Multi-source heterogeneous data fusion engineering project supervision real-time visual decision-making platform

Abstract

The invention discloses an engineering project supervision real-time visual decision-making platform based on multi-source heterogeneous data fusion, and relates to the technical field of engineering supervision. The visual decision platform is in communication connection with a data acquisition module, a project risk feature module, a risk identification model construction module, a real-time data stream processing module, a risk early warning module and a decision support module. According to the invention, the risk in the engineering project is accurately identified in combination with historical multi-source heterogeneous data and project risk characteristics, and meanwhile, the real-time data stream processing module can analyze data in real time, identify potential risks and problems and generate early warning information, so that the accuracy and timeliness of risk identification are greatly improved, and the risk identification efficiency is improved. According to the invention, powerful decision support is provided for supervisors, and through the preset early warning level and the early warning signal, managers can rapidly understand the risk condition and take corresponding measures, thereby effectively reducing the project risk.

Description

Technical Field

[0001] The present invention relates to the technical field of project supervision, and particularly to a real-time visualization decision-making platform for project supervision of multi-source heterogeneous data fusion. Background Art

[0002] With the expansion of the scale of engineering projects, the progress of technology, and the diversification of management methods, project management in engineering projects faces increasingly complex challenges. Since the project involves a wide range of fields, including construction, civil engineering, electrical, mechanical, quality, schedule, cost, etc., the data sources are scattered and heterogeneous, and traditional management methods are difficult to meet the needs of large-scale, multi-department, and cross-regional project management.

[0003] In the prior art, for the decision-making management of project supervision in engineering projects, it is necessary to process information from multiple data sources. However, the fusion of multi-source heterogeneous data will cause a large delay, which in turn affects the response processing efficiency of fault items and warning items in the project. Therefore, a real-time visualization decision-making platform for project supervision of multi-source heterogeneous data fusion is proposed to solve the above problems. Summary of the Invention

[0004] The purpose of the present invention is to provide a real-time visualization decision-making platform for project supervision of multi-source heterogeneous data fusion to solve the problems raised in the above background art.

[0005] To solve the above technical problems, the technical solution adopted by the present invention is:

[0006] A real-time visualization decision-making platform for project supervision of multi-source heterogeneous data fusion, including a visualization decision-making platform, which is communicatively connected to a data acquisition module, a project risk feature module, a risk identification model construction module, a real-time data stream processing module, a risk warning module, and a decision support module;

[0007] The visualization decision-making platform is used to intuitively present the analysis results of relevant data of the engineering project, and realize the end-to-end collaborative work between various ends, ensure the real-time transmission and processing of data, improve the response speed and stability of the entire system, and ensure that the fault items and warning items in the project can be responded to and processed in a timely and accurate manner;

[0008] The data acquisition module is used to collect multi-source heterogeneous data from the engineering project, ensure the comprehensiveness and real-time nature of the data, and preprocess and fuse the multi-source heterogeneous data to obtain a comprehensive data set;

[0009] The project risk feature module is used to analyze the historical multi-source heterogeneous data of the engineering project, and extract project risk features related to the engineering project, including quality risk features and schedule risk features;

[0010] The risk identification model construction module is used to construct a risk identification model by combining historical multi-source heterogeneous data and project risk characteristics, and utilize machine learning algorithms to identify risks in engineering projects;

[0011] The real-time data stream processing module is used to instantaneously process and analyze the multi-source heterogeneous data in the integrated dataset by using real-time stream processing technology, reduce the latency of data processing, improve the real-time response speed, and ensure timely response to the fault items and warning items in the project;

[0012] The risk warning module combines the processing results of the real-time data stream and relevant project risk characteristics, and based on the preset risk identification model, identifies potential risks and problems, and generates warning information to timely discover potential problems in engineering projects, so as to help managers quickly respond and take necessary countermeasures;

[0013] The decision support module is used to provide corresponding decision support suggestions for the abnormal project situations identified by the risk warning, assist the supervision personnel in making decisions, reduce the influence of human factors on the decisions, and improve the accuracy of the decisions.

[0014] A further improvement of the technical solution of the present invention is that: the data acquisition module includes a data collection access unit and a data fusion and integration unit;

[0015] The data collection access unit is used to collect multi-source heterogeneous data from various data sources of the engineering project, and perform preprocessing operations such as cleaning, denoising, and format conversion on the collected multi-source heterogeneous data;

[0016] The data fusion and integration unit is used to fuse the preprocessed multi-source heterogeneous data by using a data fusion algorithm, and integrate the fused multi-source heterogeneous data to obtain an integrated dataset.

[0017] A further improvement of the technical solution of the present invention is that: the data collection access unit specifically includes:

[0018] Identify the locations, types, and access methods of various data sources of the engineering project. The data sources cover sensor networks, construction equipment, design drawings, construction logs, video surveillance, and Internet of Things devices. After the identification is completed, establish connections with various data sources according to the types and access protocols of the data sources to collect construction progress records, quality inspection reports, material procurement and acceptance data, personnel attendance records, equipment operation logs, and various monitoring data;

[0019] After the data source is accessed, the data acquisition access unit obtains the required data from the data source according to the preset acquisition frequency and data type. Among them, for the real-time data streams including sensor data and video surveillance data, data is continuously received and cached to ensure the continuity and integrity of the data. For non-real-time data including construction logs and design drawings, data is collected regularly according to the data update frequency.

[0020] Preprocess the collected multi-source heterogeneous data, including data cleaning, denoising, and format conversion. By using data cleaning tools, detect and repair missing values, duplicate records, and error data in the multi-source heterogeneous data, and use image processing algorithms for denoising of video surveillance data. Then, perform format conversion and standardization on the multi-source heterogeneous data after data cleaning and denoising to ensure data consistency and compatibility.

[0021] A further improvement of the technical solution of the present invention is that: the data fusion and integration unit specifically includes:

[0022] Align and match the preprocessed multi-source heterogeneous data, including time alignment, space alignment, and attribute alignment, and associate and match the data from different data sources through the project number. Through data alignment, ensure that the data from different data sources can accurately correspond when fused.

[0023] On the basis of data alignment and matching, perform data conflict detection, and then apply data fusion algorithms to fuse the processed data. Among them, data fusion algorithms include weighted average method, Kalman filtering method, Bayesian network, etc., and are selected according to the characteristics of the data and the fusion target.

[0024] Integrate the fused multi-source heterogeneous data, form a comprehensive data set through operations such as deduplication, sorting, and induction, remove duplicate data records, sort the data in chronological order, induce and summarize similar data, and then store the comprehensive data set in the data warehouse.

[0025] A further improvement of the technical solution of the present invention is that: the project risk feature module specifically includes:

[0026] Based on the collected historical multi-source heterogeneous data including construction progress records, quality inspection reports, material procurement and acceptance data, personnel attendance records, equipment operation logs, and various monitoring data, classify and label them, and according to the nature and use of the data, divide them into quality-related data and progress-related data. Among them, for quality data, label its corresponding detection standards and results, for progress data, label its deviation from the planned progress, and at the same time, perform time series labeling on the data.

[0027] Based on the classified and labeled data, conduct mining of risk characteristics, identify the characteristics related to quality risks and schedule risks. Among them, for quality risks, mine the characteristics related to material quality, construction technology, and test results; for schedule risks, mine the characteristics related to construction progress, resource allocation, and weather impacts.

[0028] After mining the risk characteristics, construct specific project risk characteristics. According to the mined risk characteristics, define quality risk characteristics and schedule risk characteristics. Among them, quality risk characteristics include material qualification rate, construction defect rate, and quality inspection passing rate; schedule risk characteristics include schedule deviation rate, critical path delay rate, and resource utilization rate. And conduct quantitative processing on the extracted risk characteristics to objectively evaluate the risks of engineering projects.

[0029] Combined with the requirements of the engineering project, set the benchmark values of each risk characteristic, and then summarize the extracted quality risk characteristics and schedule risk characteristics to form a project risk characteristic system containing multiple risk characteristics.

[0030] A further improvement of the technical solution of the present invention lies in: the risk identification model construction module specifically includes:

[0031] Traverse the historical multi-source heterogeneous data and project risk characteristics, match the project risk characteristics in the multi-source heterogeneous data with the corresponding risk labels, integrate to obtain a feature data set, divide it into a training set and a test set, and select a neural network model as the basic architecture of the risk identification model. The training set is used for the training process of the model, and the test set is used to evaluate the performance of the model.

[0032] Input the training set into the neural network model to construct a risk identification model and output the corresponding project risk values. During the training process, by adjusting the model parameters and optimizing the model structure, enable the model to learn the relationship between the characteristics and risks in the data, and be able to better fit the risk patterns in the historical data. Use the test set to evaluate the performance of the model. The evaluation indicators include accuracy, recall rate, F1 score, etc. According to the evaluation results, adjust and optimize the model to obtain a risk identification model that performs well on the training data and has good generalization ability.

[0033] After the model training is completed, deploy the risk identification model to the real-time visualization decision-making platform for engineering project supervision, and receive new multi-source heterogeneous data inputs in real time to identify the risks in the engineering project.

[0034] A further improvement of the technical solution of the present invention lies in: the construction process of the risk identification model specifically includes:

[0035] In the initial stage of building a risk identification model, initialize the neural network model, configure the number of network layers, the number of neurons in each layer, and the activation function, and set hyperparameters including the learning rate, batch size, and number of training epochs during the training process;

[0036] Input the training set into the initialized neural network model for forward propagation calculation. During the forward propagation process, the input data passes through each layer of the neural network in sequence, undergoes matrix multiplication with the weight matrix and non-linear transformation by the activation function, and finally outputs the project risk value;

[0037] After the forward propagation is completed, calculate the loss between the risk value output by the model and the true label through the loss function. The loss function includes mean squared error and cross-entropy loss. The value of the loss function reflects the accuracy and error degree of the model prediction. Then, calculate the gradient of the loss function with respect to each model parameter through the backpropagation algorithm. The backpropagation starts from the output layer and calculates the gradient layer by layer backward until the input layer. The gradient information will be used to update the model parameters and optimize the weights and biases of the model;

[0038] According to the gradient information obtained from the backpropagation calculation, use an optimization algorithm to update the model parameters. The optimization algorithms include stochastic gradient descent, momentum optimization, and Adam optimizer. The optimization algorithm controls the step size and direction of parameter update by adjusting the learning rate and momentum parameters to ensure that the model can quickly converge to the optimal solution. In each iteration, update the weight and bias parameters of the model according to the calculated gradient and the rules of the optimization algorithm. Through multiple iterations of training, gradually learn the relationship between the project risk characteristics and risks in the data, and improve the risk identification ability;

[0039] After the model training is completed, use the test set to evaluate the performance of the model. Among them, the evaluation metrics include accuracy, recall rate, and F1 score. According to the evaluation results, analyze the performance of the model on the test set and identify the deficiencies of the model. If the model performance does not meet the expectations, adjust and optimize the model according to the feedback of the evaluation metrics, including adjusting the network structure, reselecting the optimization algorithm, and adding regularization terms, to improve the generalization ability and stability of the model. Then, through multiple iterations of evaluation and adjustment, finally obtain a risk identification model that performs well on the training data and has good generalization ability.

[0040] A further improvement of the technical solution of the present invention lies in that: the real-time data stream processing module specifically includes:

[0041] Configure real-time data stream access points, including real-time data interfaces of data sources such as sensor networks, construction equipment, and video monitoring, and establish a connection with the comprehensive data set to ensure the ability to receive multi-source heterogeneous data in real time;

[0042] Set up a data buffer and a processing queue to allocate real-time data streams. Among them, the data buffer is used to temporarily store the received data, and the processing queue is used to process the data streams in sequence to ensure the continuity and stability of data processing;

[0043] Utilize the stream processing technology based on Apache Flink to instantaneously process and analyze the allocated real-time data streams, and detect events related to project risk characteristics during the analysis process to identify existing abnormal situations.

[0044] A further improvement of the technical solution of the present invention lies in: The risk warning module specifically includes:

[0045] The risk warning module accesses the output result of the real-time data stream processing module, obtains the multi-source heterogeneous data after instant processing and analysis, extracts the analysis results related to project risk characteristics, and further extracts and matches the characteristics of the data;

[0046] Match the characteristics in the real-time data with the preset risk characteristics, utilize the preset risk identification model to evaluate and identify the matched risk characteristics, and identify events and problems related to risks, namely project risk values and abnormal project risk characteristics;

[0047] Based on the project risk values, different warning levels are divided, namely low warning level, medium warning level and high warning level. Corresponding warning thresholds are matched for each warning level, and warning signals of corresponding colors are matched for each warning level. Among them, the low warning level is a blue warning signal, the medium warning level is an orange warning signal, and the high warning level is a red warning signal to assist management personnel to respond in a timely manner;

[0048] According to the output result of the risk identification model, generate corresponding warning information. The warning information includes risk type, occurrence location and time, and analyze the warning information to mark abnormal project risk characteristics to ensure that management personnel can quickly understand the risk situation and take necessary countermeasures.

[0049] A further improvement of the technical solution of the present invention lies in: The decision support module specifically includes:

[0050] Receive and parse the warning information from the risk warning module, including the detailed situation of abnormal projects, risk types, risk levels, occurrence locations, times, and relevant project risk characteristics. After parsing the risk information, prioritize the relevant project risk characteristics in the abnormal project situation to determine the project risk characteristics that need to be processed first;

[0051] Based on the results of risk early warning and priority ranking, generate corresponding decision support solutions, including specific countermeasures, resource allocation suggestions, time arrangements, etc., and provide multiple feasible solutions in combination with the actual situation of the current project;

[0052] Output the decision support solution to the supervision personnel, presenting it in a clear and understandable manner, including text descriptions and chart displays. At the same time, provide a feedback mechanism that allows the supervision personnel to evaluate and give feedback on the suggestions to further optimize the decision support module.

[0053] Due to the adoption of the above technical solution, the technical progress achieved by the present invention compared with the prior art is:

[0054] 1. The present invention provides a real-time visualization decision-making platform for engineering project supervision with multi-source heterogeneous data fusion. By combining historical multi-source heterogeneous data with project risk characteristics, it accurately identifies risks in engineering projects. At the same time, the real-time data stream processing module can instantly analyze data, identify potential risks and problems, and generate early warning information, greatly improving the accuracy and timeliness of risk identification, providing strong decision support for supervision personnel. Through preset early warning levels and early warning signals, management personnel can quickly understand the risk situation and take corresponding countermeasures, effectively reducing project risks.

[0055] 2. The present invention provides a real-time visualization decision-making platform for engineering project supervision with multi-source heterogeneous data fusion. Through the data acquisition module, it realizes the comprehensive collection, preprocessing and fusion of multi-source heterogeneous data in engineering projects, and can process data from various data sources such as sensor networks, construction equipment, design drawings, construction logs, video monitoring and Internet of Things devices, solving the problems of data dispersion, heterogeneity and fusion delay in traditional management methods. This efficient data processing method not only improves the timeliness and accuracy of data, but also ensures timely and accurate response and processing of fault items and early warning items in the project, thus improving the management efficiency and quality of the entire project. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0057] Figure 1 It is a schematic diagram of the system function module of the present invention;

[0058] Figure 2 It is a schematic diagram of the working process of the risk early warning module of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0060] Embodiment 1, as Figure 1 shown, the present invention provides a real-time visualization decision-making platform for engineering project supervision with multi-source heterogeneous data fusion, including a visualization decision-making platform, which is communicatively connected to a data acquisition module, a project risk feature module, a risk identification model construction module, a real-time data stream processing module, a risk warning module, and a decision support module;

[0061] The visualization decision-making platform is used to intuitively present the analysis results of relevant data of the engineering project, and realize end-to-end collaborative work between various ends, ensure the real-time transmission and processing of data, improve the response speed and stability of the entire system, and ensure that the fault items and warning items in the project can be responded to and processed in a timely and accurate manner;

[0062] The data acquisition module is used to collect multi-source heterogeneous data from the engineering project, ensure the comprehensiveness and timeliness of the data, and preprocess and fuse the multi-source heterogeneous data to obtain a comprehensive data set. The data acquisition module includes a data acquisition access unit and a data fusion and integration unit;

[0063] The data acquisition and access unit is used to collect various data sources of engineering projects to acquire multi-source heterogeneous data, and perform preprocessing operations such as cleaning, denoising, and format conversion on the collected multi-source heterogeneous data. It identifies the locations, types, and access methods of various data sources of engineering projects. The data sources cover sensor networks, construction equipment, design drawings, construction logs, video surveillance, and Internet of Things devices. After identification, according to the type and access protocol of the data source, it establishes connections with various data sources to collect construction progress records, quality inspection reports, material procurement and acceptance data, personnel attendance records, equipment operation logs, and various monitoring data. Among them, the sensor network is distributed at various key positions on the construction site to monitor environmental parameters and the status of construction equipment. The construction logs and design drawings are stored in the database or file server of the project management system. The video surveillance data comes from cameras installed at the construction site. After identification, for data sources of the database type, database connection parameters are configured. For the file server, the file access path and permissions are set. For Internet of Things devices, connections are made through API interfaces. After the data sources are accessed, the data acquisition and access unit obtains the required data from the data sources according to the preset acquisition frequency and data type. Among them, for real-time data streams including sensor data and video surveillance data, data is continuously received and cached to ensure the continuity and integrity of the data. For non-real-time data including construction logs and design drawings, it is collected regularly according to the data update frequency. The collected multi-source heterogeneous data is preprocessed, including data cleaning, denoising, and format conversion. By using data cleaning tools, missing values, duplicate records, and error data in the multi-source heterogeneous data are detected and repaired, and image processing algorithms are used for denoising of video surveillance data. Furthermore, the format conversion and standardization of the multi-source heterogeneous data after data cleaning and denoising are performed to ensure the consistency and compatibility of the data;

[0064] The data fusion and integration unit is used to fuse the preprocessed multi-source heterogeneous data by using data fusion algorithms, and integrate the fused multi-source heterogeneous data to obtain a comprehensive data set. It aligns and matches the preprocessed multi-source heterogeneous data, including time alignment, spatial alignment, and attribute alignment, and associates and matches the data from different data sources through project numbers. Through data alignment, it ensures that the data from different data sources can accurately correspond during fusion. On the basis of data alignment and matching, data conflict detection is performed, and then data fusion algorithms are applied to fuse the processed data. Among them, the data fusion algorithms include weighted average method, Kalman filtering method, Bayesian network, etc., which are selected according to the characteristics of the data and the fusion goal. The fused multi-source heterogeneous data is integrated to form a comprehensive data set through operations such as deduplication, sorting, and induction, removing duplicate data records, sorting the data in chronological order, and generalizing and summarizing similar data. Furthermore, the comprehensive data set is stored in the data warehouse;

[0065] The project risk feature module is used to analyze the historical multi-source heterogeneous data of the engineering project, extract the project risk features related to the engineering project, including quality risk features and schedule risk features. Based on the collected historical multi-source heterogeneous data including construction progress records, quality inspection reports, material procurement and acceptance data, personnel attendance records, equipment operation logs, and various monitoring data, it is classified and labeled, and according to the nature and use of the data, it is divided into quality-related data and schedule-related data. Among them, for quality data, the corresponding inspection standards and results are labeled, and for schedule data, the deviation from the planned schedule is labeled. At the same time, the data is labeled in time series. Based on the classified and labeled data, risk features are mined to identify the features related to quality risk and schedule risk. Among them, for quality risk, the features related to material quality, construction technology, and inspection results are mined, and for schedule risk, the features related to construction progress, resource allocation, and weather impact are mined. After mining the risk features, specific project risk features are constructed. According to the mined risk features, quality risk features and schedule risk features are defined. Among them, quality risk features include material qualification rate, construction defect rate, and quality inspection passing rate, and schedule risk features include schedule deviation rate, critical path delay rate, and resource utilization rate. And the extracted risk features are quantified to objectively evaluate the risks of the engineering project. Combining the requirements of the engineering project, the benchmark values of each risk feature are set, and then the extracted quality risk features and schedule risk features are summarized to form a project risk feature system containing multiple risk features;

[0066] The risk identification model construction module is used to combine the historical multi-source heterogeneous data with the project risk features and use machine learning algorithms to construct a risk identification model to identify the risks in the engineering project;

[0067] The real-time data stream processing module is used to use real-time stream processing technology to immediately process and analyze the multi-source heterogeneous data in the comprehensive data set, reduce the delay of data processing, improve the real-time response speed, and ensure that the fault items and warning items in the project can be responded to in time;

[0068] The risk warning module combines the processing results of the real-time data stream and the relevant project risk features, and based on the preset risk identification model, identifies potential risks and problems and generates warning information to timely discover potential problems in the engineering project to help managers quickly respond and take necessary countermeasures;

[0069] The decision support module is used to provide corresponding decision support suggestions for the abnormal project situations identified by the risk warning, assist the supervision personnel in making decisions, reduce the influence of human factors on the decision, and improve the accuracy of the decision.

[0070] Example 2, as Figure 2 shown, on the basis of Example 1, the present invention provides a technical solution: Preferably, the risk identification model construction module specifically includes:

[0071] Traverse the historical multi-source heterogeneous data and project risk characteristics, match the project risk characteristics in the multi-source heterogeneous data with the corresponding risk labels, integrate to obtain a feature dataset, divide it into a training set and a test set, and select a neural network model as the basic architecture of the risk identification model. The training set is used for the training process of the model, and the test set is used to evaluate the performance of the model. Input the training set into the neural network model to construct a risk identification model and output the corresponding project risk value. During the training process, by adjusting the model parameters and optimizing the model structure, the model can learn the relationship between the features and risks in the data, and can better fit the risk patterns in the historical data. Use the test set to evaluate the performance of the model. The evaluation indicators include accuracy, recall rate, F1 score, etc. According to the evaluation results, adjust and optimize the model to obtain a risk identification model that performs well on the training data and has good generalization ability. After the model training is completed, deploy the risk identification model to the real-time visualization decision-making platform for engineering project supervision, receive new multi-source heterogeneous data input in real time, and identify the risks in the engineering project;

[0072] In addition, the construction process of the risk identification model specifically includes:

[0073] In the initial stage of building a risk identification model, the neural network model is initialized, the number of network layers, the number of neurons in each layer, and the activation function are configured, and hyperparameters including the learning rate, batch size, and number of training epochs are set during training. The training set is input into the initialized neural network model for forward propagation calculation. During the forward propagation process, the input data passes through each layer of the neural network in sequence, undergoes matrix multiplication with the weight matrix and the non-linear transformation of the activation function, and finally outputs the project risk value. By simulating the mapping relationship from input to output of the data, the model learns the relationship between the features in the data and the risks, and attempts to fit the risk patterns in the historical data. The result of the forward propagation is used to calculate the loss function to evaluate the difference between the current prediction of the model and the true label. After the forward propagation is completed, the loss between the risk value output by the model and the true label is calculated through the loss function. The loss function includes mean squared error and cross-entropy loss. The value of the loss function reflects the accuracy and error degree of the model prediction. Furthermore, the gradient of the loss function with respect to each model parameter is calculated through the backpropagation algorithm. The backpropagation starts from the output layer and calculates the gradient layer by layer backward until the input layer. The gradient information will be used to update the model parameters and optimize the weights and biases of the model. According to the gradient information calculated by the backpropagation, an optimization algorithm is used to update the model parameters. The optimization algorithms include stochastic gradient descent, momentum optimization, and Adam optimizer. The optimization algorithm controls the step size and direction of parameter update by adjusting the learning rate and momentum parameters to ensure that the model can quickly converge to the optimal solution. In each iteration, according to the calculated gradient and the rules of the optimization algorithm, the weight and bias parameters of the model are updated. Through multiple iterations of training, the model gradually learns the relationship between the project risk characteristics and risks in the data and improves the risk identification ability. After the model training is completed, the performance of the model is evaluated using the test set. Among them, the evaluation metrics include accuracy, recall, and F1 score, which reflect the risk identification ability and reliability of the model from different perspectives. Accuracy measures the proportion of correctly predicted risks by the model, recall measures the ability of the model to identify actual risks, and the F1 score comprehensively considers accuracy and recall and provides a balanced evaluation of the model performance. According to the evaluation results, the performance of the model on the test set is analyzed to identify the deficiencies of the model. If the model performance does not meet the expectations, the model is adjusted and optimized according to the feedback of the evaluation metrics, including adjusting the network structure, reselecting the optimization algorithm, and adding regularization terms to improve the generalization ability and stability of the model. Furthermore, through multiple iterations of evaluation and adjustment, a risk identification model that performs well on the training data and has good generalization ability is finally obtained;

[0074] The calculation formula for the project risk value is:

[0075] ;

[0076] In the formula, is the project risk value, is the actual value of the th quality risk characteristic, is the actual value of the th schedule risk characteristic, and are the baseline values of the quality risk characteristic and the schedule risk characteristic respectively, which are used to measure the deviation between the actual value and the standard value. and are weight coefficients, which are used to adjust the relative importance of the quality risk and the schedule risk in the comprehensive risk value. is the number of quality risk characteristics, is the number of schedule risk characteristics. ranges from (0, 1). When is close to 0, it indicates that the project risk is relatively low. When is close to 1, it indicates that the project risk is relatively high. When the actual values and have a large deviation from the baseline values and , will increase;

[0077] The real-time data stream processing module specifically includes:

[0078] Configure real-time data stream access points, including real-time data interfaces of data sources such as sensor networks, construction equipment, and video monitoring, and establish a connection with the comprehensive data set to ensure that multi-source heterogeneous data can be received in real time. Set up a data buffer and a processing queue to allocate the real-time data stream. Among them, the data buffer is used to temporarily store the received data, and the processing queue is used to process the data stream in sequence to ensure the continuity and stability of data processing. Use the stream processing technology based on Apache Flink to immediately process and analyze the allocated real-time data stream, and detect events related to project risk characteristics during the analysis process to identify existing abnormal situations;

[0079] The risk warning module specifically includes:

[0080] The risk warning module accesses the output results of the real-time data stream processing module, obtains multi-source heterogeneous data after immediate processing and analysis, extracts the analysis results related to the project risk characteristics, further extracts and matches the features of the data, matches the features in the real-time data with the preset risk characteristics, and uses the preset risk identification model to evaluate and identify the matched risk characteristics, identifying events and problems related to risks, namely the project risk value and abnormal project risk characteristics. Different warning levels are divided based on the project risk value, namely the low warning level, medium warning level, and high warning level. Corresponding warning thresholds are matched for each warning level, and warning signals of corresponding colors are matched for each warning level. Among them, the low warning level is a blue warning signal, the medium warning level is an orange warning signal, and the high warning level is a red warning signal to assist managers in responding in a timely manner. According to the output results of the risk identification model, corresponding warning information is generated. The warning information includes the risk type, occurrence location, and time, and the warning information is analyzed to mark abnormal project risk characteristics to ensure that managers can quickly understand the risk situation to take necessary countermeasures;

[0081] Multiple warning levels correspond one-to-one with multiple warning thresholds, and the specific correspondence is as follows:

[0082] The warning threshold for the low warning level is: ;

[0083] The warning threshold for the medium warning level is: ;

[0084] The warning threshold for the high warning level is: ;

[0085] Among them, is the project risk value, is the upper threshold of the low warning level and the lower threshold of the medium warning level, is the upper threshold of the medium warning level and the lower threshold of the high warning level;

[0086] The decision support module specifically includes:

[0087] Receive and analyze the warning information from the risk warning module, including the details of abnormal items, risk types, risk levels, occurrence locations, times, and relevant project risk characteristics. After parsing the risk information, prioritize the relevant project risk characteristics in the abnormal item situation to determine the project risk characteristics that need to be processed first. Based on the results of risk warning and priority ranking, generate corresponding decision support plans, including specific countermeasures, resource allocation suggestions, time arrangements, etc., and provide multiple feasible solutions in combination with the actual situation of the current project. Output the decision support plan to the supervision personnel in a clear and understandable manner, including written descriptions and chart displays. At the same time, provide a feedback mechanism to allow the supervision personnel to evaluate and give feedback on the suggestions to further optimize the decision support module.

[0088] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A real-time visualization decision-making platform for engineering project supervision with multi-source heterogeneous data fusion, including a visualization decision-making platform, characterized in that: The visualized decision-making platform is communicatively connected to a data acquisition module, a project risk feature module, a risk identification model construction module, a real-time data stream processing module, a risk warning module, and a decision support module; The data acquisition module is used to collect multi-source heterogeneous data from an engineering project, preprocess and fuse the multi-source heterogeneous data to obtain a comprehensive data set; The project risk feature module is used to analyze the historical multi-source heterogeneous data of the engineering project, and extract project risk features related to the engineering project, including quality risk features and schedule risk features; The risk identification model construction module is used to combine the historical multi-source heterogeneous data with the project risk features, and use machine learning algorithms to construct a risk identification model to identify risks in the engineering project; The real-time data stream processing module is used to instantaneously process and analyze the multi-source heterogeneous data in the comprehensive data set by using real-time stream processing technology; The risk warning module combines the processing results of the real-time data stream and relevant project risk features, and based on a preset risk identification model, identifies potential risks and problems, and generates warning information; The decision support module is used to provide corresponding decision support suggestions for abnormal project situations identified by risk warning; 2. The real-time visualization decision-making platform for engineering project supervision with multi-source heterogeneous data fusion according to claim 1, characterized in that: The data acquisition module includes a data collection access unit and a data fusion and integration unit; The data collection access unit is used to collect multi-source heterogeneous data from various data sources of the engineering project, and perform preprocessing operations such as cleaning, denoising, and format conversion on the collected multi-source heterogeneous data; The data fusion and integration unit is used to fuse the preprocessed multi-source heterogeneous data by using a data fusion algorithm, and integrate the fused multi-source heterogeneous data to obtain a comprehensive data set.

3. The real-time visualization decision-making platform for engineering project supervision with multi-source heterogeneous data fusion according to claim 2, characterized in that: The data collection access unit specifically includes: Identify the locations, types, and access methods of various data sources of the engineering project. The data sources cover sensor networks, construction equipment, design drawings, construction logs, video monitoring, and Internet of Things devices. After identification, establish connections with various data sources according to the types and access protocols of the data sources to collect construction progress records, quality inspection reports, material procurement and acceptance data, personnel attendance records, equipment operation logs, and various monitoring data; After the data sources are accessed, the data collection access unit obtains the required data from the data sources according to the preset collection frequency and data type. Among them, for real-time data streams including sensor data and video monitoring data, continuously receive and cache the data, and for non-real-time data including construction logs and design drawings, collect it regularly according to the data update frequency; Perform preprocessing on the collected multi-source heterogeneous data, including data cleaning, denoising, and format conversion. By using a data cleaning tool, detect and repair missing values, duplicate records, and error data in the multi-source heterogeneous data, and perform denoising processing on the video monitoring data by using an image processing algorithm, and then perform format conversion and standardization on the multi-source heterogeneous data after data cleaning and denoising processing.

4. The real-time visualization decision-making platform for engineering project supervision with multi-source heterogeneous data fusion according to claim 3, characterized in that: The data fusion and integration unit specifically includes: Align and match the preprocessed multi-source heterogeneous data, including time alignment, space alignment, and attribute alignment, and associate and match the data from different data sources through project numbers; Based on the data alignment and matching, perform data conflict detection, and then apply a data fusion algorithm to fuse the processed data; Integrate the fused multi-source heterogeneous data, and form a comprehensive data set through operations such as deduplication, sorting, and induction, remove duplicate data records, sort the data in chronological order, summarize and generalize similar data, and then store the comprehensive data set in the data warehouse.

5. The real-time visualization decision-making platform for engineering project supervision with multi-source heterogeneous data fusion according to claim 4, characterized in that: The project risk feature module specifically includes: Classify and label the historical multi-source heterogeneous data collected, including construction progress records, quality inspection reports, material procurement and acceptance data, personnel attendance records, equipment operation logs, and various monitoring data, and divide them into quality-related data and progress-related data according to the nature and use of the data. Among them, for quality data, mark its corresponding inspection standards and results, for progress data, mark its deviation from the planned progress, and at the same time, perform time series labeling on the data; Based on the classified and labeled data, mine risk features, and identify features related to quality risks and progress risks. Among them, for quality risks, mine features related to material quality, construction technology, and inspection results, and for progress risks, mine features related to construction progress, resource allocation, and weather impacts; After mining the risk features, construct specific project risk features. According to the mined risk features, define quality risk features and progress risk features. Among them, quality risk features include material qualification rate, construction defect rate, and quality inspection passing rate, and progress risk features include progress deviation rate, critical path delay rate, and resource utilization rate, and perform quantitative processing on the extracted risk features; Combined with the requirements of the engineering project, set the benchmark values of each risk feature, and then summarize the extracted quality risk features and progress risk features to form a project risk feature system containing multiple risk features.

6. The real-time visual decision-making platform for engineering project supervision with multi-source heterogeneous data fusion according to claim 5, characterized in that: The risk identification model construction module specifically includes: Traverse the historical multi-source heterogeneous data and project risk features, match the project risk features in the multi-source heterogeneous data with the corresponding risk labels, integrate to obtain a feature data set, divide it into a training set and a test set, and select a neural network model as the basic architecture of the risk identification model; Input the training set into the neural network model to construct the risk identification model and output the corresponding project risk values. During the training process, adjust the model parameters and optimize the model structure to enable the model to learn the relationship between the features and risks in the data, use the test set to evaluate the performance of the model, and adjust and optimize the model according to the evaluation results to obtain the risk identification model; After the model training is completed, deploy the risk identification model to the real-time visualization decision-making platform for engineering project supervision, receive new multi-source heterogeneous data input in real time, and identify the risks in the engineering project.

7. The real-time visualization decision-making platform for engineering project supervision with multi-source heterogeneous data fusion according to claim 6, characterized in that: The construction process of the risk identification model specifically includes: In the initial stage of building a risk identification model, initialize the neural network model, configure the number of network layers, the number of neurons in each layer, and the activation function, and set hyperparameters including the learning rate, batch size, and number of training epochs during the training process; Input the training set into the initialized neural network model for forward propagation calculation. During the forward propagation process, the input data passes through each layer of the neural network in sequence, undergoes matrix multiplication with the weight matrix and non-linear transformation by the activation function, and finally outputs the project risk value; After the forward propagation is completed, calculate the loss between the risk value output by the model and the true label through the loss function, and then calculate the gradient of the loss function with respect to each model parameter through the backpropagation algorithm. The backpropagation starts from the output layer and calculates the gradient layer by layer backward until the input layer; According to the gradient information obtained from the backpropagation calculation, use an optimization algorithm to update the model parameters. The optimization algorithms include stochastic gradient descent, momentum optimization, and the Adam optimizer. In each iteration, update the weight and bias parameters of the model according to the calculated gradient and the rules of the optimization algorithm. Through multiple iterations of training, gradually learn the relationship between the project risk characteristics and risks in the data; After the model training is completed, use the test set to evaluate the performance of the model. The evaluation metrics include accuracy, recall, and F1 score. According to the evaluation results, analyze the performance of the model on the test set, identify the deficiencies of the model. If the model performance does not meet the expectations, adjust and optimize the model according to the feedback of the evaluation metrics, including adjusting the network structure, reselecting the optimization algorithm, and adding regularization terms. Then, through multiple iterations of evaluation and adjustment, finally obtain the risk identification model.

8. The real-time visualization decision-making platform for engineering project supervision with multi-source heterogeneous data fusion according to claim 7, characterized in that: The real-time data stream processing module specifically includes: Configure real-time data stream access points, including real-time data interfaces of data sources such as sensor networks, construction equipment, and video surveillance, and establish a connection with the comprehensive data set; Set up a data buffer and a processing queue to allocate the real-time data stream. Among them, the data buffer is used to temporarily store the received data, and the processing queue is used to process the data stream in sequence; Use the stream processing technology based on Apache Flink to immediately process and analyze the allocated real-time data stream, and detect events related to project risk characteristics during the analysis process to identify existing abnormal situations.

9. The real-time visual decision-making platform for engineering project supervision with multi-source heterogeneous data fusion according to claim 8, characterized in that: The risk warning module specifically includes: The risk warning module accesses the output result of the real-time data stream processing module, obtains the multi-source heterogeneous data after immediate processing and analysis, extracts the analysis results related to project risk characteristics, and further extracts and matches the features of the data; Match the features in the real-time data with the preset risk features, and use the preset risk identification model to evaluate and identify the matched risk features, and identify events and problems related to risks, that is, the project risk value and the abnormal project risk characteristics. Based on the project risk value, different warning levels are divided, namely low warning level, medium warning level and high warning level. Corresponding warning thresholds are matched for each warning level, and warning signals of corresponding colors are matched for each warning level. Among them, the low warning level is a blue warning signal, the medium warning level is an orange warning signal, and the high warning level is a red warning signal; According to the output result of the risk identification model, corresponding warning information is generated. The warning information includes risk type, occurrence location and time, and the warning information is analyzed to mark abnormal project risk characteristics.

10. The real-time visualization decision-making platform for engineering project supervision with multi-source heterogeneous data fusion according to claim 9, characterized in that: The decision support module specifically includes: Receive and parse the warning information from the risk warning module, including the detailed situation of abnormal projects, risk types, risk levels, occurrence locations, times and relevant project risk characteristics. After parsing the risk information, prioritize the relevant project risk characteristics in the abnormal project situation to determine the project risk characteristics that need to be processed first; Based on the results of risk warning and priority ranking, generate corresponding decision support plans, including specific countermeasures, resource allocation suggestions and time arrangements, etc., and provide multiple feasible solutions in combination with the actual situation of the current project; Output the decision support plan to the supervision personnel, including text description and chart display. At the same time, provide a feedback mechanism to allow the supervision personnel to evaluate and feedback on the suggestions to further optimize the decision support module.

Images