using the actual execution data to achieve adaptive evolution of the digital twin environment, ensuring the accuracy of the next round of simulation verification and forming a dynamic collaborative management and control system for construction progress and cost through a digital twin.

[0013] Based on the aforementioned digital twin platform, construction schedule data and cost control benchmark data are imported to establish a schedule-cost correlation model. Real-time status data of construction resources is received, and the current progress and cost consumption are calculated using the schedule-cost correlation model. A comprehensive deviation value is obtained by comparing this value with the construction schedule data and cost control benchmark data, including:

[0014] Import construction schedule data and cost control benchmark data into the digital twin platform; extract features from the real-time status data of construction resources collected at the construction site, extract work efficiency features from personnel data, energy consumption level features from equipment data, and loss rate features from material data, and construct a resource allocation feature vector from the extracted work efficiency features, energy consumption level features, and loss rate features;

[0015] A directed acyclic graph is constructed based on the logical dependencies between construction procedures. Procedure completion features and critical path features are extracted from the directed acyclic graph, and the procedure completion features and critical path features are constructed into a procedure implementation feature vector.

[0016] A multi-layer neural network is constructed as a schedule-cost correlation model, and an attention mechanism module is set in the middle layer of the schedule-cost correlation model. The resource allocation feature vector and the process implementation feature vector are input into the schedule-cost correlation model, and the dynamic correlation strength between the resource allocation feature vector and the process implementation feature vector is calculated through the attention mechanism module to generate a resource-process interaction matrix.

[0017] The resource-process interaction matrix is ​​input into the schedule-cost correlation model to calculate the current construction progress and cost consumption, thereby obtaining schedule prediction data and cost prediction data. The schedule prediction data is compared with the construction schedule plan data to calculate the schedule deviation value, and the cost prediction data is compared with the cost control benchmark data to calculate the cost deviation value. The comprehensive deviation value is calculated based on the combination of the schedule deviation data and the cost deviation data.

[0018] The progress forecast data is compared with the construction progress plan data to calculate the progress deviation value, and the cost forecast data is compared with the cost control benchmark data to calculate the cost deviation value, including:

[0019] Collect basic data, including real-time construction site data and planned completion data of work processes; extract actual completion data, resource usage data, and cost consumption data of work processes from the real-time construction site data, and store the actual completion data, resource usage data, and cost consumption data of work processes, resource usage data, and cost consumption data in a sliding time window according to a time series.

[0020] A progress prediction model is constructed. The completion rate of the process is calculated based on the actual completion data of the process. The completion rate of the process is obtained by weighting the ratio of the actual completed work volume of the process to the planned completion data of the process. The position of the process in the critical path is analyzed to determine the position importance coefficient. The consumption of manpower, machinery and materials of the process is analyzed to determine the resource consumption coefficient. The dependencies between processes are analyzed to determine the process correlation coefficient.

[0021] The process weight coefficient is obtained by weighting the location importance coefficient, the resource consumption coefficient, and the process correlation coefficient; the progress prediction data is obtained by multiplying the process completion rate and the process weight coefficient; the progress prediction data is compared with the planned completion data of the process to calculate the progress deviation value.

[0022] A cost prediction model is constructed to calculate labor costs, equipment costs, and material costs based on the resource usage data and the cost consumption data; the labor cost is obtained by multiplying the actual working hours by the actual labor unit price, the equipment cost is obtained by multiplying the actual equipment usage time by the actual equipment rental rate, and the material cost is obtained by multiplying the actual material usage by the actual material unit price.

[0023] The Kalman filter algorithm is used to smooth the noise of the labor cost, the equipment cost, and the material cost. The smoothed data are combined to obtain the cost prediction data. The cost prediction data is compared with the cost control benchmark data to calculate the cost deviation value.

[0024] Based on the comprehensive deviation value, multiple sets of optimization schemes are automatically generated, including resource allocation parameters, process organization parameters, and cost control parameters. These multiple sets of optimization schemes are then input into the digital twin platform for parallel simulation to obtain simulation prediction results for each scheme. The simulation prediction results include schedule prediction results and cost prediction results.

[0025] The comprehensive deviation value includes a schedule deviation value and a cost deviation value. A schedule optimization objective function is constructed based on the schedule deviation value, and a cost optimization objective function is constructed based on the cost deviation value. The schedule optimization objective function and the cost optimization objective function are input into an improved particle swarm optimization algorithm to generate multiple sets of optimization schemes through iterative optimization. Each set of optimization schemes includes resource allocation parameters, process organization parameters, and cost control parameters. The resource allocation parameters are used to optimize the number of construction team members, the number of equipment, and the material supply. The process organization parameters are used to optimize the process overlap relationship, process duration, and process resource requirements. The cost control parameters are used to optimize the labor unit price, equipment rate, and material unit price.

[0026] A construction site model, a resource flow model, and a process implementation model are constructed on the digital twin platform. The resource configuration parameters are mapped to the resource flow model, the process organization parameters are mapped to the process implementation model, and the cost control parameters are mapped to the construction site model. A distributed computing architecture is used to distribute the mapped models to multiple computing nodes.

[0027] Parallel simulations are performed on the multiple computing nodes; resource scheduling simulations are performed based on the resource flow model to obtain simulated resource utilization prediction data; process execution simulations are performed based on the process implementation model to obtain simulated progress prediction data; and cost consumption simulations are performed based on the construction site model to obtain simulated cost prediction data; the simulated resource utilization prediction data, the simulated progress prediction data, and the simulated cost prediction data are combined to form the simulation prediction result.

[0028] Constructing a schedule optimization objective function based on the schedule deviation value, and constructing a cost optimization objective function based on the cost deviation value, includes:

[0029] The K-means clustering algorithm is used to divide the progress deviation value into deviation process groups that are higher than the preset deviation threshold and deviation process groups that are lower than the preset deviation threshold. The analytic hierarchy process is used to construct a judgment matrix for the deviation process groups that are higher than the preset deviation threshold and the deviation process groups that are lower than the preset deviation threshold from three dimensions: process importance, process criticality and process duration.

[0030] The eigenvectors of the judgment matrix are calculated using the geometric mean method to obtain the process weight coefficients; the products of the process weight coefficients and the schedule deviation values ​​are summed, and the schedule optimization objective function is constructed based on the minimum value of the summation result.

[0031] The formula for calculating the objective function for schedule optimization is as follows:

[0032]

[0033] Among them, F s The objective function for schedule optimization is given, where m is the total number of operations and w is the total number of operations. i Let d be the weighting coefficient for the i-th process. i Let be the schedule deviation value for the i-th process;

[0034] The K-means clustering algorithm is used to divide the cost deviation value into a deviation cost group that is higher than a preset deviation threshold and a deviation cost group that is lower than a preset deviation threshold. The Delphi method is used to evaluate the deviation cost group that is higher than the preset deviation threshold and the deviation cost group that is lower than the preset deviation threshold from three dimensions: cost proportion, cost controllability and cost sensitivity. The cost weight coefficient is obtained through multiple rounds of expert evaluation.

[0035] The cost weighting coefficients and the cost deviation values ​​are summed, and the minimum value of the summation is used to construct the cost optimization objective function;

[0036] The formula for calculating the cost optimization target is as follows:

[0037]

[0038] Among them, F c Let n be the total number of cost items, and W be the objective function for cost optimization. j Let C be the weighting coefficient for the j-th cost item. j Let be the cost deviation value of the j-th cost item.

[0039] The progress optimization objective function and the cost optimization objective function are input into the improved particle swarm optimization algorithm, and multiple optimization schemes are generated through iterative optimization, including:

[0040] An adaptive weighting coefficient is calculated based on the schedule deviation value and the cost deviation value. The absolute value of the deviation coefficient is obtained by weighting the absolute value of the schedule deviation value with the absolute value of the cost deviation value. The schedule weighting coefficient in the adaptive weighting coefficient is obtained based on the ratio of the absolute value of the schedule deviation value to the absolute value of the deviation coefficient. The cost weighting coefficient in the adaptive weighting coefficient is obtained based on the ratio of the absolute value of the cost deviation value to the absolute value of the deviation coefficient.

[0041] An initial particle swarm is generated using a Logistic chaotic mapping, and the initial particle swarm is mapped to a decision space to construct a particle swarm position vector. The distance between the particle swarm position vector and the centroid of the population is calculated to obtain the population aggregation degree. An adaptive mutation probability is set based on the population aggregation degree, and the adaptive mutation probability increases exponentially with the increase of the population aggregation degree.

[0042] The progress optimization objective function and the cost optimization objective function are input into the improved particle swarm optimization algorithm; the inertia weight is updated in a non-linear decreasing manner, and the inertia weight decreases according to the cosine function as the number of iterations increases; the learning factor is updated in an adaptive manner, and the learning factor includes individual learning factor and group learning factor, wherein the individual learning factor increases according to the exponential function as the number of iterations increases, and the group learning factor decreases according to the exponential function as the number of iterations increases.

[0043] In each iteration, the degree of constraint violation is calculated based on the particle swarm position vector. The fitness value is obtained by weighting the degree of constraint violation, the progress optimization objective function, and the cost objective function using the penalty function method. The crowding distance of each particle in the swarm is calculated. Solutions with crowding distances greater than a preset distance threshold are selected as elite individuals and retained for the next generation.

[0044] A pattern search method is used to perform a local search on the elite individuals to obtain local search results for the elite individuals. The step size of the pattern search method decreases exponentially with the number of iterations. The global optimal solution is updated based on the local search results, and it is determined whether the termination condition is met. When the termination condition is met, the global optimal solution is output as the optimization scheme.

[0045] A multi-objective evaluation is performed on the schedule prediction results and the cost prediction results to select the execution plan with the best overall benefits. The optimal execution plan is then distributed to the field management terminal through the digital twin platform, and the plan execution monitoring module is activated to continuously collect actual execution data, including:

[0046] Based on the schedule forecast results, schedule benefit evaluation indicators are calculated, including critical path duration reduction indicators, construction process continuity indicators, and milestone compliance indicators. Based on the cost forecast results, cost benefit evaluation indicators are calculated, including total cost savings rate indicators, cost composition rationality indicators, and cash flow matching indicators. The schedule benefit evaluation indicators and the cost benefit evaluation indicators are then weighted and combined to form a multi-objective evaluation indicator set.

[0047] A multi-objective evaluation function is constructed, and the index values ​​in the multi-objective evaluation index set are normalized. Based on the normalized index values, normalized progress weight coefficients and normalized cost weight coefficients are set. The normalized progress weight coefficients are multiplied by the normalized progress index values ​​to obtain a progress weighted score, and the normalized cost weight coefficients are multiplied by the normalized cost index values ​​to obtain a cost weighted score.

[0048] A multi-objective comprehensive evaluation is performed on the schedule-weighted score and the cost-weighted score to calculate a comprehensive benefit score; multiple optimization schemes are ranked according to the comprehensive benefit score; and the optimization scheme with the highest comprehensive benefit score is selected as the optimal execution scheme.

[0049] The optimal execution plan is converted into a standard data format to generate an execution data package containing a construction operation instruction and a resource allocation plan; the execution data package is then distributed to the site management terminal through the digital twin platform; the execution monitoring module is activated to collect actual execution data through the data acquisition equipment deployed at the site management terminal.

[0050] A second aspect of the present invention,

[0051] Provides a BIM-based dynamic management and optimization system for project schedule and cost, including:

[0052] The first unit is used to analyze the BIM model using computer vision algorithms to extract engineering spatial information and component attribute information, and establish an initial construction scene model; deploy visual sensors to collect construction implementation status data, deploy positioning sensors to collect construction personnel distribution data, and deploy IoT sensors to collect mechanical equipment operation data at the construction site; weight and combine the construction implementation status data, construction personnel distribution data, and mechanical equipment operation data to generate multi-source data; input the multi-source data into a pre-trained deep learning model for feature extraction and data fusion to generate real-time construction resource status data; dynamically map the real-time construction resource status data with the initial construction scene model to construct a digital twin basic platform with real-time perception capabilities;

[0053] The second unit is used to import construction schedule data and cost control benchmark data based on the digital twin platform to establish a schedule-cost correlation model; receive real-time status data of construction resources, calculate the current progress completion status and cost consumption status through the schedule-cost correlation model, and compare it with the construction schedule data and cost control benchmark data to obtain a comprehensive deviation value; automatically generate multiple sets of optimization schemes containing resource allocation parameters, process organization parameters, and cost control parameters based on the comprehensive deviation value; input the multiple sets of optimization schemes into the digital twin platform for parallel simulation, and obtain the simulation prediction results of each scheme, including schedule prediction results and cost prediction results;

[0054] The third unit is used to perform multi-objective evaluation of the progress prediction results and the cost prediction results, select the execution plan with the best overall benefits, distribute the optimal execution plan to the field management terminal through the digital twin basic platform, and start the plan execution monitoring module to continuously collect actual execution data. The actual execution data is compared and analyzed with the simulation prediction results in real time. When the comparison deviation exceeds a preset threshold, the progress-cost correlation model in the digital twin basic platform is updated using the actual execution data to realize the adaptive evolution of the digital twin environment, ensure the accuracy of the next round of simulation verification, and form a dynamic collaborative management and control of construction progress and cost through a digital twin.

[0055] A third aspect of the embodiments of the present invention,

[0056] An electronic device is provided, comprising:

[0057] processor;

[0058] Memory used to store processor-executable instructions;

[0059] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0060] Fourth aspect of the present invention,

[0061] A computer-readable storage medium is provided, having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0062] The beneficial effects of this application are as follows:

[0063] 1. Improved accuracy and efficiency in schedule and cost control: This method utilizes BIM models, multi-source sensor data, and deep learning models to construct a digital twin platform with real-time sensing capabilities. This platform can monitor the status of construction resources, progress completion, and cost consumption in real time, and compare and analyze this data with planned data. This allows for more accurate understanding of project progress, early detection of deviations, and avoidance of potential risks. The automated data collection and analysis process also significantly improves control efficiency.

[0064] 2. Optimized resource allocation and process organization: Based on a schedule-cost correlation model and a digital twin platform, this method can automatically generate multiple sets of optimized solutions and evaluate the schedule and cost prediction results of each solution through parallel simulation, thereby selecting the solution with the best overall benefits. This helps to optimize resource allocation, improve process organization efficiency, reduce costs, and shorten the construction period.

[0065] 3. Achieved dynamic collaborative management and control of construction progress and cost: This method compares actual execution data with simulation prediction results in real time, and uses actual data to update the progress-cost correlation model in the digital twin platform. This enables the adaptive evolution of the digital twin environment, ensures the accuracy of simulation verification, forms dynamic collaborative management and control of construction progress and cost, and improves the overall level of project management. Attached Figure Description

[0066] Figure 1 This is a flowchart illustrating the BIM-based dynamic management and optimization method for project progress and cost according to an embodiment of the present invention.

[0067] Figure 2 This is a schematic diagram of the structure of the BIM-based dynamic management and optimization system for project progress and cost according to an embodiment of the present invention. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0069] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0070] Figure 1 This is a flowchart illustrating the BIM-based dynamic management and optimization method for project progress and cost according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0071] S11. Computer vision algorithms are used to analyze the BIM model to extract engineering spatial information and component attribute information, and an initial construction scene model is established. Visual sensors are deployed at the construction site to collect construction implementation status data, positioning sensors are deployed to collect construction personnel distribution data, and IoT sensors are deployed to collect mechanical equipment operation data. The construction implementation status data, construction personnel distribution data, and mechanical equipment operation data are weighted and combined to generate multi-source data. The multi-source data is input into a pre-trained deep learning model for feature extraction and data fusion to generate real-time construction resource status data. The real-time construction resource status data is dynamically mapped to the initial construction scene model to construct a digital twin basic platform with real-time perception capabilities.

[0072] S12. Based on the digital twin platform, import construction schedule data and cost control benchmark data to establish a schedule-cost correlation model; receive real-time status data of construction resources, calculate the current progress completion status and cost consumption status through the schedule-cost correlation model, and compare it with the construction schedule data and cost control benchmark data to obtain a comprehensive deviation value; automatically generate multiple sets of optimization schemes including resource allocation parameters, process organization parameters, and cost control parameters based on the comprehensive deviation value; input the multiple sets of optimization schemes into the digital twin platform for parallel simulation, and obtain the simulation prediction results of each scheme, including schedule prediction results and cost prediction results;

[0073] S13. Perform multi-objective evaluation on the progress prediction results and the cost prediction results, and select the execution plan with the best comprehensive benefits; distribute the optimal execution plan to the field management terminal through the digital twin basic platform, and start the plan execution monitoring module to continuously collect actual execution data; compare and analyze the actual execution data with the simulation prediction results in real time, and when the comparison deviation exceeds a preset threshold, update the progress-cost correlation model in the digital twin basic platform using the actual execution data to realize the adaptive evolution of the digital twin environment, ensure the accuracy of the next round of simulation verification, and form a dynamic collaborative management and control of construction progress and cost through a digital twin.

[0074] In one optional implementation, based on the digital twin platform, construction schedule data and cost control benchmark data are imported to establish a schedule-cost correlation model; real-time status data of construction resources is received, and the current progress completion status and cost consumption status are calculated through the schedule-cost correlation model. The comprehensive deviation value is obtained by comparing this value with the construction schedule data and cost control benchmark data, including:

[0075] Import construction schedule data and cost control benchmark data into the digital twin platform; extract features from the real-time status data of construction resources collected at the construction site, extract work efficiency features from personnel data, energy consumption level features from equipment data, and loss rate features from material data, and construct a resource allocation feature vector from the extracted work efficiency features, energy consumption level features, and loss rate features;

[0076] A directed acyclic graph is constructed based on the logical dependencies between construction procedures. Procedure completion features and critical path features are extracted from the directed acyclic graph, and the procedure completion features and critical path features are constructed into a procedure implementation feature vector.

[0077] A multi-layer neural network is constructed as a schedule-cost correlation model, and an attention mechanism module is set in the middle layer of the schedule-cost correlation model. The resource allocation feature vector and the process implementation feature vector are input into the schedule-cost correlation model, and the dynamic correlation strength between the resource allocation feature vector and the process implementation feature vector is calculated through the attention mechanism module to generate a resource-process interaction matrix.

[0078] The resource-process interaction matrix is ​​input into the schedule-cost correlation model to calculate the current construction progress and cost consumption, thereby obtaining schedule prediction data and cost prediction data. The schedule prediction data is compared with the construction schedule plan data to calculate the schedule deviation value, and the cost prediction data is compared with the cost control benchmark data to calculate the cost deviation value. The comprehensive deviation value is calculated based on the combination of the schedule deviation data and the cost deviation data.

[0079] Based on a digital twin platform, a schedule-cost correlation model is established to achieve real-time monitoring and prediction of construction project schedule and cost, and to provide comprehensive deviation analysis. The core of this method lies in utilizing a multi-layer neural network combined with an attention mechanism to dynamically correlate resource allocation and process implementation, thereby more accurately predicting project schedule and cost.

[0080] First, the construction schedule data and cost control baseline data are imported into the digital twin platform. The schedule data includes the planned start time, end time, duration, and logical relationships between each process. The cost control baseline data includes the budgeted cost, actual cost, and cost composition for each process. For example, in the foundation engineering phase of a project, the planned duration is 30 days, and the budgeted cost is 1 million yuan.

[0081] Next, real-time resource status data from the construction site is collected, including personnel, equipment, and material data. Feature extraction is then performed on the collected data. From personnel data, operational efficiency features are extracted, such as the time required for a worker to complete a unit of work. From equipment data, energy consumption level features are extracted, such as the hourly fuel consumption of an excavator. From material data, loss rate features are extracted, such as the ratio of actual cement usage to planned usage. The extracted operational efficiency, energy consumption level, and loss rate features are combined into a resource allocation feature vector. For example, the resource allocation feature vector for a given day could be represented as [average worker operational efficiency: 0.8, average excavator fuel consumption: 10 liters / hour, cement loss rate: 5%].

[0082] Then, a directed acyclic graph (DAG) is constructed based on the logical dependencies between construction procedures. For example, the procedures for foundation engineering include: excavation, foundation pouring, and backfilling. Excavation is a prerequisite for foundation pouring, and foundation pouring is a prerequisite for backfilling. Procedure completion features are extracted from the DAG, such as the percentage of procedures currently completed. Critical path features are extracted, such as the length of the critical path and the completion status of procedures on the current critical path. The procedure completion features and critical path features are combined into a procedure implementation feature vector. For example, the procedure implementation feature vector for a certain day can be represented as [procedure completion: 60%, critical path length: 20 days, critical path completion: 50%].

[0083] Next, a multi-layer neural network is constructed as the schedule-cost correlation model. An attention mechanism module is set up in the middle layer of the model. Resource allocation feature vectors and process implementation feature vectors are input into the schedule-cost correlation model. The attention mechanism module calculates the dynamic correlation strength between the resource allocation feature vectors and process implementation feature vectors, generating a resource-process interaction matrix. This matrix reflects the degree of influence of different resource configurations on the completion status of different processes. For example, the resource-process interaction matrix can show that increased fuel consumption of the excavator has a significant impact on the completion schedule of the foundation pouring process.

[0084] The resource-process interaction matrix is ​​input into the schedule-cost correlation model to calculate the current construction progress and cost consumption, obtaining schedule and cost forecast data. For example, the model predicts the current progress completion rate to be 65% and the cost consumption to be 600,000 yuan. The schedule forecast data is compared with the planned construction schedule data to calculate the schedule deviation value, and the cost forecast data is compared with the cost control benchmark data to calculate the cost deviation value. For example, the schedule deviation value is +5%, and the cost deviation value is -40%. The combined deviation value is calculated based on the schedule deviation data and the cost deviation data. For example, the combined deviation value is -30%.

[0085] The solution proposed in this application can:

[0086] Improved Prediction Accuracy: By dynamically linking resource allocation and process implementation through an attention mechanism, this method can more accurately predict project progress and costs, avoiding prediction errors caused by traditional methods that ignore the impact of resource allocation on processes. Real-time Monitoring and Early Warning: This method can receive construction resource status data in real time and perform real-time calculations, promptly identifying deviations in project progress and costs and issuing early warnings, providing decision support for project managers. Optimized Resource Allocation: Through the resource-process interaction matrix, the impact of different resource configurations on process completion can be analyzed, thereby optimizing resource allocation, improving resource utilization efficiency, and reducing project costs.

[0087] In one optional implementation, comparing the progress forecast data with the construction progress plan data to calculate the progress deviation value, and comparing the cost forecast data with the cost control benchmark data to calculate the cost deviation value, includes:

[0088] Collect basic data, including real-time construction site data and planned completion data of work processes; extract actual completion data, resource usage data, and cost consumption data of work processes from the real-time construction site data, and store the actual completion data, resource usage data, and cost consumption data of work processes, resource usage data, and cost consumption data in a sliding time window according to a time series.

[0089] A progress prediction model is constructed. The completion rate of the process is calculated based on the actual completion data of the process. The completion rate of the process is obtained by weighting the ratio of the actual completed work volume of the process to the planned completion data of the process. The position of the process in the critical path is analyzed to determine the position importance coefficient. The consumption of manpower, machinery and materials of the process is analyzed to determine the resource consumption coefficient. The dependencies between processes are analyzed to determine the process correlation coefficient.

[0090] The process weight coefficient is obtained by weighting the location importance coefficient, the resource consumption coefficient, and the process correlation coefficient; the progress prediction data is obtained by multiplying the process completion rate and the process weight coefficient; the progress prediction data is compared with the planned completion data of the process to calculate the progress deviation value.

[0091] A cost prediction model is constructed to calculate labor costs, equipment costs, and material costs based on the resource usage data and the cost consumption data; the labor cost is obtained by multiplying the actual working hours by the actual labor unit price, the equipment cost is obtained by multiplying the actual equipment usage time by the actual equipment rental rate, and the material cost is obtained by multiplying the actual material usage by the actual material unit price.

[0092] The Kalman filter algorithm is used to smooth the noise of the labor cost, the equipment cost, and the material cost. The smoothed data are combined to obtain the cost prediction data. The cost prediction data is compared with the cost control benchmark data to calculate the cost deviation value.

[0093] A method for predicting project progress and cost deviations based on real-time data, the specific implementation steps of which are as follows:

[0094] First, basic data needs to be collected. This step requires obtaining real-time data from various sources, such as sensors at the construction site, material management systems, and personnel attendance systems. This includes the actual completed work volume for each process, resource usage (including labor hours, equipment usage time, and material consumption), and corresponding cost expenditures. Simultaneously, it's also necessary to collect project process completion data, such as the planned start and end times for each process and the planned completed work volume. For example, suppose process A of a project planned to complete 1000 cubic meters of work, actually completed 800 cubic meters, had 100 hours of labor, 50 hours of equipment usage, and consumed 200 tons of materials.

[0095] Next, the collected real-time construction site data and planned completion data for each work process are stored in a sliding time window and organized according to time sequence. The size of the sliding time window can be adjusted according to the specific circumstances of the project; for example, it can be set to one week or one month. The advantage of doing this is that it allows for dynamic tracking of project progress and cost changes, and timely prediction and adjustments.

[0096] Next, a schedule prediction model is constructed. First, the completion rate of each process is calculated. For example, the completion rate of process A is 80% (actual completed work volume of 800 cubic meters divided by planned completed work volume of 1000 cubic meters).

[0097] To more accurately predict schedules, the importance of each process needs to be considered. This can be determined by analyzing the process's location on the critical path, its resource consumption, and its dependencies on other processes. For example, processes located on the critical path, those with high resource consumption, and those closely related to other processes have higher importance coefficients. Assume process A has a location importance coefficient of 0.8, a resource consumption coefficient of 0.9, and a process dependency coefficient of 0.7.

[0098] The weight coefficient of the process is obtained by weighting the location importance coefficient, resource consumption coefficient, and process correlation coefficient.

[0099] Next, multiply the process completion rate by the process weight coefficient to obtain the schedule forecast data. For example, the schedule forecast data for process A is 80% * 0.79 = 63.2%. By comparing the schedule forecast data with the planned completion data for the process, the schedule deviation value can be calculated.

[0100] Subsequently, a cost prediction model was constructed. Based on the collected resource usage and cost consumption data, labor costs, equipment costs, and material costs were calculated respectively. Labor costs were calculated by multiplying actual working hours by the actual labor unit price. For example, assuming the actual labor unit price is 50 yuan / hour, the labor cost for process A would be 100 hours * 50 yuan / hour = 5000 yuan. Equipment costs were calculated by multiplying actual equipment usage time by the actual equipment rental rate. For example, assuming the actual equipment rental rate is 100 yuan / hour, the equipment cost for process A would be 50 hours * 100 yuan / hour = 5000 yuan. Material costs were calculated by multiplying actual material usage by the actual material unit price. For example, assuming the actual material unit price is 1000 yuan / ton, the material cost for process A would be 200 tons * 1000 yuan / ton = 200,000 yuan.

[0101] To mitigate the impact of data fluctuations, a Kalman filter algorithm is used to smooth the calculated labor, equipment, and material costs. The smoothed labor, equipment, and material costs are then summed to obtain the cost forecast data. By comparing the cost forecast data with the cost control baseline data, the cost deviation value can be calculated.

[0102] The solution proposed in this application can:

[0103] Improve forecast accuracy: By comprehensively considering process completion rates, location importance, resource consumption, and dependencies between processes, project schedules and costs can be predicted more accurately. Enable dynamic monitoring: Utilizing sliding time windows and real-time data, project schedule and cost changes can be dynamically tracked, and potential risks and problems can be identified promptly. Support decision-making: Predicting schedule and cost deviations provides decision support for project managers, helping them take effective measures to control project schedule and costs.

[0104] In one optional implementation, multiple sets of optimization schemes, including resource allocation parameters, process organization parameters, and cost control parameters, are automatically generated based on the comprehensive deviation value. These multiple sets of optimization schemes are then input into the digital twin platform for parallel simulation to obtain simulation prediction results for each scheme. The simulation prediction results include schedule prediction results and cost prediction results.

[0105] The comprehensive deviation value includes a schedule deviation value and a cost deviation value. A schedule optimization objective function is constructed based on the schedule deviation value, and a cost optimization objective function is constructed based on the cost deviation value. The schedule optimization objective function and the cost optimization objective function are input into an improved particle swarm optimization algorithm to generate multiple sets of optimization schemes through iterative optimization. Each set of optimization schemes includes resource allocation parameters, process organization parameters, and cost control parameters. The resource allocation parameters are used to optimize the number of construction team members, the number of equipment, and the material supply. The process organization parameters are used to optimize the process overlap relationship, process duration, and process resource requirements. The cost control parameters are used to optimize the labor unit price, equipment rate, and material unit price.

[0106] A construction site model, a resource flow model, and a process implementation model are constructed on the digital twin platform. The resource configuration parameters are mapped to the resource flow model, the process organization parameters are mapped to the process implementation model, and the cost control parameters are mapped to the construction site model. A distributed computing architecture is used to distribute the mapped models to multiple computing nodes.

[0107] Parallel simulations are performed on the multiple computing nodes; resource scheduling simulations are performed based on the resource flow model to obtain simulated resource utilization prediction data; process execution simulations are performed based on the process implementation model to obtain simulated progress prediction data; and cost consumption simulations are performed based on the construction site model to obtain simulated cost prediction data; the simulated resource utilization prediction data, the simulated progress prediction data, and the simulated cost prediction data are combined to form the simulation prediction result.

[0108] A construction scheme optimization method based on digital twins and an improved particle swarm optimization algorithm can effectively improve the efficiency and accuracy of construction progress and cost control. The core of this method is to automatically generate multiple sets of optimized schemes by comprehensively considering schedule and cost deviations, perform parallel simulations on a digital twin platform, and finally select the optimal scheme.

[0109] First, it is necessary to collect the actual progress and cost data of the construction project and compare them with the planned data to calculate the schedule deviation and cost deviation. For example, assuming the planned duration of a project is 100 days and the current actual duration is 110 days, the schedule deviation is 10 days; if the planned cost is 10 million yuan and the current actual cost is 11 million yuan, the cost deviation is 1 million yuan.

[0110] Next, schedule optimization objective functions and cost optimization objective functions are constructed based on the schedule deviation and cost deviation values, respectively. The goal of the schedule optimization objective function is to minimize the schedule deviation, and the goal of the cost optimization objective function is to minimize the cost deviation.

[0111] Then, these two objective functions are input into the improved particle swarm optimization (PSO) algorithm. PSO is an optimization algorithm that simulates the foraging behavior of bird flocks, finding the optimal solution through cooperation and information sharing among particles. This method uses an improved PSO algorithm, enhancing the algorithm's convergence speed and global search capability. Through iterative optimization, the algorithm generates multiple sets of optimized solutions, each containing resource allocation parameters, process organization parameters, and cost control parameters.

[0112] Taking one set of plans as an example, resource allocation parameters may include: increasing the number of construction team members by 10%, increasing the number of equipment by 5%, and advancing the material supply by 10 days; process organization parameters may include: changing the interleaving relationship between process A and process B from sequential execution to parallel execution, shortening the duration of process C by 2 days, and increasing the resource requirements of process D by 10%; cost control parameters may include: reducing the unit price of labor by 5%, reducing the equipment rate by 2%, and reducing the unit price of materials by 3%.

[0113] On the digital twin platform, it is necessary to construct a construction site model, a resource flow model, and a process implementation model. The generated resource allocation parameters are mapped to the resource flow model, the process organization parameters to the process implementation model, and the cost control parameters to the construction site model. For example, a 10% increase in the number of construction team members will be reflected in the personnel count of the resource flow model; a 2-day reduction in process duration will be reflected in the process duration of the process implementation model; and a 5% reduction in labor cost will be reflected in the cost accounting of the construction site model.

[0114] A distributed computing architecture is adopted to allocate the mapped model to multiple computing nodes, and parallel simulations are performed on each computing node. Resource scheduling simulation is performed based on the resource flow model to obtain simulation resource utilization prediction data, such as equipment utilization rate and personnel idle rate; process execution simulation is performed based on the process implementation model to obtain simulation progress prediction data, such as the completion time of each process and the total project duration; cost consumption simulation is performed based on the construction site model to obtain simulation cost prediction data, such as labor costs, material costs, and equipment costs.

[0115] Finally, the simulation resources, including predicted data, schedule prediction data, and cost prediction data, are combined to form the simulation prediction results. For example, the simulation prediction result for a certain scheme is: the construction period is shortened by 5 days, and the cost is reduced by 800,000 yuan. By comparing the simulation prediction results of each scheme, the optimal scheme can be selected for actual construction, thereby improving the efficiency and accuracy of construction schedule and cost control.

[0116] The solution proposed in this application can:

[0117] By automatically generating multiple sets of optimized solutions and utilizing digital twin technology for parallel simulation, the time required for manual intervention and solution evaluation is significantly reduced, thus greatly improving the efficiency of solution optimization. By comprehensively considering schedule and cost deviations and employing an improved particle swarm optimization algorithm, more comprehensive and reasonable optimized solutions can be generated, thereby improving the quality of solution optimization. Simulation on the digital twin platform can simulate real construction environments and resource flows, thereby verifying the feasibility of the solutions and reducing implementation risks.

[0118] In one optional implementation, constructing a schedule optimization objective function based on the schedule deviation value and constructing a cost optimization objective function based on the cost deviation value include:

[0119] The K-means clustering algorithm is used to divide the progress deviation value into deviation process groups that are higher than the preset deviation threshold and deviation process groups that are lower than the preset deviation threshold. The analytic hierarchy process is used to construct a judgment matrix for the deviation process groups that are higher than the preset deviation threshold and the deviation process groups that are lower than the preset deviation threshold from three dimensions: process importance, process criticality and process duration.

[0120] The eigenvectors of the judgment matrix are calculated using the geometric mean method to obtain the process weight coefficients; the products of the process weight coefficients and the schedule deviation values ​​are summed, and the schedule optimization objective function is constructed based on the minimum value of the summation result.

[0121] The formula for calculating the objective function for schedule optimization is as follows:

[0122]

[0123] Among them, Fs The objective function for schedule optimization is given, where m is the total number of operations and w is the total number of operations. i Let d be the weighting coefficient for the i-th process. i Let be the schedule deviation value for the i-th process;

[0124] The K-means clustering algorithm is used to divide the cost deviation value into a deviation cost group that is higher than a preset deviation threshold and a deviation cost group that is lower than a preset deviation threshold. The Delphi method is used to evaluate the deviation cost group that is higher than the preset deviation threshold and the deviation cost group that is lower than the preset deviation threshold from three dimensions: cost proportion, cost controllability and cost sensitivity. The cost weight coefficient is obtained through multiple rounds of expert evaluation.

[0125] The cost weighting coefficients and the cost deviation values ​​are summed, and the minimum value of the summation is used to construct the cost optimization objective function;

[0126] The formula for calculating the cost optimization target is as follows:

[0127]

[0128] Among them, F c Let n be the total number of cost items, and W be the objective function for cost optimization. j Let C be the weighting coefficient for the j-th cost item. j Let be the cost deviation value of the j-th cost item.

[0129] A method for optimizing project schedule and cost based on deviation values ​​is proposed. The core of this method is to construct optimization objective functions based on the deviation values ​​of schedule and cost, and use these functions to guide project execution, thereby achieving dual optimization of schedule and cost.

[0130] First, collect project progress data, including the planned and actual completion times for each process. Then, calculate the schedule deviation for each process, which is the difference between the actual and planned completion times.

[0131] Next, the K-means clustering algorithm is used to group these schedule deviation values. A preset deviation threshold is set, and processes with schedule deviation values ​​higher than the threshold are grouped into one group, while processes with deviation values ​​lower than the threshold are grouped into another. The purpose of this is to distinguish between processes with large and small schedule deviations, so as to carry out more targeted optimization.

[0132] For the two groups of processes after division, a judgment matrix is ​​constructed based on three dimensions: process importance, process criticality, and process duration. For example, project experts can be invited to make judgments based on their experience, comparing the importance of different processes in these three dimensions pairwise and assigning corresponding weight values ​​to form the final judgment matrix.

[0133] Then, the eigenvectors of the judgment matrix are calculated using the geometric mean method to obtain the weight coefficient of each process. This coefficient reflects the importance of different processes in the overall project schedule.

[0134] Each process's weight coefficient is multiplied by its schedule deviation value, and the results for all processes are summed to obtain a single value. A schedule optimization objective function is then constructed, aiming to minimize this value. This means that processes with larger weight coefficients have smaller schedule deviation values, resulting in less impact on the overall project schedule and thus contributing to on-time project completion.

[0135] The same approach can be applied to cost optimization. First, collect project cost data, including the budgeted and actual values ​​for each cost item. Then, calculate the cost deviation for each cost item, which is the difference between the actual and budgeted value.

[0136] The K-means clustering algorithm is used to group cost deviation values. Cost items with cost deviation values ​​higher than a preset threshold are grouped into one group, and those with deviation values ​​lower than the threshold are grouped into another group.

[0137] For the two groups of cost items after division, expert evaluation was conducted from three dimensions: cost proportion, cost controllability, and cost sensitivity. For example, financial experts were invited to score the importance of different cost items in these three dimensions, and the weight coefficient of each cost item was finally determined through multiple rounds of evaluation.

[0138] Multiply the weight coefficient of each cost item by its cost deviation value, and sum the results for all cost items to obtain a single value. Construct a cost optimization objective function with the goal of minimizing this value. This means that cost items with larger weight coefficients have smaller cost deviation values ​​and thus have a smaller impact on the overall project cost, which is more conducive to project cost control.

[0139] For example, a project contains two cost items with weight coefficients of 0.7 and 0.3, and cost deviations of 1 and -2, respectively. The objective function for cost optimization would then be 0.7*1 + 0.3*(-2) = 0.1.

[0140] The solution proposed in this application can:

[0141] Improve the accuracy of project schedule and cost management. By clustering and weighting deviation values, key processes and cost items that significantly impact project schedule and cost can be identified more accurately, allowing for more targeted optimization measures. Enhance the effectiveness of project schedule and cost control. By constructing an optimization objective function, the goals of project schedule and cost management can be quantified, guiding continuous adjustments to project execution strategies and improving the effectiveness of project schedule and cost control. Enhance the scientific and standardized nature of project management. This method employs scientific clustering algorithms, analytic hierarchy process (AHP), and Delphi method, combined with expert experience, to make project schedule and cost management more scientific and standardized, contributing to improved project management overall performance.

[0142] In one optional implementation, the progress optimization objective function and the cost optimization objective function are input into an improved particle swarm optimization algorithm, and multiple sets of optimization schemes are generated through iterative optimization, including:

[0143] An adaptive weighting coefficient is calculated based on the schedule deviation value and the cost deviation value. The absolute value of the deviation coefficient is obtained by weighting the absolute value of the schedule deviation value with the absolute value of the cost deviation value. The schedule weighting coefficient in the adaptive weighting coefficient is obtained based on the ratio of the absolute value of the schedule deviation value to the absolute value of the deviation coefficient. The cost weighting coefficient in the adaptive weighting coefficient is obtained based on the ratio of the absolute value of the cost deviation value to the absolute value of the deviation coefficient.

[0144] An initial particle swarm is generated using a Logistic chaotic mapping, and the initial particle swarm is mapped to a decision space to construct a particle swarm position vector. The distance between the particle swarm position vector and the centroid of the population is calculated to obtain the population aggregation degree. An adaptive mutation probability is set based on the population aggregation degree, and the adaptive mutation probability increases exponentially with the increase of the population aggregation degree.

[0145] The progress optimization objective function and the cost optimization objective function are input into the improved particle swarm optimization algorithm; the inertia weight is updated in a non-linear decreasing manner, and the inertia weight decreases according to the cosine function as the number of iterations increases; the learning factor is updated in an adaptive manner, and the learning factor includes individual learning factor and group learning factor, wherein the individual learning factor increases according to the exponential function as the number of iterations increases, and the group learning factor decreases according to the exponential function as the number of iterations increases.

[0146] In each iteration, the degree of constraint violation is calculated based on the particle swarm position vector. The fitness value is obtained by weighting the degree of constraint violation, the progress optimization objective function, and the cost objective function using the penalty function method. The crowding distance of each particle in the swarm is calculated. Solutions with crowding distances greater than a preset distance threshold are selected as elite individuals and retained for the next generation.

[0147] A pattern search method is used to perform a local search on the elite individuals to obtain local search results for the elite individuals. The step size of the pattern search method decreases exponentially with the number of iterations. The global optimal solution is updated based on the local search results, and it is determined whether the termination condition is met. When the termination condition is met, the global optimal solution is output as the optimization scheme.

[0148] An improved particle swarm optimization (PSO) algorithm is proposed to optimize project schedule and cost, aiming to simultaneously consider both project schedule and cost objectives and find the optimal resource allocation scheme. This method introduces improved mechanisms such as adaptive weight coefficients, chaotic mapping initialization, adaptive mutation, nonlinear decreasing inertial weights, adaptive learning factors, penalty function constraint handling, elite retention strategy, and pattern search local optimization, effectively improving the algorithm's optimization performance and convergence speed.

[0149] First, determine the objective functions for schedule optimization and cost optimization. For example, the objective function for schedule optimization can be defined as minimizing the time required to complete the project, and the objective function for cost optimization can be defined as minimizing the total cost of the project.

[0150] Next, calculate the schedule deviation and cost deviation. For example, the schedule deviation can be defined as the difference between the planned completion time and the actual completion time, and the cost deviation can be defined as the difference between the budgeted cost and the actual cost. The absolute values ​​of the schedule deviation and cost deviation are weighted and summed to obtain the absolute value of the deviation coefficient. The schedule weight coefficient is equal to the ratio of the absolute value of the schedule deviation to the absolute value of the deviation coefficient, and the cost weight coefficient is equal to the ratio of the absolute value of the cost deviation to the absolute value of the deviation coefficient. These two weight coefficients will be used to balance the importance of schedule and cost targets in the optimization process. Assuming the schedule deviation is 5 days and the cost deviation is 100,000 yuan, and setting the weighting coefficients for schedule and cost to 0.6 and 0.4 respectively, then the absolute value of the deviation coefficient is 5*0.6 + 10*0.4 = 7, the schedule weight coefficient is (5*0.6) / 7 ≈ 0.43, and the cost weight coefficient is (10*0.4) / 7 ≈ 0.57.

[0151] Next, a Logistic chaotic mapping is used to generate an initial particle swarm. This initial particle swarm is then mapped to a decision space, constructing a particle swarm position vector. The decision space can contain decision variables such as resource allocation and task scheduling. The distance between the particle swarm position vector and the population centroid is calculated to obtain the population clustering degree. An adaptive mutation probability is set based on the population clustering degree. The mutation probability increases exponentially with the population clustering degree to enhance the algorithm's ability to escape local optima. For example, the mutation probability can be set to e^(-1 / clustering degree).

[0152] The progress optimization objective function and cost optimization objective function are input into the improved particle swarm optimization algorithm. The inertia weight is updated using a non-linear decreasing method, decreasing according to a cosine function as the number of iterations increases. For example, the inertia weight can be set to 0.9*cos(iteration number*π / maximum iteration number). An adaptive method is used to update the learning factor. The individual learning factor increases exponentially with the number of iterations, while the swarm learning factor decreases exponentially with the number of iterations.

[0153] In each iteration, the degree of constraint violation is calculated based on the particle swarm position vector. For example, constraints may include resource limitations and time constraints. A penalty function method is used to weight the degree of constraint violation, the progress optimization objective function, and the cost objective function to obtain the fitness value. The crowding distance of each particle in the swarm is calculated. Solutions with a crowding distance greater than a preset threshold are selected as elite individuals and retained for the next generation to maintain population diversity.

[0154] A pattern search method is used to perform a local search on elite individuals, yielding local search results. The step size of the pattern search method decreases exponentially with the number of iterations; for example, the step size can be set to the initial step size * e^(-iteration number). The global optimal solution is updated based on the local search results, and a termination condition is determined. The termination condition can be reaching the maximum number of iterations or the improvement of the solution being less than a preset threshold. When the termination condition is met, the global optimal solution is output as the optimized solution.

[0155] The solution proposed in this application can:

[0156] Improved optimization efficiency: Through improved mechanisms such as adaptive weight coefficients, chaotic mapping initialization, adaptive mutation, nonlinear decreasing inertia weights, and adaptive learning factors, this method can find the global optimum more quickly and accurately, thereby improving optimization efficiency. Enhanced solution quality: Through mechanisms such as penalty function constraint handling, elite preservation strategy, and pattern search local optimization, this method can effectively handle constraints, maintain population diversity, and perform local fine-tuning searches, thereby enhancing solution quality. Balanced schedule and cost objectives: Through adaptive weight coefficients, this method can dynamically adjust the importance of schedule and cost objectives in the optimization process according to the actual situation, thereby achieving a balanced optimization of schedule and cost.

[0157] In one optional implementation, a multi-objective evaluation is performed on the schedule prediction results and the cost prediction results to select the execution plan with the best overall benefits; the optimal execution plan is then distributed to the field management terminal through the digital twin platform, and the plan execution monitoring module is activated to continuously collect actual execution data, including:

[0158] Based on the schedule forecast results, schedule benefit evaluation indicators are calculated, including critical path duration reduction indicators, construction process continuity indicators, and milestone compliance indicators. Based on the cost forecast results, cost benefit evaluation indicators are calculated, including total cost savings rate indicators, cost composition rationality indicators, and cash flow matching indicators. The schedule benefit evaluation indicators and the cost benefit evaluation indicators are then weighted and combined to form a multi-objective evaluation indicator set.

[0159] A multi-objective evaluation function is constructed, and the index values ​​in the multi-objective evaluation index set are normalized. Based on the normalized index values, normalized progress weight coefficients and normalized cost weight coefficients are set. The normalized progress weight coefficients are multiplied by the normalized progress index values ​​to obtain a progress weighted score, and the normalized cost weight coefficients are multiplied by the normalized cost index values ​​to obtain a cost weighted score.

[0160] A multi-objective comprehensive evaluation is performed on the schedule-weighted score and the cost-weighted score to calculate a comprehensive benefit score; multiple optimization schemes are ranked according to the comprehensive benefit score; and the optimization scheme with the highest comprehensive benefit score is selected as the optimal execution scheme.

[0161] The optimal execution plan is converted into a standard data format to generate an execution data package containing a construction operation instruction and a resource allocation plan; the execution data package is then distributed to the site management terminal through the digital twin platform; the execution monitoring module is activated to collect actual execution data through the data acquisition equipment deployed at the site management terminal.

[0162] The multi-objective optimization execution method for engineering projects based on a digital twin platform includes the following steps:

[0163] First, a digital twin model of the project is established. This model contains data throughout the entire project lifecycle, such as design drawings, construction plans, resource information, and cost budgets. Taking a construction project as an example, the digital twin model includes a Building Information Model (BIM), a project schedule, a resource database (including manpower, materials, equipment, etc.), and a cost database.

[0164] Next, we will conduct scenario prediction and evaluation. Based on the digital twin model, we will generate multiple optimized execution plans. Each plan includes different schedules, resource allocations, and cost budgets. For example, Plan 1: Adopt assembly line operations, increase manpower input, and shorten the construction period; Plan 2: Optimize resource allocation, reduce material waste, and lower costs; Plan 3: Balance the construction period and cost to find the comprehensive optimal plan.

[0165] For each option, schedule and cost forecasts are performed. Using historical project data, expert experience, and forecasting models, the critical path duration, construction continuity, milestone completion status, total cost, cost structure, and cash flow are predicted for each option. For example, Option 1 predicts a 10% reduction in duration and a 5% increase in total cost; Option 2 predicts no change in duration but a 3% reduction in total cost; and Option 3 predicts a 5% reduction in duration and a 1% reduction in total cost.

[0166] Next, schedule and cost-benefit evaluation indicators are calculated. Based on the forecast results, schedule and cost-benefit indicators are calculated for each option. Schedule indicators include critical path duration reduction, construction continuity, and milestone compliance. Cost-benefit indicators include total cost savings, cost composition rationality, and cash flow matching. For example, Option 1 has a critical path duration reduction of 10 days, a construction continuity score of 0.9, and a milestone compliance of 100%; Option 2 has a total cost savings rate of 3%, a cost composition rationality score of 0.8, and a cash flow matching of 90%; Option 3 has a critical path duration reduction of 5 days, a construction continuity score of 0.95, a milestone compliance of 95%, a total cost savings rate of 1%, a cost composition rationality score of 0.9, and a cash flow matching of 95%.

[0167] Next, a multi-objective evaluation and scheme ranking are performed. The schedule benefit index and cost benefit index are weighted and combined to form a multi-objective evaluation index set. Schedule weight coefficients and cost weight coefficients are set, for example, a schedule weight of 0.6 and a cost weight of 0.4. The weight coefficients are multiplied by the index values ​​to obtain the schedule-weighted score and cost-weighted score for each scheme. The schedule-weighted score and cost-weighted score are added together to obtain the overall benefit score. For example, Scheme 1 has an overall benefit score of 0.85, Scheme 2 has an overall benefit score of 0.75, and Scheme 3 has an overall benefit score of 0.88. Based on the overall benefit score, the multiple schemes are ranked, and the scheme with the highest overall benefit score is selected as the optimal execution scheme. In this case, Scheme 3 has the highest overall benefit score, therefore, Scheme 3 is selected as the optimal execution scheme.

[0168] Finally, the execution plan is issued and monitored. The optimal execution plan is converted into an execution data package containing construction operation instructions and resource allocation plans. This execution data package is then distributed to the on-site management terminal via the digital twin platform. The execution monitoring module is activated, and actual execution data, such as construction progress, resource consumption, and cost expenditures, is collected through data acquisition devices deployed at the on-site management terminal. The collected actual execution data is compared and analyzed with the optimal execution plan to promptly identify deviations and make adjustments, ensuring the smooth progress of the project.

[0169] The solution proposed in this application can:

[0170] Improve project management efficiency. Digital twin technology enables digital management of the entire project lifecycle, improving information transmission efficiency and reducing communication costs, thereby enhancing project management efficiency. Optimize resource allocation. By conducting multi-objective evaluations of multiple options, the solution with the best overall benefits can be selected, achieving optimal resource allocation, reducing resource waste, and lowering project costs. Enhance the scientific basis of decision-making. Based on data analysis and predictive models, project execution can be monitored and evaluated in real time, providing a scientific basis for project decisions and improving the accuracy and effectiveness of those decisions.

[0171] Figure 2 This is a schematic diagram of the structure of the BIM-based dynamic management and optimization system for project progress and cost according to an embodiment of the present invention. Figure 2 As shown, the system includes:

[0172] The first unit is used to analyze the BIM model using computer vision algorithms to extract engineering spatial information and component attribute information, and establish an initial construction scene model; deploy visual sensors to collect construction implementation status data, deploy positioning sensors to collect construction personnel distribution data, and deploy IoT sensors to collect mechanical equipment operation data at the construction site; weight and combine the construction implementation status data, construction personnel distribution data, and mechanical equipment operation data to generate multi-source data; input the multi-source data into a pre-trained deep learning model for feature extraction and data fusion to generate real-time construction resource status data; dynamically map the real-time construction resource status data with the initial construction scene model to construct a digital twin basic platform with real-time perception capabilities;

[0173] The second unit is used to import construction schedule data and cost control benchmark data based on the digital twin platform to establish a schedule-cost correlation model; receive real-time status data of construction resources, calculate the current progress completion status and cost consumption status through the schedule-cost correlation model, and compare it with the construction schedule data and cost control benchmark data to obtain a comprehensive deviation value; automatically generate multiple sets of optimization schemes containing resource allocation parameters, process organization parameters, and cost control parameters based on the comprehensive deviation value; input the multiple sets of optimization schemes into the digital twin platform for parallel simulation, and obtain the simulation prediction results of each scheme, including schedule prediction results and cost prediction results;

[0174] The third unit is used to perform multi-objective evaluation of the progress prediction results and the cost prediction results, select the execution plan with the best overall benefits, distribute the optimal execution plan to the field management terminal through the digital twin basic platform, and start the plan execution monitoring module to continuously collect actual execution data. The actual execution data is compared and analyzed with the simulation prediction results in real time. When the comparison deviation exceeds a preset threshold, the progress-cost correlation model in the digital twin basic platform is updated using the actual execution data to realize the adaptive evolution of the digital twin environment, ensure the accuracy of the next round of simulation verification, and form a dynamic collaborative management and control of construction progress and cost through a digital twin.

[0175] A third aspect of the embodiments of the present invention,

[0176] An electronic device is provided, comprising:

[0177] processor;

[0178] Memory used to store processor-executable instructions;

[0179] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0180] Fourth aspect of the present invention,

[0181] A computer-readable storage medium is provided, having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0182] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0183] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A BIM-based project progress and cost dynamic management and control optimization method, characterized in that, The application relates to a construction progress and cost control method based on digital twin technology. The method comprises the following steps: a computer vision algorithm is used to analyze a BIM model to extract engineering space information and component attribute information, and an initial construction scene model is established; visual sensors are arranged on a construction site to collect construction implementation state data, positioning sensors are arranged to collect construction personnel distribution data, and Internet of Things sensors are arranged to collect mechanical equipment operation data, the construction implementation state data, the construction personnel distribution data and the mechanical equipment operation data are combined by weighting to generate multi-source data, the multi-source data is input into a pre-trained deep learning model for feature extraction and data fusion to generate construction resource real-time state data, and the construction resource real-time state data is dynamically mapped with the initial construction scene model to construct a digital twin basic platform with real-time sensing capability; based on the digital twin basic platform, construction progress plan data and cost control benchmark data are imported to establish a progress-cost correlation model, the construction resource real-time state data is received in real time, the current progress completion condition and cost consumption condition are calculated through the progress-cost correlation model, and the comprehensive deviation value is obtained by comparing the construction progress plan data and the cost control benchmark data, which comprises the following steps: the construction progress plan data and the cost control benchmark data are imported into the digital twin basic platform, the construction resource real-time state data collected on the construction site is subjected to feature extraction, work efficiency features are extracted from personnel data, energy consumption level features are extracted from equipment data, and loss rate features are extracted from material data, and the extracted work efficiency features, energy consumption level features and loss rate features are constructed into a resource configuration feature vector; a directed acyclic graph is constructed based on the logical dependency relationship between construction processes, process completion degree features and key path features are extracted from the directed acyclic graph, and the process completion degree features and the key path features are constructed into a process implementation feature vector; a multi-layer neural network is constructed as a progress-cost correlation model, an attention mechanism module is arranged in the middle layer of the progress-cost correlation model, the resource configuration feature vector and the process implementation feature vector are input into the progress-cost correlation model, the dynamic correlation strength between the resource configuration feature vector and the process implementation feature vector is calculated through the attention mechanism module, and a resource-process interaction matrix is generated; The resource-process interaction matrix is input into the schedule-cost correlation model to calculate the current construction schedule completion and cost consumption, and obtain schedule prediction data and cost prediction data; the schedule prediction data is compared with the construction schedule plan data to calculate a schedule deviation value, the cost prediction data is compared with the cost control benchmark data to calculate a cost deviation value, and a comprehensive deviation value is calculated based on the combination of the schedule deviation value and the cost deviation value; a plurality of sets of optimization schemes including resource allocation parameters, process organization parameters, and cost control parameters are automatically generated according to the comprehensive deviation value; the plurality of sets of optimization schemes are input into the digital twin basic platform for parallel simulation to obtain simulation prediction results of each scheme, wherein the simulation prediction results include schedule prediction results and cost prediction results; The schedule prediction results and the cost prediction results are subjected to multi-objective evaluation to select an optimal execution scheme; the optimal execution scheme is issued to a field management terminal through the digital twin basic platform to start a scheme execution monitoring module to continuously collect actual execution data; the actual execution data is compared with the simulation prediction results in real time, and when the comparison deviation exceeds a preset threshold, the actual execution data is used to update the schedule-cost correlation model in the digital twin basic platform, to realize adaptive evolution of the digital twin environment, ensure the accuracy of the next round of simulation verification, and form digital twin dynamic collaborative management of construction schedule and cost.

2. The method of claim 1, wherein, The schedule prediction data is compared with the construction schedule plan data to calculate a schedule deviation value, and the cost prediction data is compared with the cost control benchmark data to calculate a cost deviation value, including: Collecting basic data, including construction site real-time data and process plan completion data; extracting process actual completion data, resource usage data, and cost consumption data from the construction site real-time data, and storing the process actual completion data, the resource usage data, and the cost consumption data in a sliding time window according to a time sequence; Building a schedule prediction model, calculating a process completion rate based on the process actual completion data, and calculating the process completion rate by weighting the ratio of process actual completion quantities to the process plan completion data; analyzing the position of the process in the critical path to determine a position importance coefficient, analyzing the consumption of man, machine, and material to determine a resource consumption coefficient, and analyzing the dependency relationship of the process to determine a process correlation coefficient; Weighting the position importance coefficient, the resource consumption coefficient, and the process correlation coefficient to obtain a process weight coefficient; multiplying the process completion rate by the process weight coefficient to obtain schedule prediction data; and comparing the schedule prediction data with the process plan completion data to calculate a schedule deviation value; constructing a cost prediction model, calculating artificial cost, equipment cost and material cost based on the resource use data and the cost consumption data; obtaining the artificial cost based on the product of actual working hours and actual artificial unit price, obtaining the equipment cost based on the product of actual equipment use time and actual equipment rental rate, and obtaining the material cost based on the product of actual material consumption and actual material unit price; adopting Kalman filtering algorithm to perform noise smoothing processing on the artificial cost, the equipment cost and the material cost, combining the smoothed data to obtain cost prediction data; comparing the cost prediction data with the cost control benchmark data to obtain a cost deviation value.

3. The method of claim 1, wherein, automatically generating a plurality of sets of optimization schemes including resource allocation parameters, process organization parameters and cost control parameters based on the comprehensive deviation value; inputting the plurality of sets of optimization schemes into the digital twin basic platform for parallel simulation to obtain simulation prediction results of each scheme, the simulation prediction results including progress prediction results and cost prediction results including: the comprehensive deviation value includes progress deviation value and cost deviation value, constructing a progress optimization objective function based on the progress deviation value, and constructing a cost optimization objective function based on the cost deviation value; inputting the progress optimization objective function and the cost optimization objective function into an improved particle swarm algorithm to generate a plurality of sets of optimization schemes through iterative optimization; each set of optimization schemes includes resource allocation parameters, process organization parameters and cost control parameters, wherein the resource allocation parameters are used to optimize the number of construction teams, the number of equipment and the amount of material supply, the process organization parameters are used to optimize the process interleaving relationship, the process duration and the process resource demand, and the cost control parameters are used to optimize the artificial unit price, the equipment rate and the material unit price; constructing a construction site model, a resource flow model and a process implementation model in the digital twin basic platform; mapping the resource allocation parameters to the resource flow model, mapping the process organization parameters to the process implementation model, and mapping the cost control parameters to the construction site model; adopting a distributed computing architecture to distribute the mapped models to a plurality of computing nodes; performing parallel simulation in the plurality of computing nodes; performing resource scheduling simulation based on the resource flow model to obtain simulation resource utilization prediction data, performing process execution simulation based on the process implementation model to obtain simulation progress prediction data, and performing cost consumption simulation based on the construction site model to obtain simulation cost prediction data; combining the simulation resource utilization prediction data, the simulation progress prediction data and the simulation cost prediction data to form simulation prediction results.

4. The method of claim 3, wherein, constructing a progress optimization objective function based on the progress deviation value, and constructing a cost optimization objective function based on the cost deviation value includes: The K-means clustering algorithm is used to divide the progress deviation values into a deviation process group higher than a preset deviation threshold and a deviation process group lower than the preset deviation threshold, an analytic hierarchy process is used to construct a judgment matrix from three dimensions of process importance, process criticality and process duration for the deviation process group higher than the preset deviation threshold and the deviation process group lower than the preset deviation threshold; A geometric mean method is used to calculate an eigenvector of the judgment matrix to obtain a process weight coefficient; a product of the process weight coefficient and the progress deviation value is added, and a minimum value of an addition result is used to construct a progress optimization objective function; The progress optimization objective function calculation formula is as follows: ； wherein F s is the progress optimization objective function, m is the total number of processes, w i is the weight coefficient of the i-th process, d i is the progress deviation value of the i-th process; The K-means clustering algorithm is used to divide the cost deviation values into a deviation cost group higher than a preset deviation threshold and a deviation cost group lower than the preset deviation threshold, a Delphi method is used to perform expert scoring from three dimensions of cost proportion, cost controllability and cost sensitivity for the deviation cost group higher than the preset deviation threshold and the deviation cost group lower than the preset deviation threshold, and a cost weight coefficient is obtained through multiple rounds of expert scoring; The cost weight coefficient and the cost deviation value are multiplied and added, and a minimum value of an addition result is used to construct a cost optimization objective function; The cost optimization objective calculation formula is as follows: ； Wherein, F c is the cost optimization objective function, n is the total number of cost items, W j is the weight coefficient of the jth cost item, C j is the cost deviation value of the jth cost item.

5. The method of claim 3, wherein, The progress optimization objective function and the cost optimization objective function are input into an improved particle swarm algorithm, and multiple sets of optimization schemes are generated through iterative optimization, including: An adaptive weight coefficient is calculated according to the progress deviation value and the cost deviation value, an absolute value of a deviation coefficient is obtained by weighting an absolute value of the progress deviation value and an absolute value of the cost deviation value, a progress weight coefficient in the adaptive weight coefficient is obtained based on a ratio of the absolute value of the progress deviation value to the absolute value of the deviation coefficient, and a cost weight coefficient in the adaptive weight coefficient is obtained based on a ratio of the absolute value of the cost deviation value to the absolute value of the deviation coefficient; An initial particle swarm is generated by using a Logistics chaotic mapping, a particle swarm position vector is constructed by mapping the initial particle swarm to a decision space, a population aggregation degree is obtained by calculating a distance between the particle swarm position vector and a population center of gravity, and an adaptive mutation probability is set based on the population aggregation degree, which increases exponentially with an increase of the population aggregation degree; The progress optimization objective function and the cost optimization objective function are input into the improved particle swarm algorithm; an inertia weight is updated in a nonlinear decreasing manner, which decreases in a cosine function law with an increase of an iteration number; a learning factor is updated in an adaptive manner, including an individual learning factor and a group learning factor, wherein the individual learning factor increases in an exponential function law with an increase of the iteration number, and the group learning factor decreases in an exponential function law with an increase of the iteration number; In each iteration, the constraint violation degree is calculated based on the particle swarm position vector, the fitness value is obtained by weighting the constraint violation degree, the progress optimization objective function and the cost optimization objective function using the penalty function method, the crowding distance of each particle in the population is calculated, and the solution with the crowding distance greater than a preset distance threshold is selected as an elite individual reserved to the next generation; The local search result of the elite individual is obtained by performing local search on the elite individual using a pattern search method, the step size of the pattern search method decreases exponentially with the number of iterations, the global optimal solution is updated based on the local search result, and it is judged whether the termination condition is met; When the termination condition is met, the global optimal solution is output as an optimization scheme.

6. The method of claim 1, wherein, The progress prediction result and the cost prediction result are multi-objective evaluated, and an optimal execution scheme is selected; The optimal execution scheme is issued to the field management terminal through the digital twin basic platform to start the scheme execution monitoring module to continuously collect actual execution data, including: The progress benefit evaluation index is calculated based on the progress prediction result, the progress benefit evaluation index includes the key path duration compression amount index, the construction process continuity index and the milestone node compliance index, the cost benefit evaluation index is calculated according to the cost prediction result, the cost benefit evaluation index includes the total cost saving rate index, the cost composition rationality index and the cash flow matching degree index, and the progress benefit evaluation index and the cost benefit evaluation index are combined to form a multi-objective evaluation index set; A multi-objective evaluation function is constructed, and the index values in the multi-objective evaluation index set are normalized; The normalized progress weight coefficient and the normalized cost weight coefficient are set based on the normalized index values; The progress weighted score is obtained by multiplying the normalized progress weight coefficient and the normalized progress index value, and the cost weighted score is obtained by multiplying the normalized cost weight coefficient and the normalized cost index value; The progress weighted score and the cost weighted score are multi-objective comprehensive evaluated, and the comprehensive benefit score is calculated; The optimization schemes are sorted according to the comprehensive benefit score; The optimization scheme with the highest comprehensive benefit score is selected as the optimal execution scheme; The optimal execution scheme is converted into a standard data format to generate an execution data package containing a construction operation guide and a resource allocation plan; The execution data package is issued to the field management terminal through the digital twin basic platform; The scheme execution monitoring module is started, and the actual execution data is collected through the data acquisition equipment deployed by the field management terminal.

7. A BIM-based project progress and cost dynamic management and optimization system for implementing the method of any one of the preceding claims 1-6, characterized in that, The first unit is configured to analyze a BIM model using a computer vision algorithm to extract engineering space information and component attribute information, and to establish an initial construction scene model. The second unit is configured to perform multi-objective optimization on the initial construction scene model to obtain a progress prediction result and a cost prediction result. In the construction site, visual sensors are deployed to collect construction implementation state data, positioning sensors are deployed to collect construction personnel distribution data, and Internet of Things sensors are deployed to collect mechanical equipment operation data. The construction implementation state data, the construction personnel distribution data, and the mechanical equipment operation data are combined to generate multi-source data. The multi-source data is input into a pre-trained deep learning model for feature extraction and data fusion to generate real-time construction resource state data. The real-time construction resource state data is dynamically mapped to the initial construction scene model to build a digital twin basic platform with real-time sensing capability. A second unit is configured to import construction progress plan data and cost control benchmark data based on the digital twin basic platform, establish a progress-cost correlation model, and receive real-time construction resource state data. The current progress completion and cost consumption are calculated through the progress-cost correlation model, and the comprehensive deviation value is obtained by comparing the construction progress plan data and the cost control benchmark data. The multi-group optimization schemes including resource allocation parameters, process organization parameters, and cost control parameters are automatically generated based on the comprehensive deviation value. The multi-group optimization schemes are input into the digital twin basic platform for parallel simulation to obtain simulation prediction results of each scheme, including progress prediction results and cost prediction results. A third unit is configured to perform multi-objective evaluation on the progress prediction results and the cost prediction results, and select an optimal execution scheme with the best comprehensive benefit. The optimal execution scheme is sent to the field management terminal through the digital twin basic platform to start the scheme execution monitoring module to continuously collect actual execution data. The actual execution data is compared with the simulation prediction results in real time. When the comparison deviation exceeds the preset threshold, the actual execution data is used to update the progress-cost correlation model in the digital twin basic platform, realize the adaptive evolution of the digital twin environment, ensure the accuracy of the next round of simulation verification, and form the digital twin dynamic collaborative management of construction progress and cost.

8. An electronic device, comprising: It includes: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to invoke the instructions stored in the memory to perform the method of any one of claims 1 to 6.

9. A computer-readable storage medium having stored thereon computer program instructions, wherein, The computer program instructions are executed by the processor to implement the method of any one of claims 1 to 6.

Images