A safety early warning method and system for multiple risk sources of an underground pipe gallery

Abstract

The present application relates to the technical field of urban underground infrastructure safety monitoring, and particularly relates to a safety early warning method and system for underground pipe gallery multi-risk sources, which comprises the following steps: obtaining physical state monitoring data through each sensor of the underground pipe gallery and calculating the time series activity factor of each data at the current time, then judging the fast or slow change type of the data according to the factor and extracting the corresponding characteristic parameters; then calculating the original abnormality degree and the conduction confidence between the same layer sensor nodes according to the parameters, and correcting the significant abnormality degree according to the confidence threshold; finally, generating the risk source confidence by aggregating the significant abnormality degrees of the same type data, and determining whether the pipe gallery has risks and triggering the early warning according to the comparison result of the risk source confidence and the risk threshold. The present application precisely adapts to the multi-type risk evolution characteristics of the pipe gallery, effectively solves the lag and false alarm problems of the traditional early warning, and improves the accuracy and reliability of the pipe gallery multi-risk source early warning.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring technology for urban underground infrastructure. In particular, it relates to a safety early warning method and system for underground utility tunnels with multiple risk sources. Background Technology

[0002] As a core carrier of urban infrastructure, underground utility tunnels integrate various pipeline systems such as electricity, communication, water supply, and gas. The tunnels are characterized by enclosed environments, long and narrow spaces, and interconnected multi-compartment structures. Risk sources exhibit significant spatial propagation characteristics. A single risk source can spread between compartments through airflow, heat conduction, or physical contact, easily triggering a chain of safety accidents. At the same time, risk sources within the tunnels exhibit the characteristic of multiple types coexisting. Environmental risks have the characteristic of rapid diffusion on a second-level scale, equipment risks exhibit slow accumulation on an hour-level scale, and human behavior risks exhibit sudden and discrete characteristics. It is necessary to construct a safety early warning system that can simultaneously identify multiple types of risks, accurately capture the spatial diffusion process of risks, and adapt to different risk evolution speeds.

[0003] Currently, multi-risk source early warning methods for underground utility tunnels are mainly divided into three categories: static alarm mechanisms based on a single threshold, statistical correlation analysis methods based on a unified time window, and risk classification methods based on machine learning models. However, when these methods are applied to specific scenarios in underground utility tunnels, they have significant technical limitations: they cannot simultaneously adapt to the multi-type risk evolution characteristics within the tunnel, including second-level suddenness, hour-level gradualism, and sudden discreteness, leading to problems such as delayed response to sudden risks and over-alarming of gradual risks. Summary of the Invention

[0004] To address the problems of existing three methods for early warning of multiple risk sources in underground utility tunnels, which employ a uniform time scale, rely solely on statistical correlation to determine risk transmission, and assess the risk of monitoring points in isolation, they cannot adapt to risk types with different evolution rates within the utility tunnel. This leads to misjudgment of risk transmission and underreporting of spatial diffusion risks, resulting in difficulties in ensuring the timeliness, sensitivity, and accuracy of early warnings. This invention provides solutions in the following aspects.

[0005] In the first aspect, a safety early warning method for multiple risk sources in underground utility tunnels includes: acquiring utility tunnel physical condition monitoring data from various sensors in the underground utility tunnel, calculating the temporal activity factor of the utility tunnel physical condition monitoring data corresponding to each sensor at the current moment; determining the feature type of the utility tunnel physical condition monitoring data of the corresponding sensor based on the temporal activity factor, and extracting feature parameters corresponding to the feature type, wherein the feature type includes: abrupt change features (rapid change type) and trend accumulation features (slow change type); calculating the original anomaly degree of each sensor based on the extracted feature parameters, and calculating the transmission confidence degree between sensor nodes on the same floor. The transmission confidence level comprehensively characterizes the statistical correlation strength between nodes and the physical spatiotemporal matching characteristics of risk propagation. Based on a preset confidence level threshold, the effectiveness of the transmission path between sensor nodes is determined. The original anomaly level of each sensor is spatially diffused and corrected in combination with the transmission confidence level to obtain the significant anomaly level of each sensor. The average of the significant anomaly levels of all sensors under the same monitoring type is used as the risk source confidence level corresponding to the same type of monitoring object. If the risk source confidence level of all sensors of the same type is less than the risk threshold, the pipe gallery is determined to be without risk; otherwise, the pipe gallery is determined to be at risk and an early warning is triggered.

[0006] Preferably, the calculation method of the time-series activity factor includes: Using any type of parameter in the physical monitoring data of the utility tunnel as the parameter to be analyzed, based on the preset number of historical time points of the parameter to be analyzed, the monitoring value of the parameter to be analyzed at the current moment and the monitoring value corresponding to each historical time point are obtained, the absolute difference between the monitoring value at the current moment and the monitoring value at the historical moment is calculated, and the ratio of the absolute difference to the corresponding time interval is used as the absolute value of the rate of change of the monitoring value per unit time of a single historical time point relative to the current moment. Calculate the absolute value of the rate of change of the monitored value per unit time for each of the preset number of historical time points, and sum them to obtain the total rate of change within the preset number of historical time points. Calculate the average of the total rate of change to obtain the time series activity factor of the parameter to be analyzed at the current moment.

[0007] Preferably, the step of determining the feature type of the pipe gallery physical condition monitoring data of the corresponding sensor based on the time-series activity factor includes: The upper quartile of the temporal activity factor sequence of each sensor within a preset time period is selected as the dynamic threshold. If the temporal activity factor at the current moment is greater than the dynamic threshold, the pipe gallery physical condition monitoring data of the corresponding sensor is determined to be of the fast-changing type; if the temporal activity factor at the current moment is less than or equal to the dynamic threshold, the pipe gallery physical condition monitoring data of the corresponding sensor is determined to be of the slow-changing type.

[0008] Preferably, the step of extracting the feature parameters corresponding to the feature type includes: In response to the feature type being fast-changing, the maximum absolute change, the variance of the rate of change, and the normalized information entropy of adjacent sampling moments within the time window corresponding to the preset time length are extracted as feature parameters. If the feature type is slow-changing, the long-term trend slope within the time window corresponding to the preset time length and the deviation between the current monitoring data value and the mean of all monitoring data values ​​within the preset time period are extracted as feature parameters.

[0009] Preferably, the step of calculating the original anomaly degree of each sensor includes: Since the physical monitoring data of the pipe gallery from the sensor is of the rapidly changing type, the original degree of anomaly is obtained by comparing the corresponding characteristic parameters with the historical maximum values ​​of each characteristic parameter over a preset number of days and then averaging them. Since the physical monitoring data of the pipe gallery from the sensor is of the slow-changing type, the absolute value of the corresponding characteristic parameter is taken and compared with the historical maximum value of each characteristic parameter over a preset number of days, and then the average value is calculated to obtain the original degree of anomaly.

[0010] Preferably, the step of calculating the conduction confidence among sensor nodes in the same layer includes: The mutual information of the original anomaly degree sequences of each pair of sensor nodes within the adaptive time window is calculated, and the mutual information is compared with the maximum mutual information of all node pairs in the same layer to obtain the statistical correlation value; then the spatiotemporal matching attenuation term is calculated based on the risk propagation speed between nodes, the time lag of the abnormal state, and the actual path distance in the utility tunnel BIM model; the statistical correlation value is multiplied by the spatiotemporal matching attenuation term to obtain the transmission confidence between each pair of sensor nodes. Each sensor node corresponds to a raw anomaly level at different times. The raw anomaly levels at different times are arranged in chronological order to form the raw anomaly level sequence of each sensor node.

[0011] Preferably, the step of determining the validity of the transmission path between sensor nodes includes: A confidence threshold is preset. The transmission confidence between each sensor node is compared with the confidence threshold. If the transmission confidence is greater than the confidence threshold, the transmission path between the corresponding sensor nodes is determined to be valid and the label is set to 1. If the transmission confidence is less than or equal to the confidence threshold, the transmission path between the corresponding sensor nodes is determined to be invalid and the label is set to 0.

[0012] Preferably, the calculation method for the significant anomaly degree of each sensor includes: Taking any sensor as the target sensor and other sensors as reference sensors, the risk transmission value from the reference sensor to the target sensor is obtained by multiplying the label of the transmission path between the target sensor and the reference sensor nodes, the validity of the transmission path, and the original anomaly degree of the reference sensor. The risk transmission values ​​from all reference sensors to the target sensor are summed to obtain the comprehensive risk value. The original anomaly level of the target sensor is added to the comprehensive risk value, and the minimum value between the sum and 1 is selected as the significant anomaly level of the target sensor.

[0013] Secondly, a safety early warning system for multiple risk sources in underground utility tunnels includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned safety early warning method for multiple risk sources in underground utility tunnels is implemented.

[0014] The present invention has the following effects: 1. This invention calculates the temporal activity factor of the utility tunnel's physical monitoring data and dynamically determines whether the data belongs to the fast-changing or slow-changing type. It then extracts matching feature parameters in a targeted manner, achieving precise adaptation of the feature extraction scale to risks with different evolutionary characteristics, such as second-level sudden events and hour-level gradual events within the utility tunnel. This effectively solves the problems of delayed response to sudden risks and excessive alarms for gradual risks caused by the uniform time scale processing of traditional methods, and significantly improves the timeliness and accuracy of identifying different types of risks.

[0015] 2. This invention constructs a transmission confidence calculation method that integrates statistical correlation strength and physical spatiotemporal matching characteristics. It combines a preset threshold to determine the validity of the transmission path between sensor nodes and performs spatial diffusion correction on the original anomaly level based on the effective transmission path. This allows the corrected significant anomaly level to simultaneously reflect the risk transmission impact of local sensor anomalies and neighboring nodes. It can accurately capture the spatial diffusion effect of pipeline corridor risks, avoid risk underreporting caused by isolated assessment monitoring points, and achieve comprehensive perception of the overall risk situation of the pipeline corridor.

[0016] 3. This invention aggregates the significant anomalies of similar sensors into risk source confidence levels, and uses the comparison between the risk source confidence levels and preset thresholds as the basis for determining the risk of underground utility tunnels and triggering early warnings. This forms a hierarchical risk determination system that ranges from single-point anomaly identification of sensors and risk transmission correction between nodes to aggregated assessment of categorized risk sources. This makes the risk determination of underground utility tunnels more in line with the actual risk evolution pattern of operation, and the early warning triggering is more targeted, effectively improving the overall reliability and practicality of early warning for multiple risk sources in underground utility tunnels. Attached Figure Description

[0017] Figure 1This is a flowchart of steps S1-S5 in a safety early warning method for multiple risk sources in underground utility tunnels according to an embodiment of the present invention.

[0018] Figure 2 This is a structural block diagram of a safety early warning system for multiple risk sources in underground utility tunnels according to an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0020] Reference Figure 1 A safety early warning method for multiple risk sources in underground utility tunnels includes steps S1-S5, as detailed below: S1: Based on the physical monitoring data of the underground utility tunnel obtained by each sensor, calculate the temporal activity factor of the physical monitoring data of the utility tunnel corresponding to each sensor at the current moment.

[0021] Personnel behavior data includes, but is not limited to: the location trajectory, movement trajectory, areas and duration of stay of maintenance personnel, access authorization, and operational behavior. Specifically, this data is acquired collaboratively through high-precision positioning base stations, access control systems, intelligent inspection terminals, and intelligent maintenance platform software systems to achieve full-process visualized control of personnel behavior. The specific data collection process includes: High-precision positioning base stations are deployed in various compartments and working shafts of the utility tunnel. Personnel entering the tunnel wear work cards or inspection terminals containing positioning chips. The positioning base stations capture personnel positioning signals in real time and upload them to the platform at a set frequency. The platform automatically generates personnel movement trajectories and dwell data. Data is collected in conjunction with the platform's inspection and maintenance module through intelligent inspection terminals (handheld devices). Inspection personnel receive inspection tasks, check in at inspection points, and upload inspection results and equipment operation records through the terminals. Terminal data is synchronized to the platform in real time, forming an operation behavior ledger. For example, the personnel behavior data collection frequency is set to 1-5 seconds / time, of which positioning trajectory data is collected at a high frequency of 1-3 seconds / time to ensure the continuity of personnel movement trajectories. Operation behavior and entry / exit data are collected in an event-triggered manner and uploaded to the platform in real time when the behavior occurs.

[0022] Equipment status data includes, but is not limited to: operating parameters, working status, energy consumption data, and fault information of various operating, monitoring, and emergency equipment. Specifically, it is acquired through the linkage of monitoring nodes, smart sensors, instruments, multi-functional monitoring base stations deployed in each compartment of the utility tunnel with the equipment management, safety management, and inspection and maintenance modules of the intelligent operation and maintenance platform. This enables real-time online perception of equipment status, remote control, and closed-loop fault management. Specific data acquisition links include: Deploy intelligent monitoring nodes at power equipment such as ventilation equipment and drainage pumps to directly collect the operation parameters and status of the equipment, and upload them to the Equipment Status Module of the Platform Equipment Management in real time through an industrial-grade communication network. The platform supports remote start / stop and mode adjustment of the equipment, and the operation results are synchronously recorded as equipment status data; collect the working status and operation parameters of sensors, intelligent manhole covers, and power equipment through multi-functional monitoring base stations and intelligent meters. Among them, core parameters such as cable temperature, voltage, and current are collected in real time by dedicated sensors, and the data is uploaded to the Power Monitoring Module of the Platform Security Management to form an operation ledger of power equipment; the collection frequency is obtained according to the specific equipment data. Exemplarily, for relatively stable operating states such as lighting systems and fire protection equipment, it is 1 - 5 seconds per time, while for key equipment such as ventilation equipment, cables, and power equipment, it is 10 - 30 seconds per time.

[0023] Environmental parameter data includes but is not limited to: temperature and humidity data, gas concentration data, water accumulation data, etc. The environmental parameter data is obtained through the linkage of multi-functional monitoring base stations deployed throughout the pipe gallery, dedicated environmental sensors, and the Safety Management and Daily Duty Modules of the intelligent operation and maintenance platform, realizing real-time monitoring, abnormal alarm, and equipment linkage of environmental parameters. The specific collection link: Uniformly deploy temperature and humidity sensors and gas sensors in each compartment and working well of the pipe gallery and other areas. All sensors are connected to the nearest multi-functional monitoring base station. The sensors collect environmental parameters in real time and transmit them to the monitoring base station. The base station uploads the data to the Temperature and Humidity Query / Temperature and Humidity Chart Module of the Safety Management of the intelligent operation and maintenance platform at the set frequency, and the platform automatically generates statistical charts and change trends of temperature, humidity, and gas concentration; deploy water level sensors in low-lying areas and drainage outlets of the pipe gallery. The water accumulation depth data collected by the sensors is transmitted to the multi-functional monitoring base station in real time and uploaded to the platform synchronously. When the water accumulation depth exceeds the preset threshold, the platform automatically triggers a water accumulation alarm; Exemplarily, environmental parameters such as temperature, humidity, and water accumulation depth are 1 - 3 seconds per time, while for flammable, explosive, and harmful gases such as methane, carbon monoxide, and hydrogen sulfide, it is 1 second per time.

[0024] Use moving window median filtering for smoothing processing, and the window size is dynamically set according to the data type; use the maximum and minimum value method to map the data to the [0,1] interval. Among them, the normalization reference takes the maximum and minimum values of the sensor data in the past 7 days; with 1 second as the unified time reference, use the forward filling strategy to align the data with different sampling frequencies in time series, and generate a standardized and time-series aligned original data set.

[0025] Select any type of data (personnel behavior data, equipment status data, and environmental parameter data) from the preprocessed raw dataset as the parameter to be analyzed. Preset the number of historical time points, obtain the current monitoring value of the parameter to be analyzed and the monitoring value corresponding to each historical time point. For each historical time point, calculate the absolute difference between the current monitoring value and the monitoring value at a historical time point. Then, use the ratio of the absolute difference to the time interval as the absolute value of the rate of change of the monitoring value per unit time relative to the current time point of a single historical time point, which represents the degree of drastic change of the monitoring value from the historical time point to the current time point.

[0026] The absolute values ​​of the rate of change of all unit time monitoring values ​​of the historical time points are summed to obtain the total rate of change within a preset number of historical time points. The average of the total rate of change is then calculated to obtain the time series activity factor of the parameter to be analyzed at the current moment.

[0027] Specifically, the time-series activity factor satisfies the following relationship: ; In the formula, Indicates the first Each sensor at the current moment Temporal activity factor; This indicates the number of historical time points. The specific value can be adjusted according to the parameter type. For example, environmental parameter data needs to capture rapidly changing characteristics. A value of 10 can be used; device status data changes relatively slowly. You can take 3; personnel behavior data You can choose 5; Indicates the first Each sensor at the current moment The monitored values; Indicates the first A sensor at a historical moment The monitored values; Indicates the current moment and the historical number. The time interval between each point in time.

[0028] In other words, Characterizing the first Each sensor at the current moment The magnitude of change relative to the average change within recent historical windows. When When the current monitored value is very close to the recent historical value, the system is in a state of instantaneous stability. This could be normal and stable operation, or it could be a phase of gradual risk accumulation (such as slow equipment aging), and the rate of change is difficult to detect within a short time window. When the current value fluctuates significantly relative to the historical window, the system is in a disturbed state. This is usually related to sudden events (such as instantaneous gas leakage or a sudden temperature rise), but it may also be caused by sensor noise or external interference.

[0029] The mean calculation is used to smooth out local extreme fluctuations within the historical time window, avoiding distortion of calculation results caused by instantaneous sensor noise and single abnormal fluctuations, and ensuring that the time series activity factor objectively analyzes the overall time series fluctuation activity at the current moment.

[0030] To further explain, considering the different temporal evolution characteristics of underground utility tunnel physical monitoring data, in order to adapt to the data characteristics and achieve accurate feature extraction, the feature type of the utility tunnel physical monitoring data is determined based on the temporal activity factor corresponding to each sensor. The specific steps are as follows: S2: Based on the time-series activity factor, determine the feature type of the pipe gallery physical state monitoring data of the corresponding sensor, and extract the feature parameters corresponding to the feature type. The feature types include: rapid change type of abrupt change feature and slow change type of trend accumulation feature.

[0031] To achieve adaptive determination of the characteristic type of utility tunnel physical monitoring data, this application selects the upper quartile of the temporal activity factor sequence of each sensor within a preset time period as a dynamic threshold to distinguish the data characteristic type. During the determination process at the current moment, if the temporal activity factor of the sensor at the current moment is greater than the dynamic threshold, the utility tunnel physical monitoring data corresponding to the sensor is determined to be a fast-changing type that adapts to abrupt change characteristics; if the temporal activity factor of the sensor at the current moment is less than or equal to the dynamic threshold, the utility tunnel physical monitoring data corresponding to the sensor is determined to be a slow-changing type that adapts to cumulative trend characteristics.

[0032] To adapt to different characteristic types of utility tunnel physical monitoring data and achieve targeted feature parameter extraction, this application extracts feature parameters within the corresponding time window based on the determined feature type. If the monitoring data is determined to be of the rapid change type, the maximum absolute change, the variance of the rate of change, and the normalized information entropy within a preset time window are extracted as feature parameters for this type. The maximum absolute change reflects the magnitude of the data mutation, the variance of the rate of change reflects the dispersion of data fluctuations, and the normalized information entropy reflects the uncertainty of the data state. If the monitoring data is determined to be of the slow change type, the long-term trend slope within a preset time window and the deviation between the current monitoring value and the average monitoring value within a preset time period are extracted as feature parameters for this type. The long-term trend slope reflects the overall trend and rate of data evolution over time, and the deviation reflects the degree of deviation of the current data value from its average level over a period of time.

[0033] S3: Calculate the original anomaly degree of each sensor based on the extracted feature parameters, calculate the transmission confidence between sensor nodes in the same layer, and the transmission confidence comprehensively characterizes the statistical correlation strength between nodes and the physical and spatiotemporal matching characteristics of risk propagation.

[0034] To address the differences in anomaly representation logic among different data feature types, this application employs a differentiated calculation method to obtain the original anomaly degree of each sensor: If the physical monitoring data of the utility tunnel is of the rapid change type, the corresponding characteristic parameters are compared with the historical maximum value of each characteristic parameter over a preset number of days (for example, the preset number of days is 7 days, but it can also be adjusted according to the specific situation), and then the average value of the multiple ratios is calculated as the original anomaly degree of the sensor. If the physical monitoring data of the utility tunnel is of the slow-changing type, first take the absolute value of each corresponding characteristic parameter, and then make a ratio with the historical maximum value of each characteristic parameter over a preset number of days (for example, the preset number of days is 7 days, but it can also be adjusted according to the specific situation). Calculate the average of the multiple ratios obtained, and use this as the original anomaly degree of the sensor.

[0035] The fast-change type directly calculates the mean based on the ratio of the feature parameter to the historical maximum value, which can intuitively reflect the magnitude and intensity of the instantaneous change in data, preserve the original change direction of the abnormal signal, and ensure the sensitivity of identifying sudden risks at the second level. The slow-change type first takes the absolute value of the feature parameter and then calculates the mean by comparing it with the historical maximum value. This can eliminate the influence of positive and negative deviation direction on the quantification of anomalies, uniformly represent the long-term deviation of data, objectively reflect the deviation of hourly gradual anomalies, and improve the accuracy and stability of gradual risk judgment.

[0036] To accurately quantify the correlation and spatiotemporal rationality of risk transmission among sensor nodes at the same level, the transmission confidence level is calculated by combining statistical correlation with spatiotemporal physical constraints: The mutual information of the original anomaly degree sequences of each pair of sensor nodes within a 60-second adaptive time window is calculated. This mutual information is then normalized with the maximum mutual information of all node pairs in the same layer to obtain a statistical correlation value that characterizes the correlation of data changes between nodes. Combining the risk propagation speed, the time lag relationship of abnormal states, and the actual physical path distance between nodes in the BIM (Building Information Modeling) model of the utility tunnel, a spatiotemporal matching attenuation term reflecting the spatiotemporal rationality of risk transmission is calculated. The statistical correlation value is multiplied by the spatiotemporal matching attenuation term to obtain the final transmission confidence degree between each pair of sensor nodes that can simultaneously reflect statistical correlation and physical rationality.

[0037] Among them, the risk propagation speed refers to the actual diffusion or transmission speed of risk factors such as gas leakage, temperature abnormality, and water accumulation within the pipe gallery; the abnormal state time lag relationship refers to the time difference between the moment when the upstream sensor shows an abnormality and the moment when the downstream sensor shows an abnormality, reflecting the temporal characteristics of risk transmission along the pipe gallery; the actual physical path distance between nodes in the pipe gallery BIM model refers to the shortest path length of the actual transmission of risk along the pipe gallery channel, determined based on the pipe gallery BIM model, rather than the straight-line distance.

[0038] By setting the above, false correlations between nodes can be filtered out, so that the transmission confidence can take into account both statistical correlation and physical rationality of the transmission of risks in the utility tunnel, thereby improving the reliability of the transmission confidence and the accuracy and anti-interference ability of multi-node joint early warning.

[0039] S4: Determine the validity of the transmission path between sensor nodes based on the preset confidence threshold, and perform spatial diffusion correction on the original anomaly degree of each sensor in combination with the transmission confidence to obtain the significant anomaly degree of each sensor.

[0040] To determine the validity of risk transmission paths between sensor nodes, this application pre-sets a confidence threshold to distinguish whether transmission is reliable. The calculated transmission confidence between each sensor node is compared with this confidence threshold. If the transmission confidence is greater than the confidence threshold, it indicates that the risk transmission relationship between nodes is significant and physically reasonable. The transmission path between the corresponding sensor nodes is determined to be valid and assigned a valid label 1. If the transmission confidence is less than or equal to the confidence threshold, it indicates that there is no real and reliable risk transmission relationship between nodes. The transmission path between the corresponding sensor nodes is determined to be invalid and assigned an invalid label 0.

[0041] For example, the confidence threshold is 0.5, which can be adjusted according to specific circumstances.

[0042] To achieve spatial diffusion correction of the original anomaly level of the sensors and accurately reflect the conduction and propagation characteristics of risks in the utility tunnel, any sensor is used as the target sensor, and all other sensors are used as reference sensors. The significant anomaly level of the target sensor is calculated: The validity label, transmission confidence, and original anomaly degree of the corresponding reference sensor are multiplied sequentially to obtain the risk transmission value of each reference sensor to the target sensor. The risk transmission values ​​of all reference sensors are accumulated to obtain the comprehensive risk value from neighboring nodes. The original anomaly degree of the target sensor itself is added to the comprehensive risk value, and the minimum value between the sum and 1 is taken as the upper limit constraint to finally obtain the significant anomaly degree of the target sensor after spatial diffusion correction.

[0043] S5: The average of the significant anomalies of all sensors under the same monitoring type is used as the risk source confidence level of the monitoring object of the same type. If the risk source confidence level of all sensors of the same type is less than the risk threshold, the utility tunnel is determined to be without risk. Otherwise, the utility tunnel is determined to be at risk and an early warning is triggered.

[0044] To achieve a quantitative assessment of overall risk from single-point sensor anomalies, this application aggregates the significant anomaly levels of all sensors under the same type of physical state monitoring: the arithmetic mean of the significant anomaly levels of each sensor under the same type is calculated, and the resulting average value is used as the confidence level of the risk source corresponding to this type of physical state monitoring, thereby objectively and stably characterizing the overall risk level of this type of monitoring object and avoiding interference from single-point anomaly fluctuations on the risk assessment results.

[0045] For example, the risk threshold is 0.3, which can be adjusted according to the specific scenario.

[0046] This invention also provides a safety early warning system for multiple risk sources in underground utility tunnels. For example... Figure 2 As shown, the system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a safety early warning method for multiple risk sources in underground utility tunnels according to the first aspect of the present invention. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface. Their configuration and functions are known in the art and will not be described in detail here.

[0047] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A safety early warning method for multiple risk sources in underground utility tunnels, characterized in that, include: Based on the physical monitoring data of the underground utility tunnel obtained by each sensor, the temporal activity factor of the physical monitoring data of the utility tunnel corresponding to each sensor at the current moment is calculated. The feature type of the pipe gallery physical condition monitoring data of the corresponding sensor is determined based on the time series activity factor, and the feature parameters corresponding to the feature type are extracted. The feature types include: rapid change type of abrupt change feature and slow change type of trend accumulation feature. The original anomaly degree of each sensor is calculated based on the extracted feature parameters, and the transmission confidence between sensor nodes in the same layer is calculated. The transmission confidence comprehensively characterizes the statistical correlation strength between nodes and the physical and spatiotemporal matching characteristics of risk propagation. The validity of the transmission path between sensor nodes is determined based on a preset confidence threshold. The original anomaly degree of each sensor is spatially diffused and corrected based on the transmission confidence to obtain the significant anomaly degree of each sensor. The average of the significant anomalies of all sensors under the same monitoring type is used as the risk source confidence level corresponding to the same type of monitoring object. If the risk source confidence level of all sensors of the same type is less than the risk threshold, the utility tunnel is determined to be without risk; otherwise, the utility tunnel is determined to be at risk and an early warning is triggered.

2. The safety early warning method for multiple risk sources in underground utility tunnels according to claim 1, characterized in that, The calculation method for the time-series activity factor includes: Using any type of parameter in the physical monitoring data of the utility tunnel as the parameter to be analyzed, based on the preset number of historical time points of the parameter to be analyzed, the monitoring value of the parameter to be analyzed at the current moment and the monitoring value corresponding to each historical time point are obtained, the absolute difference between the monitoring value at the current moment and the monitoring value at the historical moment is calculated, and the ratio of the absolute difference to the corresponding time interval is used as the absolute value of the rate of change of the monitoring value per unit time of a single historical time point relative to the current moment. Calculate the absolute value of the rate of change of the monitored value per unit time for each of the preset number of historical time points, and sum them to obtain the total rate of change within the preset number of historical time points. Calculate the average of the total rate of change to obtain the time series activity factor of the parameter to be analyzed at the current moment.

3. The safety early warning method for multiple risk sources in underground utility tunnels according to claim 1, characterized in that, The method of determining the feature type of the pipe gallery physical condition monitoring data of the corresponding sensor based on the time-series activity factor includes: The upper quartile of the temporal activity factor sequence of each sensor within a preset time period is selected as the dynamic threshold. If the temporal activity factor at the current moment is greater than the dynamic threshold, the pipe gallery physical condition monitoring data of the corresponding sensor is determined to be of the fast-changing type; if the temporal activity factor at the current moment is less than or equal to the dynamic threshold, the pipe gallery physical condition monitoring data of the corresponding sensor is determined to be of the slow-changing type.

4. A safety early warning method for multiple risk sources in underground utility tunnels according to claim 1, characterized in that, The steps for extracting feature parameters corresponding to the feature type include: In response to the feature type being fast-changing, the maximum absolute change, the variance of the rate of change, and the normalized information entropy of adjacent sampling moments within the time window corresponding to the preset time length are extracted as feature parameters. If the feature type is slow-changing, the long-term trend slope within the time window corresponding to the preset time length and the deviation between the current monitoring data value and the mean of all monitoring data values ​​within the preset time period are extracted as feature parameters.

5. A safety early warning method for multiple risk sources in underground utility tunnels according to claim 1, characterized in that, The steps for calculating the original anomaly level of each sensor include: Since the physical monitoring data of the pipe gallery from the sensor is of the rapidly changing type, the original degree of anomaly is obtained by comparing the corresponding characteristic parameters with the historical maximum values ​​of each characteristic parameter over a preset number of days and then averaging them. Since the physical monitoring data of the pipe gallery from the sensor is of the slow-changing type, the absolute value of the corresponding characteristic parameter is taken and compared with the historical maximum value of each characteristic parameter over a preset number of days, and then the average value is calculated to obtain the original degree of anomaly.

6. A safety early warning method for multiple risk sources in underground utility tunnels according to claim 1, characterized in that, The steps for calculating the conduction confidence among sensor nodes in the same layer include: The mutual information of the original anomaly degree sequences of each pair of sensor nodes within the adaptive time window is calculated, and the mutual information is compared with the maximum mutual information of all node pairs in the same layer to obtain the statistical correlation value; then the spatiotemporal matching attenuation term is calculated based on the risk propagation speed between nodes, the time lag of the abnormal state, and the actual path distance in the utility tunnel BIM model; the statistical correlation value is multiplied by the spatiotemporal matching attenuation term to obtain the transmission confidence between each pair of sensor nodes. Each sensor node corresponds to a raw anomaly level at different times. The raw anomaly levels at different times are arranged in chronological order to form the raw anomaly level sequence of each sensor node.

7. A safety early warning method for multiple risk sources in underground utility tunnels according to claim 1, characterized in that, The steps for determining the validity of the transmission path between sensor nodes include: A confidence threshold is preset. The transmission confidence between each sensor node is compared with the confidence threshold. If the transmission confidence is greater than the confidence threshold, the transmission path between the corresponding sensor nodes is determined to be valid and the label is set to 1. If the transmission confidence is less than or equal to the confidence threshold, the transmission path between the corresponding sensor nodes is determined to be invalid and the label is set to 0.

8. A safety early warning method for multiple risk sources in underground utility tunnels according to claim 1, characterized in that, The calculation methods for the significant anomalies of each sensor include: Taking any sensor as the target sensor and other sensors as reference sensors, the risk transmission value from the reference sensor to the target sensor is obtained by multiplying the label of the transmission path between the target sensor and the reference sensor nodes, the validity of the transmission path, and the original anomaly degree of the reference sensor. The risk transmission values ​​from all reference sensors to the target sensor are summed to obtain the comprehensive risk value. The original anomaly level of the target sensor is added to the comprehensive risk value, and the minimum value between the sum and 1 is selected as the significant anomaly level of the target sensor.

9. A safety early warning system for multiple risk sources in underground utility tunnels, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the safety early warning method for multiple risk sources in underground utility tunnels according to any one of claims 1-8 is implemented.

Images