Intelligent building safety management system based on big data

Abstract

The invention discloses an intelligent building safety management system based on big data, and belongs to the technical field of intelligent buildings. Comprising the following modules: a multi-source data fusion and preprocessing module, which adopts intelligent data standardization and feature extraction technologies to realize efficient integration and semantic conversion of cross-source heterogeneous data and ensure data quality and consistency; the security risk dynamic knowledge graph construction module is used for constructing a semantic association network of building security elements through a graph neural network so as to continuously self-learn and dynamically update a risk feature relationship; the intelligent risk assessment and early warning module is used for carrying out comprehensive modeling, accurate positioning, layered assessment and intelligent early warning on building safety risks; the safety decision support module is used for providing an executable risk assessment report and an emergency decision suggestion; and the safety management visualization module adopts an interactive multi-dimensional visualization technology to intuitively present the dynamic evolution process of the building safety risk, so that the risk perception and management efficiency is remarkably improved.

Description

Technical Field

[0001] The present application relates to the field of smart building technology, and more specifically, to an intelligent building safety management system based on big data. Background Art

[0002] In today's rapidly developing construction industry, buildings are becoming larger, more complex, and more diverse in function. At the same time, building safety management faces unprecedented challenges.

[0003] Traditional building safety management models rely primarily on manual patrols, regular inspections, and experience-based risk assessments. These inspections not only consume significant manpower, material resources, and time, but also hinder real-time and comprehensive monitoring of the safety status of every building component and system. Due to the complexity of building structures and equipment, potential safety hazards may be overlooked, increasing the risk of safety incidents. For example, in large commercial buildings, numerous components and subsystems, such as electrical equipment, fire protection facilities, and ventilation systems, require continuous monitoring, making manual inspections difficult to guarantee.

[0004] With the rapid development of information technology, buildings have gradually introduced automated safety monitoring devices and systems, such as fire alarms and surveillance cameras. However, these devices often operate independently, with data stored in decentralized locations and lacking effective integration and in-depth analysis mechanisms. Data from different sources comes in varying formats and semantics, making cross-system data sharing and collaboration impossible. This makes it difficult to comprehensively understand building safety status and accurately predict risks. For example, smoke data detected by the fire protection system and overload data from the electrical system cannot be automatically correlated and analyzed, potentially delaying the identification and resolution of the root cause of fire risks.

[0005] Furthermore, existing building safety management methods lack efficient data processing capabilities and intelligent decision-making support when faced with massive amounts of monitoring data. They are unable to quickly and accurately identify potential safety risks based on multiple sources of information, including real-time operating data, environmental data, and historical data. They also struggle to provide personalized, forward-looking risk warnings and response strategies. In emergencies, safety managers often need to process large amounts of complex information and make decisions in a short period of time. Traditional management models struggle to meet this demand, easily leading to misjudgments and delays in rescue efforts.

[0006] To sum up, how to integrate multi-source heterogeneous building safety data, achieve efficient data processing and in-depth analysis, build an intelligent risk assessment and early warning mechanism, and provide accurate and real-time support for safety management decisions has become a technical problem that needs to be solved urgently. Summary of the Invention

[0007] In order to overcome the shortcomings of the existing technology, the purpose of this application is to provide an intelligent building safety management system based on big data, which includes the following modules:

[0008] The multi-source data fusion and pre-processing module uses intelligent data standardization and feature extraction technology to achieve efficient integration and semantic conversion of cross-source heterogeneous data, ensuring data quality and consistency;

[0009] A safety risk dynamic knowledge graph construction module uses a graph neural network to construct a semantic association network of building safety elements to continuously self-learn and dynamically update risk feature relationships;

[0010] Intelligent risk assessment and early warning module, which comprehensively models, accurately locates, hierarchically assesses and provides intelligent early warning for building safety risks;

[0011] Safety decision support module, providing executable risk assessment reports and emergency decision-making recommendations;

[0012] The safety management visualization module uses interactive multi-dimensional visualization technology to intuitively present the dynamic evolution of building safety risks, significantly improving risk perception and management efficiency.

[0013] Furthermore, the multi-source data fusion and preprocessing module includes the following components:

[0014] The data acquisition unit is responsible for collecting raw data from different data sources in real time and performing preliminary formatting processing;

[0015] The data standardization unit standardizes data from different data sources to ensure that they have unified units, scopes, and formats, eliminate differences caused by heterogeneous data, and ensure data consistency and comparability;

[0016] The data cleaning unit performs noise filtering, outlier detection, and missing data completion on the collected raw data to ensure data quality and accuracy and remove incomplete or unreliable data;

[0017] Multidimensional data integration unit, which integrates multidimensional data from different data sources to form a comprehensive building safety data set;

[0018] Feature selection and extraction unit, based on the characteristics of the building safety field, extracts key features related to risks from the building safety dataset;

[0019] The semantic conversion unit is responsible for converting the fused data into semantic information that is easy to understand and analyze, and realizing automatic conversion of data semantics through rule reasoning.

[0020] Furthermore, based on the characteristics of the building safety field, key features related to risks are extracted from the building safety dataset, including the following steps:

[0021] S1, through expert knowledge, standardized processes and preliminary analysis, select and construct a preliminary feature set S that is highly relevant to building safety risks init , expressed as: S init ={x j |j=1,2,…,M}, where x j is the primary feature variable closely related to building safety risks; M is the number of features in the primary feature set;

[0022] S2, using multidimensional statistical methods to deeply analyze the correlation between each feature and the risk target variable y, and screen out the most informative features. For continuous data, the Pearson correlation coefficient is used to evaluate the linear correlation strength between the feature and the target variable. The formula is:

[0023]

[0024] Cov(x j ,y) is the feature x j The covariance between the target variable y; σ(y) is the standard deviation of the target variable y; σ(x j ) is the feature x j The standard deviation of I pearson (x j ,y) represents feature x j and the linear correlation between the target variable y;

[0025] For classification targets, the chi-square statistic is used to evaluate the significance of the association between features and targets. The formula is:

[0026]

[0027] Among them, O k is the observed value of a specific risk state k; E k is the expected value of a specific risk state k; Represents feature x j Independence between and the target variable y;

[0028] S3, use mutual information to perform feature redundancy analysis and remove features with too high correlation. The formula is:

[0029]

[0030] Among them, MI(x i ,x j ) is the feature x i and x jThe mutual information between them; p(x i ,x j ) is the feature x i and x j The joint probability distribution of p(x i ) is the feature x i The marginal probability distribution of p(x j ) is the feature x j The marginal probability distribution of dx i , dx j For x i and x j Small changes when performing integration;

[0031] S4 introduces an interpretable machine learning model. Based on the model’s built-in feature importance scoring mechanism, it systematically evaluates the contribution of each feature to risk prediction and iteratively optimizes the feature selection strategy. The formula is:

[0032]

[0033] Among them, Gini after (T i ,x j ) is the feature x j When used as a splitting condition, the Gini index after splitting; Gini before (T i ) is the characteristic T i Gini index before split; I model (x j ) is the model-driven importance score; N is the number of records used for calculation and analysis;

[0034] S5, conduct multi-dimensional interactive analysis on candidate features, deeply explore the synergistic impact of feature combinations on building safety risk prediction, and calculate the comprehensive importance score of each feature using the following formula:

[0035]

[0036] Among them, α1, α2, α3, and α4 are the weights of Pearson correlation coefficient, chi-square test, information gain, and model-driven evaluation respectively; W j is feature x j The comprehensive importance score of gain (x j ,y) is the information gain, which is used to measure the feature x j Information contribution to the target variable y;

[0037] S6, according to the set threshold γ, select the first K features to form the final feature subset, expressed as: S final ={x j ∣Wj ≥γ}, where S final Represents the final set of key features that are filtered out.

[0038] Furthermore, it is responsible for converting the fused data into semantic information that is easy to understand and analyze, and realizing automatic conversion of data semantics through rule reasoning, which includes the following steps:

[0039] Based on domain knowledge and expert experience, combined with data analysis, a semantic conversion rule set is constructed to define the conversion mapping from raw data to semantic labels, and to clarify the triggering conditions and their corresponding semantic results.

[0040] Through rule reasoning, the collected data is matched and inferred with the constructed rule set, automatically converting the raw data into structured and easy-to-understand semantic information;

[0041] Incremental learning and adaptive feedback mechanisms are introduced to automatically adjust the weights and trigger conditions in the rules, continuously optimizing the accuracy and real-time performance of the reasoning semantic transformation rule set.

[0042] The converted semantic information is presented through an intuitive visual interface, enabling safety managers to quickly perceive and assess building safety risks.

[0043] Furthermore, the security risk dynamic knowledge graph construction module includes the following components:

[0044] The safety factor data collection unit collects safety factor information from key features related to risks and provides raw data for building the knowledge graph;

[0045] The semantic mapping and data annotation unit uses natural language processing and ontology mapping technology to semantically annotate the collected building safety factor data to ensure the data's structure and semantic consistency;

[0046] The graph construction unit uses graph database technology and graph neural network algorithms to construct a building safety knowledge graph based on the relationship between safety elements, revealing the semantic associations between various safety elements;

[0047] The feature relationship mining unit conducts in-depth analysis of nodes and edges in the graph, explores potential association rules between security factors, and establishes a relationship network of risk characteristics;

[0048] Dynamic update and self-learning units dynamically update security elements and relationships in the knowledge graph through a continuous learning mechanism, and continuously adjust the relationship between risk features based on real-time data and event feedback;

[0049] The risk reasoning and prediction unit, based on the constructed dynamic knowledge graph, uses graph reasoning and inference technology to analyze building safety risks, predict and evaluate future risks, and provide intelligent decision-making support.

[0050] Furthermore, we conduct in-depth analysis of the nodes and edges in the graph, explore the potential association rules between security factors, and establish a relationship network of risk characteristics, including the following steps:

[0051] Calculate node v using similarity measurement method p and v q The strength of the relationship between them is used to measure the similarity or correlation between building safety elements;

[0052] The graph convolution operation is implemented through the graph neural network, and the features of each node are updated, which can be expressed as follows: in, is node v p In the feature representation of layer o+1, N(v) is the feature representation of node v. p The neighbor node set, w(e vu ) is the node v p The relationship weight with neighbor node u; Represents the feature representation of node u in the o-th layer of the graph convolutional network;

[0053] By further mining the edges in the graph, we can analyze the potential relationships between security factors and establish a correlation network of risk characteristics.

[0054] Through continuous learning and dynamic updating mechanisms, the security elements and relationships in the graph are adjusted, which can be expressed as: ΔG=G t+1 -G t , where ΔG represents the change of the knowledge graph; G t+1 Represents the knowledge graph structure at time t+1; G t Represents the knowledge graph structure at time t;

[0055] Use the constructed risk feature relationship network to infer and predict future security risks.

[0056] Furthermore, the intelligent risk assessment and early warning module includes the following components:

[0057] The risk profile modeling unit constructs a multi-dimensional profile of building safety risks based on the extracted key features, identifying potential risk factors and hazard sources;

[0058] Risk location and correlation analysis unit accurately locates the specific location of risks and analyzes the correlation between risks and building structures and environmental factors;

[0059] Intelligent grading and assessment unit classifies building safety risks according to different risk levels, provides risk assessment results at different levels, and helps formulate targeted management strategies;

[0060] The personalized early warning strategy unit generates personalized early warning strategies based on risk profiling and classification results, combined with environmental factors and historical records;

[0061] Real-time risk monitoring and dynamic adjustment unit monitors building safety risks in real time and dynamically adjusts risk assessment models and early warning strategies based on real-time data and system feedback to ensure the timeliness and accuracy of early warnings;

[0062] The early warning feedback and decision support unit converts risk assessment and early warning results into specific decision support information and provides it to safety managers to help them take timely and effective emergency response measures.

[0063] Furthermore, a multi-dimensional portrait of building safety risks is constructed based on the extracted key features to identify potential risk factors and hazard sources, including the following steps:

[0064] Identify the key risk factors that have the most significant impact on building safety and construct a risk factor impact network to reveal their complex interaction mechanisms;

[0065] Based on the correlation analysis results, the potential types of building hazards are systematically identified, and a detailed risk level assessment is conducted for each type of hazard, quantifying its probability of occurrence and potential loss extent;

[0066] Using data visualization technology, we construct a risk profile from the dimensions of structural safety, performance, and environmental adaptability, visually displaying the overall risk status of the building and the risk levels in each dimension;

[0067] Based on the newly added data, the risk profile is continuously adjusted and optimized to form a closed-loop management mechanism of "data collection-risk analysis-early warning adjustment" to ensure the scientific nature and foresight of building safety risk management.

[0068] Furthermore, the security decision support module includes the following components:

[0069] The context understanding unit, based on natural language processing technology, accurately captures and analyzes the contextual semantics of complex security scenarios, achieving a deep understanding of risk information;

[0070] The scenario simulation and prediction unit predicts the development trend and evolution path of potential risks through multi-dimensional data simulation and scenario reconstruction;

[0071] The decision generation unit uses generative AI technology to transform risk analysis results into structured, clear, and executable security decision recommendations and risk assessment reports;

[0072] The emergency plan unit intelligently matches and recommends the most suitable emergency response plan based on risk level and scenario characteristics, providing accurate risk response strategies;

[0073] The decision tracking and feedback unit records and tracks each decision-making process, builds a closed-loop learning mechanism, and continuously optimizes the accuracy and effectiveness of the decision support model.

[0074] Furthermore, the potential risk development trend and evolution path are predicted using the following formula: Among them, R pred (t+r) represents the building safety risk predicted at time point t+r; f(·) is the risk assessment function used to calculate the risk level of building safety at a certain moment; A r represents the state transition matrix from the current time t to the future time t+r; X(t) represents the multidimensional data vector at time t; r is the time step, which is used to represent the predicted future time point; A r-s It represents the state transition matrix from the current time t to the time t+s, which is used to describe how the state of the system changes in different time steps; B(t+s) represents the external influencing factors at the time t+s.

[0075] Compared with the prior art, this application has the following beneficial effects:

[0076] This application uses modules such as multi-source data fusion and preprocessing, dynamic knowledge graph construction, intelligent risk assessment and early warning, safety decision support and visual management to achieve efficient integration of cross-source heterogeneous data, dynamic update of safety risk relationship network, accurate assessment and early warning of building safety risks, and provide executable risk management reports and decision support, significantly improving the efficiency and accuracy of building safety management. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 This is a structural diagram of an intelligent building safety management system based on big data disclosed in an embodiment of the present application. DETAILED DESCRIPTION

[0078] To make the objectives, technical solutions, and advantages of the present invention more apparent, the technical solutions in the embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Throughout the drawings, identical or similar reference numerals represent identical or similar elements or elements having identical or similar functions. The described embodiments are only some, not all, of the embodiments of the present invention.

[0079] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative work shall fall within the scope of protection of the present invention.

[0080] The embodiments and directional terms described below with reference to the accompanying drawings are exemplary and intended to be used to explain the present invention, but should not be construed as limiting the present invention.

[0081] like Figure 1 As shown in the figure, an intelligent building safety management system based on big data includes the following modules:

[0082] The multi-source data fusion and pre-processing module uses intelligent data standardization and feature extraction technology to achieve efficient integration and semantic conversion of cross-source heterogeneous data, ensuring data quality and consistency;

[0083] A safety risk dynamic knowledge graph construction module uses a graph neural network to construct a semantic association network of building safety elements to continuously self-learn and dynamically update risk feature relationships;

[0084] Intelligent risk assessment and early warning module, which comprehensively models, accurately locates, hierarchically assesses and provides intelligent early warning for building safety risks;

[0085] Safety decision support module, providing executable risk assessment reports and emergency decision-making recommendations;

[0086] The safety management visualization module uses interactive multi-dimensional visualization technology to intuitively present the dynamic evolution of building safety risks, significantly improving risk perception and management efficiency.

[0087] In this embodiment, the multi-source data fusion and preprocessing module uses intelligent data standardization and feature extraction technology to achieve efficient integration and semantic conversion of cross-source heterogeneous data. In modern construction projects, data sources are diverse, including sensors, monitoring equipment, construction records, geographic information systems (GIS), weather forecasts, and social security data. These data have significant differences in format, structure, and update frequency. This module unifies the format and structure of different data sources through intelligent data standardization, thereby achieving data consistency. In addition, feature extraction technology can automatically identify key features in the data and perform abstract processing, thereby eliminating redundant information and retaining the most valuable data. This module not only improves data processing efficiency, but also ensures data quality and reduces the impact of data errors on subsequent analysis. Through this module, data information from multiple heterogeneous data sources can be efficiently integrated and semantic conversion can be achieved, so that data can be accurately connected in different application scenarios, laying a solid foundation for subsequent building safety analysis and decision-making.

[0088] In this embodiment, the safety risk dynamic knowledge graph construction module adopts graph neural network (GNN) technology to establish a semantic association network based on building safety elements, which can dynamically associate and model various elements in the field of building safety. In construction projects, safety risks are often complex multi-dimensional issues, involving factors such as construction sites, worker health, climatic conditions, building structures, historical safety data, etc. By constructing a dynamically updated knowledge graph, it is possible to capture and learn the relationships between various risk factors in real time, and continuously update risk characteristics. The advantage of graph neural networks is that they can handle complex non-Euclidean relationships between nodes, can automatically mine potential connections between building safety elements, and support online updates of graphs, and can quickly adapt to new safety risk information. This module effectively enhances the ability to understand safety risks and provides a global perspective for subsequent risk assessment and prediction.

[0089] In this embodiment, the intelligent risk assessment and early warning module is one of the core application modules of the building safety management system. It is responsible for comprehensive modeling, precise location, tiered assessment, and intelligent early warning of building safety risks. Based on machine learning and deep learning algorithms, this module comprehensively analyzes historical and real-time data to accurately assess safety risks. Specifically, by modeling construction site data, it can identify potential safety hazards and locate the source of risk, such as equipment failure during construction, worker errors, or changes in the external environment. This tiered assessment mechanism prioritizes risks at different levels based on severity, probability of occurrence, and scope of impact, helping managers make more targeted decisions during emergency response. Furthermore, the intelligent early warning function automatically issues warnings based on preset risk thresholds and model predictions, prompting relevant personnel to take preventive measures and avoid potential accidents. The key technology of this module lies in its adaptive capability, which continuously optimizes the accuracy of assessments and early warnings as more data is input, thereby enhancing the proactive and proactive nature of building safety management.

[0090] In this embodiment, the safety decision support module systematically analyzes building safety risks to provide actionable risk assessment reports and emergency decision-making recommendations. This module integrates multiple data analysis and decision-making algorithms to help decision-makers quickly identify potential safety hazards and generate the most appropriate emergency response plans based on specific scenarios. In traditional building safety management, decisions often rely on experience and historical cases. However, the intelligent safety decision support module uses precise data analysis to provide decision-making recommendations based on real-time information. Using optimized decision trees or expert system algorithms, combined with safety risk assessment results, this module automatically generates specific emergency measures, such as adjusting construction plans, adding monitoring equipment, and conducting personnel evacuations. Furthermore, this module features intelligent recommendation capabilities that automatically adjust decision-making strategies based on the characteristics of different construction projects, ensuring personalized and efficient risk response. This module enables building safety managers to respond quickly to sudden safety incidents and effectively coordinate resources to minimize the risk of accidents.

[0091] In this embodiment, the safety management visualization module uses interactive multi-dimensional visualization technology to intuitively present the dynamic evolution of building safety risks to managers. By integrating data visualization and 3D modeling technologies, this module displays the real-time safety status of the building site, historical risk data, monitoring information, and risk assessment results through a graphical interface, greatly improving the operability of building safety management. Through the visualization interface, managers can quickly view information such as the distribution of risk points, predict accident development trends, and understand the current safety management status, avoiding the information lag and high processing complexity encountered in traditional management. The interactivity of visualization technology allows users to flexibly adjust the view according to actual needs, focusing on specific risk areas or switching between different layers to analyze different types of safety data. This multi-dimensional display allows managers to grasp the dynamic changes of building safety risks from a global perspective, enhancing their perception of complex building safety issues, thereby improving management efficiency and response speed.

[0092] In summary, the big data-based intelligent building safety management system, through the collaborative work of multiple technical modules, has built an efficient, accurate, and dynamic safety management platform. The multi-source data fusion and preprocessing module ensures efficient data integration and consistency, providing a solid data foundation for the system. The dynamic safety risk knowledge graph construction module gives the system powerful self-learning and adaptability, making risk assessment and prediction more accurate. The intelligent risk assessment and early warning module enhances the system's proactiveness, enabling early identification and response to potential safety risks. The safety decision support module provides scientific and precise decision-making recommendations, helping managers make rapid and effective emergency responses. The safety management visualization module significantly improves risk perception and management efficiency, helping users quickly identify risks and make decisions through an intuitive interface. Overall, this system has greatly improved the accuracy, real-time nature, and flexibility of building safety management through intelligent means, providing strong support for ensuring construction site safety.

[0093] Furthermore, the multi-source data fusion and preprocessing module includes the following components:

[0094] The data acquisition unit is responsible for collecting raw data from different data sources in real time and performing preliminary formatting processing;

[0095] The data standardization unit standardizes data from different data sources to ensure that they have unified units, scopes, and formats, eliminate differences caused by heterogeneous data, and ensure data consistency and comparability;

[0096] The data cleaning unit performs noise filtering, outlier detection, and missing data completion on the collected raw data to ensure data quality and accuracy and remove incomplete or unreliable data;

[0097] Multidimensional data integration unit, which integrates multidimensional data from different data sources to form a comprehensive building safety data set;

[0098] Feature selection and extraction unit, based on the characteristics of the building safety field, extracts key features related to risks from the building safety dataset;

[0099] The semantic conversion unit is responsible for converting the fused data into semantic information that is easy to understand and analyze, and realizing automatic conversion of data semantics through rule reasoning.

[0100] In summary, the multi-source data fusion and preprocessing module, as the core component of an intelligent building safety management system, achieves efficient processing and accurate analysis of building safety data through a series of steps, including data collection, standardization, cleaning, integration, feature extraction, and semantic transformation. These technical approaches not only ensure data quality, reliability, and consistency, but also, through intelligent processing, transform complex and heterogeneous building safety data into clear and easy-to-use risk information, providing a scientific basis for subsequent risk assessment and decision-making. Data standardization and cleaning ensure data accuracy, multidimensional data integration provides a global perspective, feature selection enhances the model's predictive capabilities, and semantic transformation makes complex data more operational and interpretable. Through this series of technical approaches, the building safety management system can efficiently integrate multi-source information, achieve comprehensive control of building safety risks, and provide data support for timely warnings and decision-making.

[0101] Furthermore, based on the characteristics of the building safety field, key features related to risks are extracted from the building safety dataset, including the following steps:

[0102] S1, through expert knowledge, standardized processes and preliminary analysis, select and construct a preliminary feature set S that is highly relevant to building safety risks init , expressed as: S init ={x j |j=1,2,…,M}, where x j is the primary feature variable closely related to building safety risks; M is the number of features in the primary feature set;

[0103] S2, using multidimensional statistical methods to deeply analyze the correlation between each feature and the risk target variable y, and screen out the most informative features. For continuous data, the Pearson correlation coefficient is used to evaluate the linear correlation strength between the feature and the target variable. The formula is:

[0104]

[0105] Cov(x j ,y) is the feature x j The covariance between the target variable y; σ(y) is the standard deviation of the target variable y; σ(x j ) is the feature x j The standard deviation of I pearson (x j ,y) represents feature x j and the linear correlation between the target variable y;

[0106] For classification targets, the chi-square statistic is used to evaluate the significance of the association between features and targets. The formula is:

[0107]

[0108] Among them, O k is the observed value of a specific risk state k; E k is the expected value of a specific risk state k; Represents feature x j Independence between and the target variable y;

[0109] S3, use mutual information to perform feature redundancy analysis and remove features with too high correlation. The formula is:

[0110]

[0111] Among them, MI(x i ,x j ) is the feature x i and x j The mutual information between them; p(x i ,x j ) is the feature x i and x j The joint probability distribution of p(x i ) is the feature x i The marginal probability distribution of p(x j ) is the feature x j The marginal probability distribution of dx i , dx j For x i and x j Small changes when performing integration;

[0112] S4 introduces an interpretable machine learning model. Based on the model’s built-in feature importance scoring mechanism, it systematically evaluates the contribution of each feature to risk prediction and iteratively optimizes the feature selection strategy. The formula is:

[0113]

[0114] Among them, Gini after (T i ,x j ) is the feature x j When used as a splitting condition, the Gini index after splitting; Gini before (T i ) is the characteristic T i Gini index before split; I model (x j ) is the model-driven importance score; N is the number of records used for calculation and analysis;

[0115] S5, conduct multi-dimensional interactive analysis on candidate features, deeply explore the synergistic impact of feature combinations on building safety risk prediction, and calculate the comprehensive importance score of each feature using the following formula:

[0116]

[0117] Among them, α1, α2, α3, and α4 are the weights of Pearson correlation coefficient, chi-square test, information gain, and model-driven evaluation respectively; W j is feature x j The comprehensive importance score of gain (x j ,y) is the information gain, which is used to measure the feature x j Information contribution to the target variable y;

[0118] S6, according to the set threshold γ, select the first K features to form the final feature subset, expressed as: S final ={x j ∣W j ≥γ}, where S final Represents the final set of key features filtered out.

[0119] Furthermore, it is responsible for converting the fused data into semantic information that is easy to understand and analyze, and realizing automatic conversion of data semantics through rule reasoning, which includes the following steps:

[0120] Based on domain knowledge and expert experience, combined with data analysis, a semantic conversion rule set is constructed to define the conversion mapping from raw data to semantic labels, and to clarify the triggering conditions and their corresponding semantic results.

[0121] Through rule reasoning, the collected data is matched and inferred with the constructed rule set, automatically converting the raw data into structured and easy-to-understand semantic information;

[0122] Incremental learning and adaptive feedback mechanisms are introduced to automatically adjust the weights and trigger conditions in the rules, continuously optimizing the accuracy and real-time performance of the reasoning semantic transformation rule set.

[0123] The converted semantic information is presented through an intuitive visual interface, enabling safety managers to quickly perceive and assess building safety risks.

[0124] In summary, data mapping and rule-based reasoning transform raw data into easily understandable semantic information, enabling safety managers to quickly identify and assess building safety risks. By constructing a set of transformation rules tailored to the specific characteristics of the building safety domain, combined with incremental learning and adaptive feedback mechanisms, the semantic transformation rules can be continuously optimized and updated, improving their accuracy and real-time performance. Furthermore, with the help of an intuitive visual interface, the transformed semantic information can better support managers' decision-making, enhancing their risk perception and response capabilities. This systematic semantic transformation process not only improves data operability but also lays a solid foundation for automated, intelligent, and precise building safety management. Faced with complex building safety environments, the semantic transformation module enables the system to respond quickly and provide managers with real-time, clear safety risk assessments, ultimately achieving intelligent building safety management.

[0125] Furthermore, the security risk dynamic knowledge graph construction module includes the following components:

[0126] The safety factor data collection unit collects safety factor information from key features related to risks and provides raw data for building the knowledge graph;

[0127] The semantic mapping and data annotation unit uses natural language processing and ontology mapping technology to semantically annotate the collected building safety factor data to ensure the data's structure and semantic consistency;

[0128] The graph construction unit uses graph database technology and graph neural network algorithms to construct a building safety knowledge graph based on the relationship between safety elements, revealing the semantic associations between various safety elements;

[0129] The feature relationship mining unit conducts in-depth analysis of nodes and edges in the graph, explores potential association rules between security factors, and establishes a relationship network of risk characteristics;

[0130] Dynamic update and self-learning units dynamically update security elements and relationships in the knowledge graph through a continuous learning mechanism, and continuously adjust the relationship between risk features based on real-time data and event feedback;

[0131] The risk reasoning and prediction unit, based on the constructed dynamic knowledge graph, uses graph reasoning and inference technology to analyze building safety risks, predict and evaluate future risks, and provide intelligent decision-making support.

[0132] In summary, the dynamic knowledge graph construction module for safety risks achieves comprehensive identification, in-depth analysis, and future prediction of safety risks through multi-level data collection, semantic mapping, relationship mining, and reasoning analysis. The collaborative work of various sub-units ensures that the knowledge graph can not only accurately reveal the relationship between building safety elements, but also be continuously updated and optimized in a dynamic environment. Over time, it can continuously adapt to new building safety risk characteristics and enhance the decision-making support capabilities of managers. Through risk reasoning and prediction units, potential risks can be identified in advance, intelligent risk warnings can be provided, and construction project teams can take forward-looking measures. In general, the construction and application of the dynamic knowledge graph for safety risks provides an intelligent and systematic solution for building safety management, which helps to improve the monitoring efficiency, risk identification capabilities, and decision-making support effects of building safety.

[0133] Furthermore, we conduct in-depth analysis of the nodes and edges in the graph, explore the potential association rules between security factors, and establish a relationship network of risk characteristics, including the following steps:

[0134] Calculate node v using similarity measurement method p and v q The strength of the relationship between them is used to measure the similarity or correlation between building safety elements;

[0135] The graph convolution operation is implemented through the graph neural network, and the features of each node are updated, which can be expressed as follows: in, is node v p In the feature representation of layer o+1, N(v) is the feature representation of node v. p The neighbor node set, w(e vu ) is the node v p The relationship weight with neighbor node u; Represents the feature representation of node u in the o-th layer of the graph convolutional network;

[0136] By further mining the edges in the graph, we can analyze the potential relationships between security factors and establish a correlation network of risk characteristics.

[0137] Through continuous learning and dynamic updating mechanisms, the security elements and relationships in the graph are adjusted, which can be expressed as: ΔG=G t+1 -G t , where ΔG represents the change of the knowledge graph; G t+1 Represents the knowledge graph structure at time t+1; G t Represents the knowledge graph structure at time t;

[0138] Use the constructed risk feature relationship network to infer and predict future security risks.

[0139] Furthermore, the intelligent risk assessment and early warning module includes the following components:

[0140] The risk profile modeling unit constructs a multi-dimensional profile of building safety risks based on the extracted key features, identifying potential risk factors and hazard sources;

[0141] Risk location and correlation analysis unit accurately locates the specific location of risks and analyzes the correlation between risks and building structures and environmental factors;

[0142] Intelligent grading and assessment unit classifies building safety risks according to different risk levels, provides risk assessment results at different levels, and helps formulate targeted management strategies;

[0143] The personalized early warning strategy unit generates personalized early warning strategies based on risk profiling and classification results, combined with environmental factors and historical records;

[0144] Real-time risk monitoring and dynamic adjustment unit monitors building safety risks in real time and dynamically adjusts risk assessment models and early warning strategies based on real-time data and system feedback to ensure the timeliness and accuracy of early warnings;

[0145] The early warning feedback and decision support unit converts risk assessment and early warning results into specific decision support information and provides it to safety managers to help them take timely and effective emergency response measures.

[0146] In summary, the intelligent risk assessment and early warning module, through the close collaboration of its various subunits, achieves comprehensive monitoring, dynamic assessment, and timely early warning of building safety risks. The risk profiling and risk location and correlation analysis units enable in-depth identification and analysis of key building safety risk factors. The intelligent grading and assessment unit provides a systematic, hierarchical perspective for risk assessment, ensuring that managers can accurately locate and respond to risks at different levels. The personalized early warning strategy unit makes early warnings more targeted and adaptable, while the real-time risk monitoring and dynamic adjustment unit ensures that adjustments can be made at any time based on environmental changes and data feedback. Finally, the early warning feedback and decision support unit provides intelligent decision support, helping managers quickly implement emergency response measures, ensuring more efficient and scientific safety management for construction projects. Overall, the intelligent risk assessment and early warning module, through the application of multi-level and multi-dimensional intelligent technologies, has significantly enhanced the intelligent level of building safety management, helping to achieve comprehensive, real-time, and accurate safety monitoring and emergency response for construction projects.

[0147] Furthermore, a multi-dimensional portrait of building safety risks is constructed based on the extracted key features to identify potential risk factors and hazard sources, including the following steps:

[0148] Identify the key risk factors that have the most significant impact on building safety and construct a risk factor impact network to reveal their complex interaction mechanisms;

[0149] Based on the correlation analysis results, the potential types of building hazards are systematically identified, and a detailed risk level assessment is conducted for each type of hazard, quantifying its probability of occurrence and potential loss extent;

[0150] Using data visualization technology, we construct a risk profile from the dimensions of structural safety, performance, and environmental adaptability, visually displaying the overall risk status of the building and the risk levels in each dimension;

[0151] Based on the newly added data, the risk profile is continuously adjusted and optimized to form a closed-loop management mechanism of "data collection-risk analysis-early warning adjustment" to ensure the scientific nature and foresight of building safety risk management.

[0152] In summary, constructing a multidimensional profile of building safety risks based on extracted key features is one of the core technologies in modern building safety management systems. By identifying key risk factors, constructing a risk factor impact network, and conducting risk assessments on potential sources of harm, it is possible to comprehensively reveal the potential sources of building safety risks and quantify the probability of occurrence and potential losses of each risk. Data visualization technology allows building safety risks to be presented intuitively, helping managers quickly understand the overall risk profile of building safety. The mechanism of continuously adjusting and optimizing risk profiles based on newly added data ensures the dynamic adaptability of building safety risk management and forms an efficient closed-loop management process. The combined application of these technologies has greatly enhanced the intelligent level of building safety management, helping managers to promptly identify, warn, and respond to safety risks in buildings, thereby ensuring the safety and reliability of construction projects.

[0153] Furthermore, the security decision support module includes the following components:

[0154] The context understanding unit, based on natural language processing technology, accurately captures and analyzes the contextual semantics of complex security scenarios, achieving a deep understanding of risk information;

[0155] The scenario simulation and prediction unit predicts the development trend and evolution path of potential risks through multi-dimensional data simulation and scenario reconstruction;

[0156] The decision generation unit uses generative AI technology to transform risk analysis results into structured, clear, and executable security decision recommendations and risk assessment reports;

[0157] The emergency plan unit intelligently matches and recommends the most suitable emergency response plan based on risk level and scenario characteristics, providing accurate risk response strategies;

[0158] The decision tracking and feedback unit records and tracks each decision-making process, builds a closed-loop learning mechanism, and continuously optimizes the accuracy and effectiveness of the decision support model.

[0159] In summary, the role of the safety decision support module in the building safety management system cannot be underestimated. Through the deep semantic analysis of the situational understanding unit, the forward-looking risk trend prediction of the scenario simulation and prediction unit, the generative AI technical support of the decision generation unit, the intelligent emergency plan recommendation of the emergency plan unit, and the closed-loop learning mechanism of the decision tracking and feedback unit, comprehensive decision support can be provided for building safety management. Each component plays a key role in improving the efficiency and quality of building safety management, ensuring that in a complex and dynamically changing building safety environment, managers can make timely and accurate decisions, effectively reduce safety risks and respond to emergencies. Through continuous optimization and intelligence, the safety decision support module not only improves the safety of construction projects, but also provides a feasible technical solution for the intelligent management of the construction industry.

[0160] Furthermore, the potential risk development trend and evolution path are predicted using the following formula: Among them, R pred (t+r) represents the building safety risk predicted at time point t+r; f(·) is the risk assessment function used to calculate the risk level of building safety at a certain moment; A r represents the state transition matrix from the current time t to the future time t+r; X(t) represents the multidimensional data vector at time t; r is the time step, which is used to represent the predicted future time point; A r-s It represents the state transition matrix from the current time t to the time t+s, which is used to describe how the state of the system changes in different time steps; B(t+s) represents the external influencing factors at the time t+s.

[0161] Finally, it should be noted that the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art will appreciate that modifications may be made to the technical solutions described in the above embodiments, or that some of the technical features may be replaced with equivalents; such modifications or replacements do not deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An intelligent building safety management system based on big data, characterized in that: Includes the following modules: The multi-source data fusion and pre-processing module uses intelligent data standardization and feature extraction technology to achieve efficient integration and semantic conversion of cross-source heterogeneous data, ensuring data quality and consistency; A dynamic knowledge graph construction module for safety risks uses a graph neural network to construct a semantic association network of building safety elements, enabling continuous self-learning and dynamic updating of risk feature relationships. Intelligent risk assessment and early warning module, which comprehensively models, accurately locates, hierarchically assesses and provides intelligent early warning for building safety risks; Safety decision support module, providing executable risk assessment reports and emergency decision-making recommendations; The safety management visualization module uses interactive multi-dimensional visualization technology to intuitively present the dynamic evolution of building safety risks, significantly improving risk perception and management efficiency.

2. The intelligent building safety management system based on big data according to claim 1 is characterized in that: The multi-source data fusion and preprocessing module includes the following components: The data acquisition unit is responsible for collecting raw data from different data sources in real time and performing preliminary formatting processing; The data standardization unit standardizes data from different data sources to ensure that they have unified units, scopes, and formats, eliminate differences caused by heterogeneous data, and ensure data consistency and comparability; The data cleaning unit filters noise, detects outliers, and completes missing data on the collected raw data to ensure data quality and accuracy and remove incomplete or unreliable data; Multidimensional data integration unit, which integrates multidimensional data from different data sources to form a comprehensive building safety data set; Feature selection and extraction unit, based on the characteristics of the building safety field, extracts key features related to risks from the building safety dataset; The semantic conversion unit is responsible for converting the fused data into semantic information that is easy to understand and analyze, and realizing automatic conversion of data semantics through rule reasoning.

3. The intelligent building safety management system based on big data according to claim 2 is characterized in that: Based on the characteristics of the building safety field, key features related to risks are extracted from the building safety dataset, including the following steps: S1, through expert knowledge, standardized processes and preliminary analysis, select and construct a preliminary feature set S that is highly relevant to building safety risks init , expressed as: S init ={x j |j=1,2,…,M}, where x j is the primary feature variable closely related to building safety risks; M is the number of features in the primary feature set; S2, using multidimensional statistical methods to deeply analyze the correlation between each feature and the risk target variable y, and screen out the most informative features. For continuous data, the Pearson correlation coefficient is used to evaluate the linear correlation strength between the feature and the target variable. The formula is: Cov(x j ,y) is the feature x j The covariance between the target variable y; σ(y) is the standard deviation of the target variable y; σ(x j ) is the feature x j The standard deviation of I pearson (x j ,y) represents feature x j and the linear correlation between the target variable y; For classification targets, the chi-square statistic is used to evaluate the significance of the association between features and targets. The formula is: Among them, O k is the observed value of a specific risk state k; E k is the expected value of a specific risk state k; Represents feature x j Independence between and the target variable y; S3, use mutual information to perform feature redundancy analysis and remove features with too high correlation. The formula is: Among them, MI(x i ,x j ) is the feature x i and x j The mutual information between them; p(x i ,x j ) is the feature x i and x j The joint probability distribution of p(x i ) is the feature x i The marginal probability distribution of p(x j ) is the feature x j The marginal probability distribution of dx i , dx j For x i and x j Small changes when performing integration; S4 introduces an interpretable machine learning model. Based on the model’s built-in feature importance scoring mechanism, it systematically evaluates the contribution of each feature to risk prediction and iteratively optimizes the feature selection strategy. The formula is: Among them, Gini after (T i ,x j ) is the feature x j When used as a splitting condition, the Gini index after splitting; Gini before (T i ) is the characteristic T i Gini index before split; I model (x j ) is the model-driven importance score; N is the number of records used for calculation and analysis; S5, conduct multi-dimensional interactive analysis on candidate features, deeply explore the synergistic impact of feature combinations on building safety risk prediction, and calculate the comprehensive importance score of each feature using the following formula: Among them, α1, α2, α3, and α4 are the weights of Pearson correlation coefficient, chi-square test, information gain, and model-driven evaluation respectively; W j is feature x j The comprehensive importance score of gain (x j ,y) is the information gain, which is used to measure the feature x j Information contribution to the target variable y; S6, according to the set threshold γ, select the first K features to form the final feature subset, expressed as: S final ={x j ∣W j ≥γ}, where S final Represents the final set of key features filtered out.

4. The intelligent building safety management system based on big data according to claim 2 is characterized in that: Responsible for converting the fused data into semantic information that is easy to understand and analyze, and realizing automatic conversion of data semantics through rule reasoning, including the following steps: Based on domain knowledge and expert experience, combined with data analysis, a semantic conversion rule set is constructed to define the conversion mapping from raw data to semantic labels, and to clarify the triggering conditions and their corresponding semantic results. Through rule reasoning, the collected data is matched and inferred with the constructed rule set, automatically converting the raw data into structured and easy-to-understand semantic information; Incremental learning and adaptive feedback mechanisms are introduced to automatically adjust the weights and trigger conditions in the rules, continuously optimizing the accuracy and real-time performance of the reasoning semantic transformation rule set. The converted semantic information is presented through an intuitive visual interface, enabling safety managers to quickly perceive and assess building safety risks.

5. The intelligent building safety management system based on big data according to claim 1 is characterized in that: The security risk dynamic knowledge graph construction module includes the following components: The safety factor data collection unit collects safety factor information from key features related to risks and provides raw data for building the knowledge graph; The semantic mapping and data annotation unit uses natural language processing and ontology mapping technology to semantically annotate the collected building safety factor data to ensure the data's structure and semantic consistency; The graph construction unit uses graph database technology and graph neural network algorithms to construct a building safety knowledge graph based on the relationship between safety elements, revealing the semantic associations between various safety elements; The feature relationship mining unit conducts in-depth analysis of nodes and edges in the graph, explores potential association rules between security factors, and establishes a relationship network of risk characteristics; Dynamic update and self-learning units dynamically update security elements and relationships in the knowledge graph through a continuous learning mechanism, and continuously adjust the relationship between risk features based on real-time data and event feedback; The risk reasoning and prediction unit, based on the constructed dynamic knowledge graph, uses graph reasoning and inference technology to analyze building safety risks, predict and evaluate future risks, and provide intelligent decision-making support.

6. The intelligent building safety management system based on big data according to claim 5 is characterized in that: Perform in-depth analysis of the nodes and edges in the graph, explore potential association rules between security factors, and establish a relationship network of risk characteristics, including the following steps: Calculate node v using similarity measurement method p and v q The strength of the relationship between them is used to measure the similarity or correlation between building safety elements; The graph convolution operation is implemented through the graph neural network, and the features of each node are updated, which can be expressed as follows: in, is node v p In the feature representation of layer o+1, N(v) is the feature representation of node v. p The neighbor node set, w(e vu ) is the node v p The relationship weight with neighbor node u; Represents the feature representation of node u in the o-th layer of the graph convolutional network; By further mining the edges in the graph, we can analyze the potential relationships between security factors and establish a correlation network of risk characteristics. Through continuous learning and dynamic updating mechanisms, the security elements and relationships in the graph are adjusted, which can be expressed as: ΔG=G t+1 -G t , where ΔG represents the change of the knowledge graph; G t+1 Represents the knowledge graph structure at time t+1; G t Represents the knowledge graph structure at time t; Use the constructed risk feature relationship network to infer and predict future security risks.

7. The intelligent building safety management system based on big data according to claim 1 is characterized in that: The intelligent risk assessment and early warning module includes the following components: The risk profile modeling unit constructs a multi-dimensional profile of building safety risks based on the extracted key features, identifying potential risk factors and hazard sources; Risk location and correlation analysis unit accurately locates the specific location of risks and analyzes the correlation between risks and building structures and environmental factors; Intelligent grading and assessment unit classifies building safety risks according to different risk levels, provides risk assessment results at different levels, and helps formulate targeted management strategies; The personalized early warning strategy unit generates personalized early warning strategies based on risk profiling and classification results, combined with environmental factors and historical records; Real-time risk monitoring and dynamic adjustment unit monitors building safety risks in real time and dynamically adjusts risk assessment models and early warning strategies based on real-time data and system feedback to ensure the timeliness and accuracy of early warnings; The early warning feedback and decision support unit converts risk assessment and early warning results into specific decision support information and provides it to safety managers to help them take timely and effective emergency response measures.

8. The intelligent building safety management system based on big data according to claim 7 is characterized in that: Based on the extracted key features, a multi-dimensional portrait of building safety risks is constructed to identify potential risk factors and hazard sources, including the following steps: Identify the key risk factors that have the most significant impact on building safety and construct a risk factor impact network to reveal their complex interaction mechanisms; Based on the correlation analysis results, the potential types of building hazards are systematically identified, and a detailed risk level assessment is conducted for each type of hazard, quantifying its probability of occurrence and potential loss extent; Using data visualization technology, we construct a risk profile from the dimensions of structural safety, performance, and environmental adaptability, visually displaying the overall risk status of the building and the risk levels in each dimension; Based on the newly added data, the risk profile is continuously adjusted and optimized to form a closed-loop management mechanism of "data collection-risk analysis-early warning adjustment" to ensure the scientific nature and foresight of building safety risk management.

9. The intelligent building safety management system based on big data according to claim 1, characterized in that: The Security Decision Support Module includes the following components: The context understanding unit, based on natural language processing technology, accurately captures and analyzes the contextual semantics of complex security scenarios, achieving a deep understanding of risk information; The scenario simulation and prediction unit predicts the development trend and evolution path of potential risks through multi-dimensional data simulation and scenario reconstruction; The decision generation unit uses generative AI technology to transform risk analysis results into structured, clear, and executable security decision recommendations and risk assessment reports; The emergency plan unit intelligently matches and recommends the most suitable emergency response plan based on risk level and scenario characteristics, providing accurate risk response strategies; The decision tracking and feedback unit records and tracks each decision-making process, builds a closed-loop learning mechanism, and continuously optimizes the accuracy and effectiveness of the decision support model.

10. The intelligent building safety management system based on big data according to claim 9, characterized in that: The potential risk development trend and evolution path are predicted using the following formula: Among them, R pred (t+r) represents the building safety risk predicted at time point t+r; f(·) is the risk assessment function used to calculate the risk level of building safety at a certain moment; A r represents the state transition matrix from the current time t to the future time t+r; X(t) represents the multidimensional data vector at time t; r is the time step, which is used to represent the predicted future time point; A r-s It represents the state transition matrix from the current time t to the time t+s, which is used to describe how the state of the system changes in different time steps; B(t+s) represents the external influencing factors at the time t+s.

Images