Infrastructure engineering project management system based on big data

Abstract

The invention relates to the field of construction engineering project management, in particular to an infrastructure engineering project management system based on big data. The method comprises the steps of obtaining multi-modal engineering project data in a system, performing space-time alignment on the multi-modal engineering project data, processing abnormal values in the aligned data to obtain complete engineering project data, performing multi-modal fusion on the complete engineering project data, and constructing a feature library of the fused engineering project data. And establishing a ResNet-Transform risk prediction model, optimizing hyper-parameters of the model by using an IWOA improved whale optimization algorithm, inputting the feature engineering project data into the target risk prediction model for prediction, and performing construction planning based on the construction engineering project state. The problem of data islands is solved, a complete data set covering engineering total elements is formed, the system can automatically trigger a hierarchical response strategy, and closed-loop management from risk identification to planning adjustment is achieved.

Description

Technical Field

[0001] This invention relates to the field of engineering construction technology, and in particular to a big data-based infrastructure project management system. Background Technology

[0002] In the field of engineering construction, the construction process involves multi-dimensional data interaction and complex risk management, posing significant challenges to traditional management methods. Currently, project data exists in multiple independent modalities, such as BIM models, sensor monitoring, supply chain records, meteorological and environmental data, and project logs. This heterogeneous data format and inconsistent spatiotemporal benchmarks lead to prominent data silos, making it difficult to form a comprehensive understanding of the project's status. Furthermore, traditional risk prediction models struggle to simultaneously extract spatial features and address time-series dependencies, and their limited generalization capabilities result in delayed adjustments to construction planning, failing to adapt to dynamically changing engineering environments. As engineering complexity increases, integrating multimodal data, constructing high-precision prediction models, and achieving intelligent construction planning have become pressing technical challenges for the industry. Summary of the Invention

[0003] The purpose of this invention is to solve the above problems by designing a big data-based infrastructure project management system.

[0004] To achieve the above objectives, the technical solution of the present invention further includes the following modules in the aforementioned infrastructure project management system based on big data:

[0005] The engineering data acquisition module is used to acquire multimodal engineering project data in the system, including at least BIM model data, IoT sensor data, supply chain data, geographic and meteorological data, and engineering log text.

[0006] The data preprocessing module is used to perform spatiotemporal alignment on the multimodal engineering project data, process outliers in the aligned data to obtain complete engineering project data, and perform multimodal fusion on the complete engineering project data to obtain fused engineering project data.

[0007] The data feature fusion module is used to construct a feature library of the fused engineering project data, which includes at least physical features, behavioral features and derived features, to obtain feature engineering project data.

[0008] The risk model building module is used to build a ResNet-Transformer risk prediction model. It uses the improved whale optimization algorithm of IWOA to optimize the hyperparameters of the model and obtain the target IWOA-ResNet-Transformer risk prediction model.

[0009] The project construction planning module is used to input the characteristic engineering project data into the target IWOA-ResNet-Transformer risk prediction model for prediction, obtain the construction project status, and carry out construction planning based on the construction project status.

[0010] Furthermore, in the aforementioned infrastructure project management system based on big data, the project data acquisition module includes the following sub-modules:

[0011] The BIM model data extraction submodule is used to extract geometric attribute data, material parameters, and construction sequence logical relationship data of building components.

[0012] The IoT sensor data acquisition submodule is used to collect environmental data, structural data, and equipment data. The environmental data includes wind speed, temperature and humidity, and rainfall. The structural data includes concrete strain, tower crane tilt angle, and foundation settlement. The equipment data includes excavator hydraulic pressure and pump truck flow rate.

[0013] The supply chain data acquisition submodule is used to acquire material inventory data, GPS tracks of transport vehicles, and supplier credit rating data.

[0014] The geographic meteorological data acquisition submodule is used to acquire radar detection results data, wind speed probability distribution for the next 72 hours, and tide tables.

[0015] Furthermore, in the aforementioned infrastructure project management system based on big data, the data preprocessing module includes the following sub-modules:

[0016] The spatiotemporal alignment submodule is used to synchronize the device clocks in the multimodal engineering project data using the DTW dynamic time warping algorithm, convert the BIM model coordinate system into the WGS84 geographic coordinate system, and match it with the sensor data space to obtain aligned engineering project data.

[0017] The outlier submodule is used to detect abnormal sensor data in the aligned project data using the 3σ principle, and to use an LSTM neural network to identify faults in the mechanical record text data in the data, thereby deleting outliers and fault data to obtain complete project data.

[0018] The data fusion submodule is used to establish a unified spatiotemporal index by associating BIM components, sensor readings and construction logs with timestamps and projects as keys. It uses the GraphSAGE algorithm to embed unstructured text into a 128-dimensional vector and inputs it together with the sensor numerical data into a graph neural network to obtain fused engineering project data.

[0019] Furthermore, in the aforementioned infrastructure project management system based on big data, the data feature fusion module includes the following sub-modules:

[0020] A submodule is constructed to build a feature library of the fused engineering project data. The physical features in the fused engineering project data are extracted using an improved Arrhenius equation, and the calculation formula is as follows;

[0021] ;

[0022] in, Represents the material property constants in an engineering project. Indicates the activation energy of the hydration reaction. Represents the ideal gas constant. Indicates time Real-time temperature, Indicates a correction term. This indicates the effect of curing temperature and time on strength.

[0023] The indicator submodule is used to obtain dynamic indicators reflecting the interaction behavior of people, machines and materials during the construction process from the integrated engineering project data based on the process network topology and real-time progress calculation, and to obtain behavioral characteristics.

[0024] The acquisition submodule is used to calculate the integrated engineering project data to obtain supply chain resilience indicators and spatiotemporal correlation characteristics, and to obtain derived features.

[0025] The calculation submodule is used to perform streaming calculations on the physical features, behavioral features, and derived features using the sliding window method to obtain feature engineering project data.

[0026] Furthermore, in the aforementioned infrastructure project management system based on big data, the risk model establishment module includes the following sub-modules:

[0027] A submodule is established to build a ResNet-Transformer risk prediction model. This module acquires historical engineering project data from a feature library and uses the model's input layer to map the physical, behavioral, and derived features from the historical engineering project data into high-dimensional vectors. ;

[0028] ;

[0029] in, This indicates a feature concatenation operation. This represents a learnable embedding matrix, used to map the concatenated features to a high-dimensional space; , This represents the physical, behavioral, and derived characteristics of historical engineering project data. Indicates the bias term;

[0030] The residual submodule is used to extract local spatiotemporal patterns using ResNet residual blocks in the ResNet-Transformer risk prediction model.

[0031] ;

[0032] in, Residual functions containing Conv1D convolutional layers and activation functions, Indicates the th ResNet residual block Layer output; Representing the residual function The weight parameters included;

[0033] The capture submodule is used to capture long-range dependencies between features using Multi-Head Attention in the Transformer encoding layer of the ResNet-Transformer risk prediction model.

[0034] ;

[0035] in, Indicates a causal mask. Represents the query vector. Represents the key vector. Represents a value vector; Represents the key vector The dimension used to scale the dot product. ;

[0036] The enhancement submodule is used to enhance the nonlinearity during data training based on the FFN feedforward network in the model:

[0037] ;

[0038] in, This represents the input to the FFN feedforward network. This represents the weight matrix, which participates in two linear transformations. and This represents the bias term, which adds an offset to the model in each of the two linear transformations; This represents the activation function, which introduces nonlinear characteristics into the model.

[0039] Furthermore, in the aforementioned infrastructure project management system based on big data, the risk model establishment module includes the following units:

[0040] The population generation unit is used to optimize the hyperparameters of the model using the improved whale optimization algorithm of IWOA. It randomly generates a whale population, where each individual in the whale population represents a set of model hyperparameter combinations and a pre-trained ResNet-Transformer model skeleton.

[0041] The loss function computation unit is used to compute the loss function of each hyperparameter combination, including at least the number of Transformer heads, the number of residual block layers, the learning rate, and the batch size, on the validation set. ;

[0042] ;

[0043] in, This represents the cross-entropy loss in binary classification. Indicates task weight, Represents the L2 regularization coefficient. This represents the combination of model hyperparameters. Indicates the calculation of the first Real labels in each task With predictive labels The differences between them This represents the squared L2 norm of the model hyperparameter combination;

[0044] The model fine-tuning unit is used to update the whale's position according to the IWOA rules, retain the historical best solution, and trigger Cauchy mutation if the best fitness does not improve for 10 consecutive generations. The model is fine-tuned by selecting the hyperparameter combination with the highest fitness to obtain the target IWOA-ResNet-Transformer risk prediction model.

[0045] Furthermore, in the aforementioned infrastructure project management system based on big data, the project construction planning module includes the following units:

[0046] The indicator prediction unit is used to input the characteristic engineering project data into the target IWOA-ResNet-Transformer risk prediction model for prediction, and obtain multi-dimensional status indicators of the construction project, including at least risk level, status parameters and trend prediction.

[0047] The status judgment unit is used to integrate the risk level, status parameters and trend predictions output by the model to obtain a project status report. According to the engineering management standards, the project status is divided into three levels: normal status, early warning status and emergency status.

[0048] The construction planning unit is used to plan construction based on the project status. If it is in an early warning state, it assesses available resources and adjusts the priority of local processes. If it is in an emergency state, it suspends regional construction and recalculates the BIM model.

[0049] Furthermore, in the method for implementing the above-mentioned infrastructure project management system based on big data, the method includes the following steps:

[0050] Acquire multimodal engineering project data from the system, including at least BIM model data, IoT sensor data, supply chain data, geographic and meteorological data, and engineering log text;

[0051] The multimodal engineering project data is spatiotemporally aligned, outliers in the aligned data are processed to obtain complete engineering project data, and multimodal fusion is performed on the complete engineering project data to obtain fused engineering project data.

[0052] Construct a feature library for the fused engineering project data, including at least physical features, behavioral features, and derived features, to obtain feature engineering project data;

[0053] A ResNet-Transformer risk prediction model is established, and the hyperparameters of the model are optimized using the improved whale optimization algorithm of IWOA to obtain the target IWOA-ResNet-Transformer risk prediction model.

[0054] The characteristic engineering project data is input into the target IWOA-ResNet-Transformer risk prediction model for prediction to obtain the construction project status, and construction planning is carried out based on the construction project status.

[0055] Furthermore, in the aforementioned infrastructure project management method based on big data, the method includes the following steps:

[0056] Extract geometric attribute data, material parameters, and logical relationship data of construction procedures from building components;

[0057] Collect environmental data, structural data, and equipment data. The environmental data includes wind speed, temperature, humidity, and rainfall. The structural data includes concrete strain, tower crane tilt angle, and foundation settlement. The equipment data includes excavator hydraulic pressure and pump truck flow rate.

[0058] Obtain material inventory data, GPS tracks of transport vehicles, and supplier credit rating data;

[0059] Acquire radar detection data, wind speed probability distribution for the next 72 hours, and tide tables.

[0060] Furthermore, in the aforementioned infrastructure project management method based on big data, the step of acquiring real-time voice data from the environment via a microphone array includes:

[0061] The device clocks in the multimodal engineering project data are synchronized by using the DTW dynamic time warping algorithm, the BIM model coordinate system is converted into the WGS84 geographic coordinate system, and the sensor data space is matched to obtain aligned engineering project data.

[0062] The sensor abnormal data of the aligned project data is detected using the 3σ principle, and the mechanical record text data in the data is identified by using an LSTM neural network. The abnormal values ​​and fault data are deleted to obtain the complete project data.

[0063] A unified spatiotemporal index is established by using timestamps and projects to associate key-value BIM components, sensor readings, and construction logs. The GraphSAGE algorithm is used to embed unstructured text into a 128-dimensional vector, which is then input into a graph neural network along with the sensor numerical data to obtain fused engineering project data.

[0064] Its beneficial effects lie in unifying heterogeneous BIM models, sensor data, and supply chain data into a standardized framework through spatiotemporal alignment and outlier handling, solving the data silo problem and providing a high-quality data foundation. A comprehensive feature library containing physical, behavioral, and derived features is constructed. Combined with the ResNet-Transformer model, it can extract spatial structural features through ResNet and capture temporal process connection patterns using Transformer, achieving multi-scale characterization of engineering status. The introduction of the improved IWOA whale optimization algorithm further enhances the model's hyperparameter search efficiency, improving prediction accuracy compared to traditional optimization methods. It can automatically trigger hierarchical response strategies, achieving closed-loop management from risk identification to planning adjustments. It reduces reliance on human experience, improving management efficiency and precision. Real-time access to engineering logs and sensor data, along with iterative model optimization, enables the system to learn independently and adapt to complex scenarios in different projects. Attached Figure Description

[0065] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.

[0066] Figure 1 This is a schematic diagram of the first embodiment of the infrastructure construction project management system based on big data in this invention.

[0067] Figure 2 This is a schematic diagram of the second embodiment of the infrastructure construction project management system based on big data in this invention.

[0068] Figure 3This is a schematic diagram of the third embodiment of the infrastructure project management system based on big data in this invention. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0070] Those skilled in the art will understand that, unless otherwise stated, the singular forms “a,” “an,” and “” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0071] The present invention will now be described in detail with reference to the accompanying drawings, such as... Figure 1 As shown, the infrastructure project management system based on big data includes the following modules:

[0072] 101. Engineering data acquisition module, used to acquire multimodal engineering project data in the system, including at least BIM model data, IoT sensor data, supply chain data, geographic and meteorological data, and engineering log text;

[0073] Specifically, this implementation also includes a BIM model data extraction submodule, which is used to extract geometric attribute data, material parameters and construction procedure logical relationship data of building components;

[0074] The IoT sensor data acquisition submodule is used to collect environmental data, structural data and equipment data. Environmental data includes wind speed, temperature and humidity and rainfall. Structural data includes concrete strain, tower crane tilt angle and foundation settlement. Equipment data includes excavator oil pressure and pump truck flow rate.

[0075] The supply chain data acquisition submodule is used to acquire material inventory data, GPS tracks of transport vehicles, and supplier credit rating data.

[0076] The geographic meteorological data acquisition submodule is used to acquire radar detection results data, wind speed probability distribution for the next 72 hours, and tide tables.

[0077] 102. Data preprocessing module, used to perform spatiotemporal alignment on multimodal engineering project data, process outliers in the aligned data to obtain complete engineering project data, and perform multimodal fusion on the complete engineering project data to obtain fused engineering project data;

[0078] Specifically, this implementation also includes a spatiotemporal alignment submodule, which is used to synchronize the device clocks in the multimodal engineering project data through the DTW dynamic time warping algorithm, convert the BIM model coordinate system into the WGS84 geographic coordinate system, match it with the sensor data space, and obtain aligned engineering project data.

[0079] The outlier submodule is used to detect abnormal sensor data in aligned project data using the 3σ principle, and to use an LSTM neural network to identify faults in the mechanical record text data in the data, thereby deleting outliers and fault data to obtain complete project data.

[0080] The data fusion submodule is used to establish a unified spatiotemporal index by associating BIM components, sensor readings and construction logs with timestamps and projects as keys. It uses the GraphSAGE algorithm to embed unstructured text into a 128-dimensional vector and inputs it together with the sensor numerical data into a graph neural network to obtain fused engineering project data.

[0081] 103. Data feature fusion module, used to build a feature library of fused engineering project data, including at least physical features, behavioral features and derived features, to obtain feature engineering project data;

[0082] Specifically, this implementation also includes a construction submodule, which is used to build a feature library of integrated engineering project data. The physical features in the integrated engineering project data are extracted through an improved Arrhenius equation, and the calculation formula is as follows;

[0083] ;

[0084] in, Represents the material property constants in an engineering project. Indicates the activation energy of the hydration reaction. Represents the ideal gas constant. Indicates time Real-time temperature, Indicates a correction term. This indicates the effect of curing temperature and time on strength.

[0085] The indicator submodule is used to obtain dynamic indicators reflecting the interaction behavior of people, machines and materials during the construction process from the integrated engineering project data based on the process network topology and real-time progress calculation, and to obtain behavioral characteristics.

[0086] The acquisition submodule is used to calculate supply chain resilience indicators and spatiotemporal correlation characteristics by integrating engineering project data, and obtain derived features.

[0087] The computation submodule is used to perform streaming computation on physical features, behavioral features, and derived features using the sliding window method to obtain feature engineering project data.

[0088] 104. Risk Model Building Module: This module is used to build a ResNet-Transformer risk prediction model. It optimizes the hyperparameters of the model using the improved whale optimization algorithm of IWOA to obtain the target IWOA-ResNet-Transformer risk prediction model.

[0089] Specifically, this implementation also includes establishing a submodule for building a ResNet-Transformer risk prediction model, acquiring historical engineering project data from the feature library, and using the model's input layer to map the physical features, behavioral features, and derived features from the historical engineering project data into high-dimensional vectors. ;

[0090] ;

[0091] in, This indicates a feature concatenation operation. This represents a learnable embedding matrix, used to map the concatenated features to a high-dimensional space; , This represents the physical, behavioral, and derived characteristics of historical engineering project data. Indicates the bias term;

[0092] The residual submodule is used to extract local spatiotemporal patterns using ResNet residual blocks in the ResNet-Transformer risk prediction model.

[0093] ;

[0094] in, Residual functions containing Conv1D convolutional layers and activation functions, Indicates the th ResNet residual block Layer output; Representing the residual function The weight parameters included;

[0095] The capture submodule is used to capture long-range dependencies between features using Multi-Head Attention in the Transformer encoding layer of the ResNet-Transformer risk prediction model.

[0096] ;

[0097] in, Indicates a causal mask. Represents the query vector. Represents the key vector. Represents a value vector; Represents the key vector The dimension used to scale the dot product. ;

[0098] The enhancement submodule is used to enhance the nonlinearity during data training based on the FFN feedforward network in the model:

[0099] ;

[0100] in, This represents the input to the FFN feedforward network. This represents the weight matrix, which participates in two linear transformations. and This represents the bias term, which adds an offset to the model in each of the two linear transformations; This represents the activation function, which introduces nonlinear characteristics into the model.

[0101] The population generation unit is used to optimize the hyperparameters of the model using the improved whale optimization algorithm of IWOA. It randomly generates a whale population, where each individual in the whale population represents a set of model hyperparameter combinations and a pre-trained ResNet-Transformer model skeleton.

[0102] The loss function computation unit is used to compute the loss function of each hyperparameter combination, including at least the number of Transformer heads, the number of residual block layers, the learning rate, and the batch size, on the validation set. ;

[0103] ;

[0104] in, This represents the cross-entropy loss in binary classification. Indicates task weight, Represents the L2 regularization coefficient. This represents the combination of model hyperparameters. Indicates the calculation of the first Real labels in each task With predictive labels The differences between them This represents the squared L2 norm of the model hyperparameter combination;

[0105] The model fine-tuning unit is used to update the whale's position according to the IWOA rules, retain the historical best solution, and trigger Cauchy mutation if the best fitness does not improve for 10 consecutive generations. The model is fine-tuned by selecting the hyperparameter combination with the highest fitness to obtain the target IWOA-ResNet-Transformer risk prediction model.

[0106] Improving the whale optimization algorithm involves enhancing hyperparameter search efficiency and avoiding getting trapped in local optima. The improved strategies include: 1. Dynamically adjusting the search step size and using an adaptive inertia weight adjustment strategy to balance global exploration and local development; 2. Applying Cauchy perturbation to the optimal solution and using a Cauchy mutation strategy to enhance the ability to escape local optima; 3. Introducing a dynamic spiral encirclement mechanism with inertia weights and mutation factors when updating the whale position.

[0107] The improved strategy also includes optimizing the objective function with hyperparameters to minimize multi-task prediction error and model complexity.

[0108] 105. Project construction planning module, used to input characteristic engineering project data into the target IWOA-ResNet-Transformer risk prediction model for prediction, obtain the construction project status, and carry out construction planning based on the construction project status.

[0109] Specifically, this implementation also includes,

[0110] The indicator prediction unit is used to input the characteristic engineering project data into the target IWOA-ResNet-Transformer risk prediction model for prediction, and obtain multi-dimensional status indicators of the construction project, including at least risk level, status parameters and trend prediction.

[0111] The status judgment unit is used to integrate the risk level, status parameters and trend predictions output by the model to obtain a project status report. According to the engineering management standards, the project status is divided into three levels: normal status, early warning status and emergency status.

[0112] The construction planning unit is used to plan construction based on the project status. If it is in an early warning state, it assesses the available resources and adjusts the priority of local processes. If it is in an emergency state, it suspends construction in the area and recalculates the BIM model.

[0113] I. Status Analysis and Classification of Construction Projects

[0114] Multi-dimensional status integration: The risk level, key parameters, trend predictions and other information output by the model are integrated into an intuitive status report. For example, the risk heat map, progress Gantt chart deviation, resource supply and demand balance table of each construction area can be displayed through the dashboard.

[0115] Status Classification and Early Warning: Based on engineering management standards, project status is divided into three levels:

[0116] Normal state: All key indicators are within the preset thresholds (e.g., schedule deviation ≤5%, safety risk score <60 points), with no significant abnormal trends;

[0117] Warning status: Some indicators are close to or exceed the threshold (such as progress deviation of 5%-15%, equipment failure rate ≥10%), which require close monitoring;

[0118] Emergency: Key indicators are severely exceeded (e.g., schedule deviation > 15%, safety risk score ≥ 80 points), or trends indicate that the risk may deteriorate rapidly.

[0119] Abnormal status location: For early warning or emergency situations, trace back to the specific source of the abnormality through the feature library (such as a conflict between the structural design shown in the BIM model of a certain area and the data from on-site IoT sensors, or a delay in delivery by a material supplier shown in the supply chain data), and identify the link in which the problem occurred (design, construction, supply, environment, etc.).

[0120] II. State-Based Construction Planning Adjustment

[0121] Develop a preliminary adjustment plan: Based on the project status classification, the engineering management team will develop targeted measures.

[0122] Normal state: Maintain the existing construction plan and strengthen daily monitoring (such as increasing the frequency of project log recording and regularly calibrating sensor data).

[0123] Warning status:

[0124] For delays: assess available resources (such as increasing construction teams or extending working hours) and adjust the priority of local processes (such as prioritizing tasks on the critical path).

[0125] To address safety risks: increase the frequency of on-site inspections and add temporary protective measures for high-risk areas (such as deep foundation pits and high-altitude work areas);

[0126] In response to resource shortages: initiate the procurement process of alternative suppliers and stockpile materials that are likely to be in short supply in advance (such as stockpiling cement in advance based on geographical and meteorological data predicting that heavy rain may affect transportation).

[0127] State of emergency:

[0128] Suspend high-risk operations: Immediately stop work and investigate the problem (if abnormal structural stress is found, suspend construction in that area and recalculate the BIM model).

[0129] Reorganize resource allocation: Utilize emergency reserve resources (such as activating backup equipment and coordinating idle manpower from other projects to provide support);

[0130] Adjust the overall plan: Communicate with the owner and supervision unit to revise the construction schedule (if the progress is seriously delayed due to extreme weather, apply for an extension of the construction period or shorten the time for non-critical tasks).

[0131] Feasibility assessment: A multidisciplinary team, including design, construction, materials, and safety professionals, will review the proposed adjustments, focusing on verifying the following:

[0132] Technical feasibility (e.g., whether the adjusted procedures comply with specifications, whether the BIM model supports the new design);

[0133] Resource matching (whether existing inventory / supply chain can meet new demand, and whether there are conflicts in personnel and equipment scheduling);

[0134] Impact on cost and schedule (whether the additional costs of implementing the solution are within budget, and whether the schedule adjustment meets the contract requirements).

[0135] Approval and Release of the Plan: The approved plan will be submitted to the project leader for approval. After approval, it will be issued to each construction team in the form of an official document (such as a construction liaison form or a schedule adjustment plan), which will specify the adjustment content, responsible persons and time nodes (such as "Section 3 needs to complete the procurement of spare materials within 48 hours and add 20 construction workers within 5 days").

[0136] III. Construction Planning Implementation and Dynamic Monitoring

[0137] Implementation of the adjustment plan: Each responsible party shall implement the adjustment measures according to the plan (such as the materials department purchasing materials, the safety department adding protective facilities, and the construction team adjusting the work sequence), and record the actual implementation status (such as the material arrival time and the number of personnel on site) in the project log.

[0138] Real-time data feedback and model iteration:

[0139] Regularly (e.g., daily / weekly), input newly generated engineering logs, IoT sensor data, and supply chain update data into the model to verify whether the current status is improving in the expected direction (e.g., whether the schedule deviation has narrowed or the risk level has decreased).

[0140] If the adjusted status does not meet expectations (e.g., the warning status is not lifted for 3 consecutive days), re-analyze the model prediction results, investigate loopholes in the implementation of the plan (e.g., inadequate resource allocation, incomplete handling of abnormal data), and restart the planning adjustment process if necessary.

[0141] Document archiving and experience accumulation: The results of this status analysis, adjustment plan and execution effect will be entered into the project management system to form a historical case library, which will provide a reference for risk prediction and planning of similar projects in the future (such as supply chain risk response strategies under a certain type of geographical and meteorological data).

[0142] Its beneficial effects lie in unifying heterogeneous BIM models, sensor data, and supply chain data into a standardized framework through spatiotemporal alignment and outlier handling, thus solving the data silo problem and providing a high-quality data foundation. A comprehensive feature library encompassing physical, behavioral, and derived features is constructed. Combined with the ResNet-Transformer model, it can extract spatial structural features through ResNet and capture temporal process connection patterns using Transformer, achieving multi-scale characterization of engineering status. The introduction of an improved IWOA whale optimization algorithm further enhances the model's hyperparameter search efficiency, improving prediction accuracy compared to traditional optimization methods. It can automatically trigger hierarchical response strategies, achieving closed-loop management from risk identification to planning adjustments. It reduces reliance on human experience, improving management efficiency and precision. Real-time access to engineering logs and sensor data, along with iterative model optimization, enables the system to learn independently and adapt to complex scenarios in different projects.

[0143] In this embodiment, please refer to Figure 2 The second embodiment of the infrastructure project management system based on big data in this invention includes a data feature fusion module comprising the following sub-modules:

[0144] The construction submodule is used to build a feature library of integrated engineering project data. The physical features in the integrated engineering project data are extracted through the improved Arrhenius equation. The calculation formula is as follows;

[0145]

[0146] in, Represents the material property constants in an engineering project. Indicates the activation energy of the hydration reaction. Represents the ideal gas constant. Indicates time Real-time temperature, Indicates a correction term. This indicates the effect of curing temperature and time on strength.

[0147] The indicator submodule is used to obtain dynamic indicators reflecting the interaction behavior of people, machines and materials during the construction process from the integrated engineering project data based on the process network topology and real-time progress calculation, and to obtain behavioral characteristics.

[0148] The acquisition submodule is used to calculate supply chain resilience indicators and spatiotemporal correlation characteristics by integrating engineering project data, and obtain derived features.

[0149] The computation submodule is used to perform streaming computation on physical features, behavioral features, and derived features using the sliding window method to obtain feature engineering project data.

[0150] Its beneficial effects lie in unifying heterogeneous BIM models, sensor data, supply chain data, and other data into a standardized framework through spatiotemporal alignment and outlier handling. This solves the problem of data silos, forming a complete dataset covering all elements of the project and providing a high-quality data foundation for accurate analysis.

[0151] In this embodiment, please refer to Figure 3 The third embodiment of the infrastructure construction project management system based on big data in this invention includes a project construction planning module comprising the following sub-units:

[0152] The indicator prediction unit is used to input the characteristic engineering project data into the target IWOA-ResNet-Transformer risk prediction model for prediction, and obtain multi-dimensional status indicators of the construction project, including at least risk level, status parameters and trend prediction.

[0153] The status judgment unit is used to integrate the risk level, status parameters and trend predictions output by the model to obtain a project status report. According to the engineering management standards, the project status is divided into three levels: normal status, early warning status and emergency status.

[0154] The construction planning unit is used to plan construction based on the project status. If it is in an early warning state, it assesses the available resources and adjusts the priority of local processes. If it is in an emergency state, it suspends construction in the area and recalculates the BIM model.

[0155] Its beneficial effects lie in the fact that, based on real-time project status predictions (such as schedule deviations and resource shortage risk levels), the system can automatically trigger tiered response strategies (such as adjusting process priorities and activating backup supply chains), achieving closed-loop management from risk identification to planning adjustments. When the model predicts that heavy rain may cause delays in material transportation, inventory resources can be allocated in advance and the construction sequence optimized.

[0156] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various modifications and alterations can be made without departing from the spirit and scope of the invention, and all such modifications and alterations fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A big data-based infrastructure project management system, characterized in that: The infrastructure construction project management system includes the following modules: The engineering data acquisition module is used to acquire multimodal engineering project data in the system, including at least BIM model data, IoT sensor data, supply chain data, geographic and meteorological data, and engineering log text. The data preprocessing module is used to perform spatiotemporal alignment on the multimodal engineering project data, process outliers in the aligned data to obtain complete engineering project data, and perform multimodal fusion on the complete engineering project data to obtain fused engineering project data. The data feature fusion module is used to construct a feature library of the fused engineering project data, which includes at least physical features, behavioral features and derived features, to obtain feature engineering project data. The risk model building module is used to build a ResNet-Transformer risk prediction model. It uses the improved whale optimization algorithm of IWOA to optimize the hyperparameters of the model and obtain the target IWOA-ResNet-Transformer risk prediction model. The project construction planning module is used to input the characteristic engineering project data into the target IWOA-ResNet-Transformer risk prediction model for prediction, obtain the construction project status, and carry out construction planning based on the construction project status.

2. The infrastructure project management system based on big data as described in claim 1, characterized in that, The engineering data acquisition module includes the following sub-modules: The BIM model data extraction submodule is used to extract geometric attribute data, material parameters, and construction sequence logical relationship data of building components. The IoT sensor data acquisition submodule is used to collect environmental data, structural data, and device data. The supply chain data acquisition submodule is used to acquire material inventory data, GPS tracks of transport vehicles, and supplier credit rating data. The geographic meteorological data acquisition submodule is used to acquire radar detection results data, wind speed probability distribution for the next 72 hours, and tide tables.

3. The infrastructure project management system based on big data as described in claim 1, characterized in that, The data preprocessing module includes the following sub-modules: The spatiotemporal alignment submodule is used to synchronize the device clocks in the multimodal engineering project data using the DTW dynamic time warping algorithm, convert the BIM model coordinate system into the WGS84 geographic coordinate system, and match it with the sensor data space to obtain aligned engineering project data. The outlier submodule is used to detect abnormal sensor data in the aligned project data using the 3σ principle, and to use an LSTM neural network to identify faults in the mechanical record text data in the data, thereby deleting outliers and fault data to obtain complete project data. The data fusion submodule is used to establish a unified spatiotemporal index by associating BIM components, sensor readings and construction logs with timestamps and projects as keys. It uses the GraphSAGE algorithm to embed unstructured text into a 128-dimensional vector and inputs it together with the sensor numerical data into a graph neural network to obtain fused engineering project data.

4. The infrastructure project management system based on big data as described in claim 1, characterized in that, The data feature fusion module includes the following sub-modules: A submodule is constructed to build a feature library of the fused engineering project data and extract physical features from the fused engineering project data using an improved Arrhenius equation. The indicator submodule is used to obtain dynamic indicators reflecting the interaction behavior of people, machines and materials during the construction process from the integrated engineering project data based on the process network topology and real-time progress calculation, and to obtain behavioral characteristics. The acquisition submodule is used to calculate the integrated engineering project data to obtain supply chain resilience indicators and spatiotemporal correlation characteristics, and to obtain derived features. The calculation submodule is used to perform streaming calculations on the physical features, behavioral features, and derived features using the sliding window method to obtain feature engineering project data.

5. The infrastructure project management system based on big data as described in claim 1, characterized in that, The risk model building module includes the following sub-modules: A submodule is established to build a ResNet-Transformer risk prediction model, obtain historical engineering project data from the feature library, and use the input layer of the model to map the physical features, behavioral features and derived features in the historical engineering project data into high-dimensional vectors; The residual submodule is used to extract local spatiotemporal patterns using ResNet residual blocks in the ResNet-Transformer risk prediction model; The capture submodule is used to capture long-range dependencies between features by leveraging the Multi-HeadAttention in the Transformer encoding layer of the ResNet-Transformer risk prediction model. The enhancement submodule is used to enhance the nonlinearity of the data training process based on the FFN feedforward network in the model.

6. The infrastructure project management system based on big data as described in claim 1, characterized in that, The risk model building module includes the following units: The population generation unit is used to optimize the hyperparameters of the model using the improved whale optimization algorithm of IWOA. It randomly generates a whale population, where each individual in the whale population represents a set of model hyperparameter combinations and a pre-trained ResNet-Transformer model skeleton. The loss function calculation unit is used to calculate the loss function of each hyperparameter combination on the validation set. The model fine-tuning unit is used to update the whale's position according to the IWOA rules, retain the historical best solution, and trigger Cauchy mutation if the best fitness does not improve for 10 consecutive generations. The model is fine-tuned by selecting the hyperparameter combination with the highest fitness to obtain the target IWOA-ResNet-Transformer risk prediction model.

7. The infrastructure project management system based on big data as described in claim 1, characterized in that, The project construction planning module includes the following units: The indicator prediction unit is used to input the characteristic engineering project data into the target IWOA-ResNet-Transformer risk prediction model for prediction, and obtain multi-dimensional status indicators of the construction project, including at least risk level, status parameters and trend prediction. The status judgment unit is used to integrate the risk level, status parameters and trend predictions output by the model to obtain a project status report. According to the engineering management standards, the project status is divided into three levels: normal status, early warning status and emergency status. The construction planning unit is used to plan construction based on the project status. If it is in an early warning state, it assesses available resources and adjusts the priority of local processes. If it is in an emergency state, it suspends regional construction and recalculates the BIM model.

8. The method for implementing the big data-based infrastructure project management system as described in claim 1, characterized in that, The method includes the following steps: Acquire multimodal engineering project data from the system, including at least BIM model data, IoT sensor data, supply chain data, geographic and meteorological data, and engineering log text; The multimodal engineering project data is spatiotemporally aligned, outliers in the aligned data are processed to obtain complete engineering project data, and multimodal fusion is performed on the complete engineering project data to obtain fused engineering project data. Construct a feature library for the fused engineering project data, including at least physical features, behavioral features, and derived features, to obtain feature engineering project data; A ResNet-Transformer risk prediction model is established, and the hyperparameters of the model are optimized using the improved whale optimization algorithm of IWOA to obtain the target IWOA-ResNet-Transformer risk prediction model. The characteristic engineering project data is input into the target IWOA-ResNet-Transformer risk prediction model for prediction to obtain the construction project status, and construction planning is carried out based on the construction project status.

9. A method for implementing the big data-based infrastructure project management system as described in claim 1, characterized in that, The method includes the following steps: Extract geometric attribute data, material parameters, and logical relationship data of construction procedures from building components; Collect environmental data, structural data, and equipment data. The environmental data includes wind speed, temperature, humidity, and rainfall. The structural data includes concrete strain, tower crane tilt angle, and foundation settlement. The equipment data includes excavator hydraulic pressure and pump truck flow rate. Obtain material inventory data, GPS tracks of transport vehicles, and supplier credit rating data; Acquire radar detection data, wind speed probability distribution for the next 72 hours, and tide tables.

10. A method for implementing the big data-based infrastructure project management system as described in claim 1, characterized in that, The method includes the following steps: The device clocks in the multimodal engineering project data are synchronized by using the DTW dynamic time warping algorithm, the BIM model coordinate system is converted into the WGS84 geographic coordinate system, and the sensor data space is matched to obtain aligned engineering project data. The sensor abnormal data of the aligned project data is detected using the 3σ principle, and the mechanical record text data in the data is identified by using an LSTM neural network. The abnormal values ​​and fault data are deleted to obtain the complete project data. A unified spatiotemporal index is established by using timestamps and projects to associate key-value BIM components, sensor readings, and construction logs. The GraphSAGE algorithm is used to embed unstructured text into a 128-dimensional vector, which is then input into a graph neural network along with the sensor numerical data to obtain fused engineering project data.

Images