Foundation pit engineering dynamic risk index assessment method and system based on multi-source heterogeneous data fusion

Abstract

The invention discloses a foundation pit engineering dynamic risk index assessment method based on multi-source heterogeneous data fusion, and the method comprises the steps: obtaining multi-source heterogeneous data from on-site patrol, a monitoring system and video patrol, and the multi-source heterogeneous data comprises personnel safety, foundation pit operation, environment safety and hidden danger record data corresponding to mechanical equipment; mapping hidden danger events identified in the multi-source heterogeneous data to nodes capable of being scored in a preset four-level index system; based on the scoring results of all the fourth-level specific hidden danger indexes under each second-level index, calculating a comprehensive score of each second-level index; establishing a dynamic weight model; carrying out weighted fusion on the comprehensive score of each secondary index and the corresponding dynamic weight, and calculating to obtain a dynamic safety index of the foundation pit engineering; and determining a current risk early warning level according to the numerical range to which the dynamic safety index belongs, and outputting a corresponding risk management and control suggestion.

Description

Technical Field

[0001] This invention relates to the field of engineering construction safety assessment, and in particular to a method and system for assessing the dynamic risk index of foundation pit engineering based on multi-source heterogeneous data fusion. Background Technology

[0002] Currently, safety management and risk assessment in foundation pit engineering mainly rely on traditional monitoring and early warning systems and manual experience-based judgment. On the one hand, monitoring points for displacement, settlement, tilt angle, seepage pressure, water level, earth pressure, and support axial force are typically set up at the engineering site. Monitoring units regularly issue daily monitoring reports or periodic analysis reports, comparing the monitoring data with standard limits to determine whether the project is in a "safe" and "controllable" state. On the other hand, project managers and safety managers subjectively assess potential risks based on construction progress and on-site observations. This system, primarily based on "monitoring data exceeding limits and early warning + manual experience-based judgment," has played a certain role in long-term engineering practice, but it primarily assesses compliance with single-point indicators from a static perspective, lacking the ability to dynamically depict risks that change over time and with varying working conditions.

[0003] From the perspective of monitoring data processing, existing technologies mostly treat monitored values ​​as static samples, judging primarily whether the data at a certain moment exceeds a preset alarm threshold, while rarely considering the characteristics of changes over time, such as the rate of change, acceleration of change, and the degree of deviation between short-term and long-term trends. In many engineering accident cases, although the monitoring data had not exceeded the absolute limit for a period of time before the accident, there was already a significant and continuous increase or amplified fluctuation, indicating that the structure or soil condition was evolving in an unfavorable direction. However, traditional assessment methods still remain within the framework of "safe as long as the limit is not exceeded," failing to respond sensitively to this type of risk state of "not exceeding the limit but obviously deteriorating," leading to delayed early warnings. In addition, existing monitoring data analysis methods usually examine a single monitoring item in isolation, lacking systematic analysis of the coupling relationship between different types of monitoring quantities. For example, a continuous rise in seepage pressure is often accompanied by changes in displacement and settlement, but existing methods struggle to quantitatively assess potential surge or slip risks through multi-indicator linkage analysis.

[0004] Regarding the indicator system and weighting, existing methods for assessing foundation pit safety typically employ pre-defined indicator weights, fixing the importance of different risk factors once through expert scoring or experience-based judgment, and then maintaining this weighting throughout the construction cycle. This "static weighting" model fails to reflect the significant phased characteristics of deep foundation pit projects. In reality, the focus of risk control differs significantly across different construction phases. For example, during the support construction and early dewatering phases, greater attention needs to be paid to the quality of the support structure's formation and the effectiveness of dewatering; during foundation pit excavation, soil stability and bottom heave risks dominate; and during the installation and removal of internal supports, the deformation and axial force changes of the support system have a more pronounced impact on overall stability. If the weighting system is not adjusted according to the changes in construction phases, it will be difficult to accurately reflect the key risk factors at a particular time in the comprehensive assessment, easily leading to "averaging" and "blunting" of assessment results, and failing to provide guidance for differentiated allocation and focused attention of management resources.

[0005] Regarding data sources and integration, with the development of engineering informatization and smart construction site technologies, foundation pit engineering not only utilizes traditional monitoring data but also gradually introduces intelligent inspection methods such as on-site video surveillance, mobile terminal photography, and drone inspection images. This inspection data visually reflects changes in construction activities, the working environment, the form of support components, and on-site water accumulation, providing a wealth of supplementary information for risk assessment. However, existing management systems often treat this data only as post-event records and supplementary evidence, lacking a mechanism to incorporate it into a unified assessment system. On the one hand, monitoring data is a structured time series, while video and image data are unstructured, highly redundant, and exhibit discontinuous scene changes, making it difficult to align them on temporal and spatial scales. On the other hand, most projects currently lack a mature method for "quantifying" intelligent inspection data into indices that can be included in the index, resulting in a large amount of dynamic information that could have been utilized being left idle. Consequently, while multi-source data appears "abundant," at the index assessment level, "usable information is limited."

[0006] In terms of the way assessment results are expressed, existing safety assessments for foundation pit projects mostly present results in written conclusions, itemized comments, or simple ratings such as "qualified / average / requires attention / unqualified," lacking a unified, continuous, and quantifiable safety index expression mechanism. This method of expression is not conducive to managers intuitively grasping the overall changes in the safety status of the project, nor is it convenient for horizontal comparisons between different time periods and different projects. For example, when the safety status of a project gradually evolves from "basically stable" to "slightly unfavorable," if only discrete written comments and scattered alarms are relied upon, management will find it difficult to identify the trend changes in a timely manner, let alone quantify them as the magnitude and speed of changes in risk level. Without the support of dynamic indices, safety management often exhibits the characteristics of "more post-event summaries and weak process pre-control."

[0007] Furthermore, from the perspective of practical engineering management, monitoring data itself commonly suffers from problems such as missing data, noise, and instrument malfunctions. Due to complex on-site environments, untimely equipment maintenance, and communication interruptions, monitoring curves frequently exhibit short-term data loss, abnormal jumps, or unreasonable fluctuations. If these problems are not properly cleaned and supplemented, directly using them for risk assessment will significantly reduce the reliability of the assessment results; on the other hand, simply discarding abnormal data may result in the loss of crucial information that is indicative of risk evolution. In current engineering practice, data cleaning and supplementation largely rely on the experience of monitoring personnel and manual processing, lacking systematic and model-based methodological support, and rarely are these preprocessing steps incorporated into the formal assessment method for unified design and quantitative expression.

[0008] In summary, existing safety management technologies for foundation pit engineering are generally static, experience-based, and fragmented in terms of data processing, indicator construction, weighting, and evaluation expression. They struggle to reflect the continuous evolution of risks over time and the changes in dominant risks under different working conditions. The lack of an evaluation method based on multi-source monitoring data and intelligent inspection data, with dynamic weighting and scoring at its core, capable of outputting a single comprehensive index that automatically updates over time, has become a significant technical bottleneck restricting the improvement of safety management in foundation pit engineering. Summary of the Invention

[0009] The purpose of this invention is to provide a dynamic risk index assessment method for foundation pit engineering based on multi-source heterogeneous data fusion. By constructing a systematic dynamic index assessment method, it provides a real, continuous, and quantifiable tool for describing the safety status of foundation pit engineering, and solves the above-mentioned technical problems pointed out in the prior art.

[0010] This invention provides a method for assessing the dynamic risk index of foundation pit engineering based on multi-source heterogeneous data fusion, comprising the following steps: Acquire multi-source heterogeneous data from on-site inspections, monitoring systems, and video surveillance. The multi-source heterogeneous data includes records of potential hazards related to personnel safety, foundation pit operations, environmental safety, and mechanical equipment. Identify potential hazards in the current multi-source heterogeneous data; map the identified potential hazards in the multi-source heterogeneous data to scoreable nodes in a preset four-level indicator system, the four-level indicator system including first-level indicators, second-level indicators, third-level indicators and fourth-level specific hazard indicators (i.e., fourth-level indicators); score each mapped fourth-level specific hazard indicator according to dynamic scoring rules that match the indicator type; Based on the scoring results of all specific hazard indicators at the fourth level under each secondary indicator, calculate the comprehensive score of each secondary indicator; A dynamic weighting model is established, which dynamically calculates and adjusts the weighting coefficients of each secondary indicator based at least on the current construction stage and the changing trends of each secondary indicator in the recent evaluation period. The comprehensive scores of each secondary indicator are weighted and integrated with their corresponding dynamic weights to calculate the dynamic safety index of the foundation pit project; based on the numerical range of the dynamic safety index, the current risk warning level is determined and corresponding risk control recommendations are output.

[0011] Preferably, in the four-level indicator system, the primary indicators include personnel safety, work behavior, environmental safety, and machinery and equipment; the secondary indicators for personnel safety include personnel accidents, electric shock accidents, and object strike accidents; the secondary indicators for work behavior include foundation pit collapse or instability and foundation pit leakage and water inrush accidents; the secondary indicators for environmental safety include the impact on the surrounding environment and fire accidents; and the secondary indicators for machinery and equipment include machinery and equipment accidents.

[0012] Preferably, the method involves identifying potential incidents in the current multi-source heterogeneous data; mapping the identified potential incidents in the multi-source heterogeneous data to scoreable nodes in a preset four-level indicator system, including: Identify potential events in the current multi-source heterogeneous data and output result information; the result information includes text information. Then, semantic recognition is performed on the text information; By semantic analysis, the key feature words corresponding to the text information of the potential hazard event are matched with the definition description of the level 4 indicators, and then the event is classified into the corresponding scoreable nodes.

[0013] Preferably, each mapped Level 4 specific hazard indicator is scored according to a dynamic scoring rule that matches the indicator type, specifically including: The dynamic scoring rules are pre-set in a dynamic scoring table, and different deduction or score values ​​are set for different level four specific hidden danger indicators. Among them, the scores for the indicators of "personnel not wearing safety belts", "safety passages not being set up properly", "edge guardrails not being in place", "temporary power supply not being standardized", "excavation of foundation pits not being over-excavated or not being sloped according to design", "fire-fighting equipment not being configured or being ineffective", "excessive stacking of combustibles" and "non-standard operation of mechanical equipment" are 0 points. The score for the "lack of safety helmets for workers" indicator is 15 points; The score for the "cracks or damage to the enclosure wall" indicator is 6 points.

[0014] Preferably, based on the scoring results of all the specific hazard indicators of the fourth level under each secondary indicator, the comprehensive score of each secondary indicator is calculated, including: summing up the scores of all the specific hazard indicators of the fourth level corresponding to each type of accident, such as personnel accident, electric shock accident, object strike accident, pit collapse or instability, seepage and water inrush accident, impact on the surrounding environment, fire accident, and mechanical equipment accident, to obtain the comprehensive score of the corresponding current secondary indicator.

[0015] Preferably, the dynamic weighting model predefines a set of weight coefficients corresponding to the secondary indicators for each construction stage, and the sum of the weight coefficients is 1.

[0016] Preferably, the weighted fusion is performed by multiplying the comprehensive score of each secondary indicator by its corresponding dynamic weight to obtain the weighted score of each secondary indicator, and then summing the weighted scores of all secondary indicators to calculate the dynamic security index.

[0017] Preferably, the risk warning levels include four levels: safe, mild warning, moderate warning, and high warning; wherein, the mild warning status corresponds to the risk control recommendation of "there is a general risk and a general inspection mechanism needs to be activated".

[0018] Preferably, before acquiring the multi-source heterogeneous data from on-site inspections, monitoring systems, and video surveillance, the method further includes: Multi-source heterogeneous data is collected through the following detection systems or sensors, wherein the detection systems include data from the foundation pit engineering monitoring system, video inspection system, and manual inspection system; and the sensors include at least environmental monitoring sensors and video image monitoring sensors. Simultaneously, the collected multi-source heterogeneous data is processed through a unified data formatting, cleaning, completion, noise reduction, and spatiotemporal alignment process to obtain preprocessed multi-source heterogeneous data; Accordingly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in Embodiment 1.

[0019] Accordingly, the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in Embodiment 1.

[0020] Accordingly, the present invention provides a dynamic risk index assessment system for foundation pit engineering based on multi-source heterogeneous data fusion, comprising: The data acquisition module is used to acquire multi-source heterogeneous data from on-site inspections, monitoring systems, and video surveillance. The multi-source heterogeneous data includes records of potential hazards related to personnel safety, foundation pit operations, environmental safety, and mechanical equipment. The data mapping and scoring module is used to identify potential events in the current multi-source heterogeneous data; it maps the identified potential events in the multi-source heterogeneous data to scoreable nodes in a preset four-level indicator system, which includes first-level indicators, second-level indicators, third-level indicators, and fourth-level specific potential event indicators (i.e., fourth-level indicators); and it scores each mapped fourth-level specific potential event indicator according to dynamic scoring rules that match the indicator type. The secondary scoring module is used to calculate the comprehensive score of each secondary indicator based on the scoring results of all specific hazard indicators at the fourth level under each secondary indicator. The weight acquisition module is used to establish a dynamic weight model. The dynamic weight model dynamically calculates and adjusts the weight coefficients of each secondary indicator based at least on the current construction stage and the changing trends of each secondary indicator in the recent evaluation period. The index calculation and early warning output module is used to weight and integrate the comprehensive scores of each secondary indicator with their corresponding dynamic weights to calculate the dynamic safety index of the foundation pit project; based on the numerical range of the dynamic safety index, the current risk warning level is determined and corresponding risk control suggestions are output.

[0021] Compared with the prior art, the embodiments of the present invention have at least the following technical advantages: Analysis of the above-mentioned method for assessing the dynamic risk index of foundation pit engineering based on multi-source heterogeneous data fusion provided by the present invention shows that, in specific applications, multi-source heterogeneous data from on-site inspections, monitoring systems, and video inspections are obtained. The multi-source heterogeneous data includes personnel safety, foundation pit operations, environmental safety, and hazard record data corresponding to mechanical equipment. Identify potential hazards in the current multi-source heterogeneous data; map the identified hazards to scoreable nodes in a pre-defined four-level indicator system, which includes primary, secondary, tertiary, and specific hazard indicators (i.e., hazard indicators); score each mapped hazard indicator according to dynamic scoring rules matching the indicator type; calculate the comprehensive score of each secondary indicator based on the scoring results of all hazard indicators under each secondary indicator; establish a dynamic weighting model, which dynamically calculates and adjusts the weight coefficients of each secondary indicator based at least on the current construction stage and the changing trends of each secondary indicator within the recent assessment period; weight and fuse the comprehensive scores of each secondary indicator with their corresponding dynamic weights to calculate the dynamic safety index of the foundation pit project; determine the current risk warning level based on the numerical range of the dynamic safety index and output corresponding risk control recommendations.

[0022] Analysis of the above steps shows that the processing method of this invention successively goes through the processes of acquiring multi-source heterogeneous data, identifying potential incidents, mapping to a four-level indicator system, calculating dynamic safety index, and dynamic scoring.

[0023] When processing multi-source heterogeneous data, the collected information is more comprehensive and detailed, ensuring improved information integrity. Traditional methods rely on a single source (such as manual records), which is prone to omissions. This invention actively integrates multi-source data such as on-site inspections (personnel behavior), monitoring systems (structural deformation, water level, etc.), and video inspections (real-time footage), ensuring the breadth and depth of risk information collection and making the "portrait" of the risk status more complete. The original description of hidden dangers (such as "workers not wearing safety belts" or "cracks in support beams") is qualitative and unstructured, and cannot be directly used for calculation. By mapping to a preset four-level indicator system, all hidden dangers are standardized into specific indicator nodes within the system (such as "personnel accident - personnel not wearing safety belts" or "foundation pit collapse - support structure problems - retaining wall cracks"). This step is a key transformation, providing a unified "language" and "input unit" for subsequent quantitative calculations.

[0024] After mapping, scoring is performed according to "dynamic scoring rules matched to the indicator type." This means that hazards of different severity and nature (such as "not wearing a safety helmet" and "cracked retaining wall") will be assigned different deduction values ​​(such as 15 points and 6 points), rather than a simple "present / absent" judgment. This significantly improves the accuracy of the assessment, making the scoring results more reflective of the actual risk level of the hazard. These three steps constitute the transformation channel from raw data to quantitative initial values. Fusion ensures comprehensive input, mapping achieves data isomorphism, and dynamic scoring gives the data computable value. Their combined effect is to transform chaotic on-site information into standardized indicator scores with clear structure and comparable values, laying a solid foundation for subsequent comprehensive analysis.

[0025] Subsequently, the comprehensive score of the secondary indicators is calculated. This involves calculating the comprehensive score of each secondary indicator based on the scoring results of all specific hazard indicators at the fourth level under each secondary indicator. The weights mentioned above consider not only the stage but also the trend of indicator changes. For example, even if the absolute value of "foundation pit deformation" is not currently exceeding the limit, if its monitored value continues to increase rapidly within the recent assessment period (an unfavorable trend), the system can automatically increase the weight of this indicator, thereby reflecting this potential deterioration signal earlier and more significantly in the comprehensive index. This achieves a leap from "state-based assessment" to "state + trend-based assessment," greatly enhancing the effectiveness of risk management. Then, by weighted fusion of scores and weights, and calculation of the dynamic safety index, the warning level is determined: based on the index falling into a preset range (e.g., >90 safe, 70-90 mild warning, <70 severe warning), the system automatically determines the current warning level (e.g., "mild warning"). The above process realizes an automated and standardized closed loop from risk perception to response recommendations, significantly improving the timeliness and pertinence of management response.

[0026] In summary, the dynamic risk index assessment method for foundation pit engineering based on multi-source heterogeneous data fusion provided by this invention constructs a risk management and control system for foundation pit engineering that features real-time perception, dynamic adjustment, quantitative expression, and intelligent early warning. This system comprehensively improves the scientific nature and accuracy of risk management, providing a solid data intelligence guarantee for the smooth implementation of the project. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the main process of a dynamic risk index assessment method for foundation pit engineering based on multi-source heterogeneous data fusion; Figure 2 This is a schematic diagram illustrating a specific process for a dynamic risk index assessment method for foundation pit engineering based on multi-source heterogeneous data fusion. Figure 3 This is a schematic diagram of the early warning status levels in a dynamic risk index assessment method for foundation pit engineering based on multi-source heterogeneous data fusion. Figure 4 This is a schematic diagram of a case study in a dynamic risk index assessment method for foundation pit engineering based on multi-source heterogeneous data fusion. Figure 5 This is a schematic diagram of a dynamic risk index assessment system for foundation pit engineering based on multi-source heterogeneous data fusion.

[0028] Labels: Data acquisition module 10; Data mapping and scoring module 20; Secondary scoring module 30; Weight acquisition module 40; Index calculation and early warning output module 50. Detailed Implementation

[0029] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0031] Example 1 like Figure 1 As shown, Embodiment 1 of the present invention provides a method for assessing the dynamic risk index of foundation pit engineering based on multi-source heterogeneous data fusion, including the following steps: S10. Acquire multi-source heterogeneous data from on-site inspections, monitoring systems, and video inspections. The multi-source heterogeneous data includes hazard record data corresponding to personnel safety, foundation pit operations, environmental safety, and mechanical equipment. It should be noted that the above-mentioned hazard record data are all acquired through the following detection systems or sensor data, wherein the detection systems include data from foundation pit engineering monitoring systems, video inspection systems, and manual inspection systems, and the sensors include at least environmental monitoring sensors (such as high-temperature sensors), video image monitoring sensors, etc. S20. Identify potential incidents in the current multi-source heterogeneous data; map the identified potential incidents in the multi-source heterogeneous data to scoreable nodes in a preset four-level indicator system, the four-level indicator system including first-level indicators, second-level indicators, third-level indicators and fourth-level specific incident indicators (i.e., fourth-level indicators); score each mapped fourth-level specific incident indicator according to dynamic scoring rules that match the indicator type. It should be noted that the above steps are a mapping of potential hazards to the indicator system. Then, according to the four-level indicator system of the present invention, different potential hazards are mapped to corresponding scoreable nodes: prior to this, a four-level indicator system for foundation pit engineering safety risk assessment is constructed. This four-level indicator system includes multiple dimensions such as engineering entity risk, behavioral risk, and environmental risk, and has a hierarchical structure from first-level indicators to fourth-level indicators. In specific operation, each potential hazard is mapped to a corresponding scoreable indicator node (i.e., a scoreable node) in the four-level indicator system based on its risk attribute. S30. Based on the scoring results of all fourth-level specific hidden danger indicators under each secondary indicator, calculate the comprehensive score of each secondary indicator; S40. Establish a dynamic weighting model. The dynamic weighting model dynamically calculates and adjusts the weight coefficients of each secondary indicator based on the current construction stage and the changing trends of each secondary indicator within the recent evaluation period. It should be noted that the dynamic weighting model dynamically calculates and adjusts the weight coefficients of each secondary indicator based on the current construction stage and the changing trends of each secondary indicator within the recent evaluation period (the dynamic weights corresponding to each secondary indicator are obtained from the preset dynamic weighting model according to the current construction stage of the foundation pit project. For example, in a certain case, when the safety hazard occurred, the project was in the underground structure construction stage, so the weight selection was determined according to the predetermined weight values). S50. The comprehensive scores of each secondary indicator are weighted and fused with their corresponding dynamic weights to calculate the dynamic safety index of the foundation pit project; based on the numerical range of the dynamic safety index, the current risk warning level is determined, and corresponding risk control suggestions are output; (that is, it should be noted that the comprehensive scores of each secondary indicator are weighted and fused with their corresponding dynamic weights to calculate the dynamic safety index of the foundation pit project; the dynamic safety index is compared with the preset warning threshold range to output the current risk level and warning status of the foundation pit project, which will not be elaborated further). During the execution of S50 above, the main steps implemented are dynamic safety index fusion calculation and status output, which include: based on the weight coefficients of each secondary indicator determined by the dynamic weight model, the scores of each mapped quaternary indicator node are aggregated and calculated, firstly weighted and comprehensively obtained to get the scores of each secondary indicator, and then the final dynamic safety index is calculated by fusion of the weight coefficients; the dynamic safety index is compared with the preset warning threshold range, and the risk level and warning status of the current foundation pit project are output.

[0032] Analysis of the above steps shows that the processing method of this invention successively goes through the processes of acquiring multi-source heterogeneous data, identifying potential incidents, mapping to a four-level indicator system, calculating dynamic safety index, and dynamic scoring.

[0033] When processing multi-source heterogeneous data, the collected information is more comprehensive and detailed, ensuring improved information integrity. Traditional methods rely on a single source (such as manual records), which is prone to omissions. This invention actively integrates multi-source data such as on-site inspections (personnel behavior), monitoring systems (structural deformation, water level, etc.), and video inspections (real-time footage), ensuring the breadth and depth of risk information collection and making the "portrait" of the risk status more complete. The original description of hidden dangers (such as "workers not wearing safety belts" or "cracks in support beams") is qualitative and unstructured, and cannot be directly used for calculation. By mapping to a preset four-level indicator system, all hidden dangers are standardized into specific indicator nodes within the system (such as "personnel accident - personnel not wearing safety belts" or "foundation pit collapse - support structure problems - retaining wall cracks"). This step is a key transformation, providing a unified "language" and "input unit" for subsequent quantitative calculations.

[0034] After mapping, scoring is performed according to "dynamic scoring rules matched to the indicator type." This means that hazards of different severity and nature (such as "not wearing a safety helmet" and "cracked retaining wall") will be assigned different deduction values ​​(such as 15 points and 6 points), rather than a simple "present / absent" judgment. This significantly improves the accuracy of the assessment, making the scoring results more reflective of the actual risk level of the hazard. These three steps constitute the transformation channel from raw data to quantitative initial values. Fusion ensures comprehensive input, mapping achieves data isomorphism, and dynamic scoring gives the data computable value. Their combined effect is to transform chaotic on-site information into standardized indicator scores with clear structure and comparable values, laying a solid foundation for subsequent comprehensive analysis.

[0035] Subsequently, the comprehensive score of the secondary indicators is calculated. This involves calculating the comprehensive score of each secondary indicator based on the scoring results of all specific hazard indicators at the fourth level under each secondary indicator. The weights mentioned above consider not only the stage but also the trend of indicator changes. For example, even if the absolute value of "foundation pit deformation" is not currently exceeding the limit, if its monitored value continues to increase rapidly within the recent assessment period (an unfavorable trend), the system can automatically increase the weight of this indicator, thereby reflecting this potential deterioration signal earlier and more significantly in the comprehensive index. This achieves a leap from "state-based assessment" to "state + trend-based assessment," greatly enhancing the effectiveness of risk management. Then, by weighted fusion of scores and weights, and calculation of the dynamic safety index, the warning level is determined: based on the index falling into a preset range (e.g., >90 safe, 70-90 mild warning, <70 severe warning), the system automatically determines the current warning level (e.g., "mild warning"). The above process realizes an automated and standardized closed loop from risk perception to response recommendations, significantly improving the timeliness and pertinence of management response.

[0036] In summary, the dynamic risk index assessment method for foundation pit engineering based on multi-source heterogeneous data fusion provided by this invention constructs a risk management and control system for foundation pit engineering that features real-time perception, dynamic adjustment, quantitative expression, and intelligent early warning. This system comprehensively improves the scientific nature and accuracy of risk management, providing a solid data intelligence guarantee for the smooth implementation of the project.

[0037] In one specific implementation, the four-level indicator system includes primary indicators for personnel safety, work behavior, environmental safety, and machinery and equipment; secondary indicators for personnel safety include personnel accidents, electric shock accidents, and falling object accidents; secondary indicators for work behavior include foundation pit collapse or instability and foundation pit leakage / water inrush accidents; secondary indicators for environmental safety include impact on the surrounding environment and fire accidents; and secondary indicators for machinery and equipment include machinery and equipment accidents. It should be noted that in the above technical solution, the primary indicators are based on four main dimensions: personnel safety, work behavior, environmental safety, and machinery and equipment, which can be broadly summarized as engineering entity risk, behavioral risk, and environmental risk.

[0038] In one specific implementation, the step of identifying potential incidents in the current multi-source heterogeneous data and mapping the identified potential incidents in the multi-source heterogeneous data to scoreable nodes in a preset four-level indicator system includes: Identify potential events in the current multi-source heterogeneous data and output result information; the result information includes text information. Then, semantic recognition is performed on the text information; By semantic analysis, the key feature words corresponding to the text information of the potential hazard event are matched with the definition description of the level 4 indicators, and then the event is classified into the corresponding scoreable nodes.

[0039] It should be noted that during the above execution process, potential hazards in the current multi-source heterogeneous data are identified; these hazards are then mapped to scoreable nodes in a preset four-level indicator system. The mapping process for identifying potential hazards is as follows: for example, an image detection system can directly identify abnormal behavior and output the result information directly; the result information includes text information, such as workers not wearing safety helmets; then, semantic analysis is used to match the key feature words of the potential hazard with the definition description of the four-level indicators, and it is classified into the corresponding scoreable node (i.e., the four-level indicator node).

[0040] For example, the process of matching the key feature words corresponding to the text information of the potential hazard event through semantic analysis with the definition description of the fourth-level indicator, and classifying it to the corresponding scoreable node, that is, mapping the potential hazard event to the scoreable node, can take many forms. A brief example is given below: For example, "personnel not wearing safety belts", "safety passage not set up properly", and "edge guardrail not in place" are mapped to the fourth-level indicators under the second-level indicator "personnel accidents"; for example, "improper temporary power supply" is mapped to the fourth-level indicators under the second-level indicator "electric shock accident"; for example, "workers missing safety helmets" is mapped to the fourth-level indicators under the second-level indicator "object falling accident"; For example, “cracks or damage to the retaining wall” and “over-excavation or failure to slope the foundation pit according to the design” are mapped to the fourth-level indicators of the third-level indicator “support structure problems” under the second-level indicator “foundation pit collapse or instability”. For example, "fire-fighting equipment not configured or malfunctioning" and "excessive storage of combustibles" are mapped to the fourth-level indicators under the second-level indicator "fire accident"; For example, "non-standard operation of mechanical equipment" can be mapped to the fourth-level indicator under the second-level indicator "mechanical equipment accident".

[0041] In one specific implementation, each mapped level-four specific hazard indicator is scored according to a dynamic scoring rule that matches the indicator type, specifically including: The dynamic scoring rules are pre-set in a dynamic scoring table, and different deduction or score values ​​are set for different level four specific hidden danger indicators. Among them, the scores for the indicators of "personnel not wearing safety belts", "safety passages not being set up properly", "edge guardrails not being in place", "temporary power supply not being standardized", "excavation of foundation pits not being over-excavated or not being sloped according to design", "fire-fighting equipment not being configured or being ineffective", "excessive stacking of combustibles" and "non-standard operation of mechanical equipment" are 0 points. The score for the "lack of safety helmets for workers" indicator is 15 points; The score for the "cracks or damage to the enclosure wall" indicator is 6 points.

[0042] In one specific implementation, based on the scoring results of all fourth-level specific hazard indicators under each secondary indicator, the comprehensive score of each secondary indicator is calculated, including: summing up the scores of all fourth-level specific hazard indicators corresponding to each type of accident, such as personnel accident, electric shock accident, object strike accident, pit collapse or instability, seepage and water inrush accident, surrounding environmental impact, fire accident, and mechanical equipment accident, to obtain the comprehensive score of the corresponding current secondary indicator.

[0043] In one specific implementation, the dynamic weighting model predefines a set of weight coefficients corresponding to secondary indicators for each construction stage, and the sum of the weight coefficients is 1.

[0044] (That is, the dynamic weighting model mentioned above needs to load the dynamic weighting table when it is executed, i.e., Table 2 in the specific case is the dynamic weighting table); It should be noted that in the dynamic weighting table mentioned above, in the "underground structure construction stage", the weights of each secondary indicator are: personnel accident 0.15, electric shock accident 0.05, object strike accident 0.10, foundation pit collapse or instability 0.15, seepage and water inrush accident 0.15, surrounding environmental impact 0.15, fire accident 0.05, and mechanical equipment accident 0.20.

[0045] In one specific implementation, the weighted fusion is achieved by multiplying the comprehensive score of each secondary indicator by its corresponding dynamic weight to obtain the weighted score of each secondary indicator, and then summing the weighted scores of all secondary indicators to calculate the dynamic security index.

[0046] In one specific implementation, the risk warning level includes four levels: safe, mild warning, moderate warning, and high warning; wherein, the mild warning state corresponds to the risk control recommendation of "there is a general risk and a general inspection mechanism needs to be activated".

[0047] See Figure 2 In one specific embodiment, before acquiring the multi-source heterogeneous data from on-site inspections, monitoring systems, and video surveillance, the method further includes: S11. Collect multi-source heterogeneous data through the following detection systems or sensor data, wherein the detection systems include data from the foundation pit engineering monitoring system, video inspection system and manual inspection system; wherein the sensors include at least environmental monitoring sensors and video image monitoring sensors; S12. Simultaneously, the collected multi-source heterogeneous data is processed through a unified data formatting, cleaning, completion, noise reduction, and spatiotemporal alignment process to obtain preprocessed multi-source heterogeneous data. It should be noted that, in the above execution process, preprocessing includes steps such as data cleaning, missing data completion, noise removal, spatiotemporal alignment, and keyframe filtering. Specifically, this includes: multi-source data fusion and preprocessing: This invention uses monitoring data and intelligent patrol data as inputs. Monitoring data includes horizontal displacement, settlement, tilt angle, seepage pressure, water level, soil pressure, support axial force, etc.; patrol data includes images or videos taken by fixed cameras, mobile camera equipment, and drones.

[0048] To establish a unified data format, this invention employs the following preprocessing steps: ① Data cleaning • IQR-based anomaly detection and removal of jump values: or ;in, x i To monitor the i-th observation in the sequence;Q 1 and Q 3 represents the first and third quartiles of the sequence, used to characterize the upper and lower quartile features of the monitoring data; Q 3- Q 1 represents the interquartile range, used to measure the dispersion of monitoring data; a coefficient of 1.5 reflects the anomaly identification interval commonly used in statistics. When x i If the value exceeds this range, it indicates that it deviates from the normal fluctuation range and is considered an abnormal jump value, which should be removed or corrected before subsequent calculations.

[0049] • The monitoring sequence was smoothed using a moving average: ;in, For a moment t The smoothed monitoring value, x i The first in the original monitoring sequence i One observation value, t The current time number. k The moving average formula, defined as the sliding window size (window length), is used to smooth continuously monitored data, shifting the most recent... k The average value of the monitoring data from each sampling point is taken to reduce high-frequency noise and random fluctuations, making the sequence more reflective of actual engineering behavior characteristics. This method results in smoother and more reliable trend changes in the monitoring data, which is beneficial for improving the stability and accuracy of subsequent risk index assessments.

[0050] • Detect single points of failure through multi-point consistency checks: ;in, x i ( t ) is the first i The observed values ​​at each monitoring point at time t. x j ( t ) is the first j The observed values ​​at each monitoring point at time t. d i,j ( t The difference between two monitoring points at the same time is the absolute value. This formula is used to determine whether there are significant inconsistencies in the observed values ​​of different monitoring points at the same time. If the difference between a monitoring point and its adjacent or similar monitoring points exceeds the normal error range, it can be considered that the monitoring point may have instrument malfunction or noise interference, and should be identified and removed or corrected during the data cleaning stage.

[0051] ② Missing information completion • Short-term missing: linear interpolation • Continuous missing data: cubic spline interpolation ;in, S ( t () represents the estimated value after interpolation. t For the time that needs to be estimated, t i The start time of the segment. a i , b i , c i , d i These are the four coefficients of the spline function. This cubic spline interpolation formula is a continuous and smooth reconstruction method that constructs a third-order polynomial within a segment. By ensuring the continuity of the function values, first derivative, and second derivative at the segment boundaries, the interpolation results possess good smoothness and differentiability. This method can effectively restore the continuous trend of monitoring data within missing intervals, avoiding the abrupt changes in the broken line phenomenon caused by linear interpolation, thereby improving the stability and accuracy of subsequent dynamic scoring and index evaluation.

[0052] • Long-term missing data: ARIMA prediction ;in, x t For a moment t The monitored values ​​(original or differential sequence). B For the backshift operator, (1- B )^ d For difference operators, ϕ ( B ) is an autoregressive (AR) polynomial. θ ( B ) is a moving average (MA) polynomial. ϵ t The sequence is white noise (random disturbance term). This formula indicates that the ARIMA model uses an autoregressive (AR) term to describe the inertia of the monitoring sequence, an inverse (I) term to eliminate trends, and a moving average (MA) term to characterize random disturbances, thus enabling high-precision prediction of data in missing intervals. The model can reflect the trend, periodicity, and noise characteristics of the foundation pit monitoring sequence and is suitable for scenarios involving long-term continuous missing data.

[0053] ③ Noise Removal • The monitoring sequence uses Kalman filtering: ;in, For a moment t The optimal estimate, This is the predicted value (prior estimate). z t For a momentt The observed values, K t This refers to Kalman Gain. H This is the observation matrix. The Kalman filter update formula generates the optimal estimate by weightedly fusing the predicted values ​​and the actual observed values. Kalman gain K t This invention uses a method to dynamically adjust the relative weights of prediction and observation, ensuring that the estimation results approximate the actual engineering change trend to the greatest extent possible even in the presence of noise. This method effectively eliminates random noise and transient jumps in monitoring data, improves the stationarity of the monitoring sequence, and provides reliable input for subsequent dynamic weight and exponent calculations.

[0054] The video was deblurred using Gaussian filtering and optical flow vectors.

[0055] ④ Spatiotemporal alignment • Unified monitoring and video timestamps: ; • Achieving image-to-engineering space mapping using camera calibration: ;in, P world For engineering space coordinates (world coordinate system). P image Image coordinates (pixel coordinate system). K -1 It is the inverse of the camera intrinsic parameter matrix. R -1 The inverse of the extrinsic rotation matrix. t This is the translation vector of the extrinsic parameters. This formula, through the camera's intrinsic and extrinsic parameter matrices, realizes the process of converting image coordinates into engineering spatial coordinates. This process can accurately map potential hazards identified by video inspections (such as crack locations, load locations, and machinery operation distances) to the foundation pit engineering coordinate system, thereby achieving spatial correlation with monitoring points, BIM models, and indicator systems. This is a key step in realizing multi-source data fusion.

[0056] ⑤ Keyframe filtering Key frames reflecting changes in operating conditions are selected based on a comprehensive analysis of frame difference, optical flow intensity, and histogram differences.

[0057] To illustrate, the following is a specific example: After multi-source data fusion and preprocessing, an indicator system is constructed: The indicator system of this invention is based on the typical risk mechanisms of foundation pit engineering, and takes "personnel safety, work behavior, environmental safety, and mechanical equipment" as the main dimensions, systematically decomposing them in the form of "first-level indicators, second-level indicators, third-level indicators, and fourth-level indicators". The construction of the indicator system is based on foundation pit engineering accident statistics, major accident hazard judgment standards, hazard identification atlases, and enterprise inspection systems, and introduces a risk chain model to decompose each type of risk from cause, process to manifestation, so that each indicator can be quantitatively recorded through monitoring data, inspection images, or manual verification, ensuring that the index model has structured input.

[0058] Table 1 Indicator System Table

[0059] Then, dynamic weight modeling: The dynamic weight model of this invention takes the "change of dominant risk factors in the construction phase" of foundation pit engineering as its core logic, and constructs a multi-level weight system that can automatically adjust with time, working conditions, and risk trends. Its innovation lies in introducing a fusion modeling method of "literature statistical weight + improved analytic hierarchy process weight + construction phase dynamic factors + trend amplification factors", so that the index weights can reflect the differences in dominant risks in different construction phases of deep foundation pit engineering, and are updated in real time with changes in monitoring data and inspection results.

[0060] Table 2 Dynamic Weighting System

[0061] Then, a dynamic scoring mechanism is implemented: The dynamic scoring mechanism of this invention designs differentiated scoring methods for different types of indicators, so that monitoring indicators, inspection indicators, environmental indicators, and behavioral indicators can be scored independently and quantitatively according to their risk performance; and combined with a dynamic weight model, the scoring results can be updated over time and reflect the evolution trend of the engineering safety status in real time.

[0062] Table 3 Dynamic Scoring Table

[0063] (5) Index calculation and updating, i.e., the final dynamic index is: ; The index supports real-time or periodic updates, generating dynamic curves for risk trend analysis and early warning determination. The index model supports hierarchical management, classifying levels into green, yellow, orange, and red based on thresholds to meet engineering early warning and management needs. See also Figure 3 , Figure 3Index risk grading; through the above technical solutions, this invention forms a dynamic index evaluation method that can adapt to the entire process of foundation pit engineering, realizing the time-series, adaptive and quantitative expression of the index.

[0064] In a specific construction project, the first step is to obtain on-site inspection data. The data source is a foundation pit engineering management platform in Shenzhen. Example content is as follows: Figure 4 As shown (i.e., the semantic list after identifying various potential hazards at the construction site). Then, the mapping process from hazard events to the indicator system is performed. According to the four-level indicator system of this invention, different hazards are mapped to corresponding indicator nodes: (1) Personnel safety hazards: A worker on the third basement level was working at height on the scaffold without wearing a safety belt → Level 2 indicator "Personnel accident" Level 4 indicator "Personnel not wearing a safety belt"; The operator of tower crane #2 was not wearing a safety helmet when going up and down the tower crane → Level 2 indicator "Object falling accident" Level 4 indicator "Worker's safety helmet is missing"; The ladder steps in the mud pit on site were too far apart and no step boards were installed → Level 2 indicator "Personnel accident" Level 4 indicator "Safety passage not set up properly"; The edge protection fence on site was not stable in some places, the protection height was insufficient in some places, and the enclosure was not complete in some places → Level 2 indicator "Personnel accident" Level 4 indicator "Edge protection railing not in place"; A spot check of the power distribution of construction machinery in the switch box on the 21st floor showed that it did not comply with the "one machine, one switch" power supply regulations → Level 2 indicator "Electric shock accident" Level 4 indicator "Temporary power supply not up to standard" (2) Hidden dangers in foundation pit operations: The bottom of the second support beam in the northeast corner is cracked → Level II indicator "foundation pit collapse or instability", Level III indicator "support structure problem", Level IV indicator "cracks or damage to the retaining wall"; Spot check shows that the foundation excavation of the tower crane foundation in the foundation pit on the east side was not sloped → Level II indicator "foundation pit collapse or instability", Level III indicator "support structure problem", Level IV indicator "over-excavation of foundation pit or failure to slope according to design"; (3) Environmental safety hazards: Carbon dioxide cylinders, oxygen cylinders, and liquefied propane cylinders are stored in the on-site gas cylinder storage area on the first floor. The safety distance is less than 10 meters → Level II indicator "Fire accident (foundation pit construction site)" Level IV indicator "Fire-fighting equipment not configured or ineffective"; A spot check of the basement mezzanine of Building A revealed that a large amount of decoration materials had not been cleaned up in time, posing a fire hazard → Level II indicator "Fire accident (foundation pit construction site)" Level IV indicator "Excessive stacking of combustibles" (4) Hidden dangers related to mechanical equipment: There is no warning area for mechanical equipment operation or the operator enters the dangerous area → secondary indicator "mechanical equipment accident" and quaternary indicator "non-standard operation of mechanical equipment"; all hidden dangers are successfully mapped to the "scoring nodes" in the indicator system of this invention.

[0065] Then, the potential hazards are scored according to the dynamic scoring table in Table 3: (1) Personnel safety hazards; Level 2 indicator "Personnel accident" Level 4 indicator "Personnel not wearing safety belt" gets 0 points; Level 2 indicator "Object striking accident" Level 4 indicator "Workers missing safety helmet" gets 15 points; Level 2 indicator "Personnel accident" Level 4 indicator "Safety passage not set up properly" gets 0 points; Level 2 indicator "Personnel accident" Level 4 indicator "Edge protection railing not in place" gets 0 points; Level 2 indicator "Electric shock accident" Level 4 indicator "Temporary power use not up to standard" gets 0 points; (2) Hidden dangers in foundation pit operations; Level 2 indicator "foundation pit collapse or instability", Level 3 indicator "support structure problems", Level 4 indicator "cracks or damage to retaining wall" get 6 points; Level 2 indicator "foundation pit collapse or instability", Level 3 indicator "support structure problems", Level 4 indicator "over-excavation of foundation pit or failure to slope according to design" get 0 points; (3) Environmental safety hazards; Level 2 indicator "Fire accident (foundation pit construction site)" Level 4 indicator "Fire-fighting equipment not configured or malfunctioning" gets 0 points; Level 2 indicator "Fire accident (foundation pit construction site)" Level 4 indicator "Excessive stacking of combustibles" gets 0 points; (4) Mechanical equipment hazards; Level 2 indicator "Mechanical equipment accident" Level 4 indicator "Improper operation of mechanical equipment" gets 0 points Then, each secondary indicator is scored: that is, based on the scoring results of all the specific hazard indicators of the fourth level under each secondary indicator, the comprehensive score of each secondary indicator is calculated, including: summing up the scores of all the specific hazard indicators of the fourth level corresponding to each secondary indicator (that is, the secondary indicators include those belonging to each type of personnel accident, electric shock accident, object strike accident, foundation pit collapse or instability, seepage and water inrush accident, surrounding environmental impact, fire accident, and mechanical equipment accident) to obtain the comprehensive score of the corresponding secondary indicator. Based on the above examples, the comprehensive scores of the secondary indicators are as follows: personnel accidents 40 points (i.e., the sum of the scores of all the specific hazard indicators of the fourth level corresponding to personnel accidents, which is 40 points), electric shock accidents 70 points, object strike accidents 95 points, foundation pit collapse or instability 86 points, seepage and water inrush accidents 100 points, surrounding environmental impact 100 points, fire accidents 0 points, and mechanical equipment accidents 80 points.

[0066] The inspection revealed that the safety hazard occurred during the underground structure construction phase of the project; therefore, the weighting was selected as follows:

[0067] The weighted score is then used to output a dynamic security index.

[0068] At the same time, combined with the following Figure 3Analysis shows that the overall dynamic safety index of the project is 77.9 points, which is within the range of a mild warning. Therefore, the project is currently in a mild warning state, with general risks, and a general inspection mechanism needs to be activated.

[0069] Example 2 See Figure 5 Accordingly, the present invention provides a dynamic risk index assessment system for foundation pit engineering based on multi-source heterogeneous data fusion, comprising: Data acquisition module 10 is used to acquire multi-source heterogeneous data from on-site inspections, monitoring systems and video inspections. The multi-source heterogeneous data includes personnel safety, pit operation, environmental safety and mechanical equipment corresponding hazard record data. The data mapping and scoring module 20 is used to identify potential events in the current multi-source heterogeneous data; map the identified potential events in the multi-source heterogeneous data to scoreable nodes in a preset four-level indicator system, the four-level indicator system including first-level indicators, second-level indicators, third-level indicators and fourth-level specific potential event indicators (i.e., fourth-level indicators); and score each mapped fourth-level specific potential event indicator according to dynamic scoring rules that match the indicator type. The secondary scoring module 30 is used to calculate the comprehensive score of each secondary indicator based on the scoring results of all specific hazard indicators of the fourth level under each secondary indicator. The weight acquisition module 40 is used to establish a dynamic weight model. The dynamic weight model dynamically calculates and adjusts the weight coefficients of each secondary indicator based at least on the current construction stage and the changing trends of each secondary indicator in the recent evaluation period. The index calculation and early warning output module 50 is used to weight and integrate the comprehensive scores of each secondary indicator with their corresponding dynamic weights to calculate the dynamic safety index of the foundation pit project; based on the numerical range of the dynamic safety index, the current risk warning level is determined and corresponding risk control suggestions are output.

[0070] Example 3 Accordingly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in Embodiment 1.

[0071] In summary, this invention constructs a dynamic index assessment method based on the fusion of multi-source heterogeneous data, enabling the continuous expression of the risk status of foundation pit engineering using a unified indicator system and calculation logic. Compared with traditional qualitative judgment methods based on static monitoring values ​​and manual inspections, this invention firstly significantly improves the completeness and accuracy of on-site status information through the fusion processing of multi-source data such as monitoring data and video inspection data; secondly, by introducing a dynamic weight model that automatically adjusts with the construction stage and indicator trend changes, the assessment results can truly reflect the changes in the dominant risk factors under different working conditions, thereby overcoming the limitations of fixed weights and static assessments that are difficult to apply to full-process management. Furthermore, by designing matching scoring methods for different types of indicators, various engineering entity risks, behavioral risks, and environmental risks can all be quantitatively incorporated into a unified model, significantly improving the accuracy and interpretability of the assessment. The resulting dynamic safety index can continuously reflect the risk evolution process, facilitating managers to promptly identify adverse trends and implement targeted measures. It also provides more reliable quantitative basis for engineering monitoring and early warning, construction organization optimization, and regulatory evaluation, thereby enhancing the scientific rigor and foresight of foundation pit engineering risk management.

[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; those skilled in the art can modify the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for assessing the dynamic risk index of foundation pit engineering based on multi-source heterogeneous data fusion, characterized in that, Risk assessment for foundation pit engineering includes the following steps: Acquire multi-source heterogeneous data from on-site inspections, monitoring systems, and video surveillance. The multi-source heterogeneous data includes records of potential hazards related to personnel safety, foundation pit operations, environmental safety, and mechanical equipment. Identify potential hazards in the current multi-source heterogeneous data; map the identified potential hazards in the multi-source heterogeneous data to scoreable nodes in a preset four-level indicator system, the four-level indicator system including first-level indicators, second-level indicators, third-level indicators and fourth-level specific hazard indicators; score each mapped fourth-level specific hazard indicator according to dynamic scoring rules that match the indicator type; Based on the scoring results of all specific hazard indicators at the fourth level under each secondary indicator, calculate the comprehensive score of each secondary indicator; A dynamic weighting model is established, which dynamically calculates and adjusts the weighting coefficients of each secondary indicator based at least on the current construction stage and the changing trends of each secondary indicator in the recent evaluation period. The comprehensive scores of each secondary indicator are weighted and integrated with their corresponding dynamic weights to calculate the dynamic safety index of the foundation pit project; based on the numerical range of the dynamic safety index, the current risk warning level is determined and corresponding risk control recommendations are output.

2. The method according to claim 1, characterized in that, In the four-level indicator system, the first-level indicators include personnel safety, work behavior, environmental safety, and machinery and equipment; the second-level indicators of personnel safety include personnel accidents, electric shock accidents, and object strike accidents; the second-level indicators of work behavior include foundation pit collapse or instability and foundation pit leakage and water inrush accidents; the second-level indicators of environmental safety include the impact on the surrounding environment and fire accidents; and the second-level indicators of machinery and equipment include machinery and equipment accidents.

3. The method according to claim 2, characterized in that, The identification of potential events in the current multi-source heterogeneous data; The potential hazards identified from the multi-source heterogeneous data are mapped to scoreable nodes in a pre-defined four-level indicator system, including: Identify potential events in the current multi-source heterogeneous data and output result information; the result information includes text information. Then, semantic recognition is performed on the text information; By semantic analysis, the key feature words corresponding to the text information of the potential hazard event are matched with the definition description of the level 4 indicators, and then the event is classified into the corresponding scoreable nodes.

4. The method according to claim 1, characterized in that, Based on dynamic scoring rules that match the indicator type, each mapped Level 4 specific hazard indicator is scored, including: The dynamic scoring rules are pre-set in a dynamic scoring table, and different deduction or score values ​​are set for different level four specific hidden danger indicators. Among them, the scores for the following indicators are 0: "personnel not wearing safety belts", "safety passage not set up properly", "edge guardrails not in place", "temporary power supply not in accordance with regulations", "excessive excavation of foundation pit or failure to slope according to design", "fire-fighting equipment not configured or malfunctioning", "excessive stacking of combustibles" and "non-standard operation of mechanical equipment". The score for the "lack of safety helmets for workers" indicator is 15 points; The score for the "cracks or damage to the enclosure wall" indicator is 6 points.

5. The method according to claim 4, characterized in that, Based on the scoring results of all Level IV specific hazard indicators under each Level II indicator, the comprehensive score of each Level II indicator is calculated, including: summing up the scores of all Level IV specific hazard indicators corresponding to each type of accident, such as personnel accident, electric shock accident, object strike accident, pit collapse or instability, seepage and water inrush accident, impact on the surrounding environment, fire accident, and mechanical equipment accident, to obtain the comprehensive score of the corresponding current Level II indicator.

6. The method according to claim 1, characterized in that, The dynamic weighting model predefines a set of weight coefficients corresponding to the secondary indicators for each construction stage, and the sum of the weight coefficients is 1.

7. The method according to claim 6, characterized in that, The weighted fusion is achieved by multiplying the comprehensive score of each secondary indicator by its corresponding dynamic weight to obtain the weighted score of each secondary indicator, and then summing the weighted scores of all secondary indicators to calculate the dynamic security index.

8. The method according to claim 1, characterized in that, The risk warning levels include four levels: safe, mild warning, moderate warning, and high warning; among them, the mild warning status corresponds to the risk control recommendation of "there is a general risk and a general inspection mechanism needs to be activated".

9. The method according to claim 1, characterized in that, Before acquiring the multi-source heterogeneous data from on-site inspections, monitoring systems, and video surveillance, the process also includes: Multi-source heterogeneous data is collected through the following detection systems or sensors, wherein the detection systems include data from the foundation pit engineering monitoring system, video inspection system, and manual inspection system; and the sensors include at least environmental monitoring sensors and video image monitoring sensors. Simultaneously, the collected multi-source heterogeneous data undergoes a unified data formatting, cleaning, completion, noise reduction, and spatiotemporal alignment process to obtain preprocessed multi-source heterogeneous data.

10. A dynamic risk index assessment system for foundation pit engineering based on multi-source heterogeneous data fusion, characterized in that, include: The data acquisition module is used to acquire multi-source heterogeneous data from on-site inspections, monitoring systems, and video surveillance. The multi-source heterogeneous data includes records of potential hazards related to personnel safety, foundation pit operations, environmental safety, and mechanical equipment. The data mapping and scoring module is used to identify potential events in the current multi-source heterogeneous data; map the identified potential events in the multi-source heterogeneous data to scoreable nodes in a preset four-level indicator system, which includes first-level indicators, second-level indicators, third-level indicators, and fourth-level specific potential event indicators; and score each mapped fourth-level specific potential event indicator according to dynamic scoring rules that match the indicator type. The secondary scoring module is used to calculate the comprehensive score of each secondary indicator based on the scoring results of all specific hazard indicators at the fourth level under each secondary indicator. The weight acquisition module is used to establish a dynamic weight model. The dynamic weight model dynamically calculates and adjusts the weight coefficients of each secondary indicator based at least on the current construction stage and the changing trends of each secondary indicator in the recent evaluation period. The index calculation and early warning output module is used to weight and integrate the comprehensive scores of each secondary indicator with their corresponding dynamic weights to calculate the dynamic safety index of the foundation pit project. Based on the numerical range of the dynamic security index, the current risk warning level is determined, and corresponding risk management recommendations are output.

Images