Intelligent data fusion and dynamic early warning system for underground comprehensive pipe gallery

Abstract

The invention relates to the technical field of intelligent electronic data processing in an underground comprehensive pipe gallery, and discloses an underground comprehensive pipe gallery intelligent data fusion and dynamic early warning system and method, and the system comprises a data collection module which is used for receiving monitoring data containing timestamps from a plurality of sensors in the underground comprehensive pipe gallery in real time; the time sequence causal chain synchronization module dynamically constructs a causal time sequence chain between the data according to a predefined physical topological structure; the lightweight causal pruning device module predefines a causal trunk path based on physical topology and activates complete causal chain construction when the causal trunk path is abnormal; and the dynamic early warning module is used for sending out early warning information based on a causal time sequence chain analysis result. According to the method, through micro timestamp deviation calibration and dynamic causal chain construction based on physical topology, real causes of abnormal events can be accurately identified, early symptoms omissed due to neglecting of time sequence relevance of a traditional system are captured, and precious time is won for timely disposal.

Description

Technical Field

[0001] The present invention relates to an intelligent data fusion and dynamic early warning system for an underground integrated pipe gallery, belonging to the technical field of intelligent electronic data processing in the underground integrated pipe gallery. Background Art

[0002] As an important urban infrastructure, the underground integrated pipe gallery integrates various pipelines such as power, communication, water supply, drainage, and gas, playing a crucial role in ensuring the safe operation of the city and enhancing the comprehensive carrying capacity of the city. In order to ensure the safe and stable operation of the pipe gallery, it is necessary to monitor it in real time so as to timely detect and handle potential risks and hidden dangers, such as pipeline leakage, structural deformation, fire, etc. At present, the monitoring of the underground integrated pipe gallery mainly relies on deploying various sensors inside the pipe gallery, such as vibration sensors, temperature sensors, pressure sensors, etc., to collect the operation data of the pipe gallery in real time.

[0003] In the prior art, the processing and analysis of these monitoring data usually adopt an alarm mechanism based on thresholds. That is, one or more early warning thresholds are set for each sensor, and when the reading of the sensor exceeds these thresholds, the system will issue an alarm. This method is simple and intuitive, and can detect obvious abnormal situations to a certain extent. However, with the expansion of the scale of the pipe gallery and the complexity of the operating environment, this traditional monitoring method gradually shows its limitations. Specifically, the prior art mainly has the following problems:

[0004] 1. Lack of analysis of event timing and causality: Traditional threshold alarms often analyze the readings of individual sensors in isolation, ignoring the time sequence and potential causal relationships between the data of different sensors. For example, pipeline leakage may first cause slight vibration, then lead to changes in temperature and humidity, and finally a significant drop in pressure. A system based on a single threshold may only issue an alarm when the pressure drops significantly, missing the best opportunity for early warning.

[0005] 2. Difficulty in accurately locating the root cause of faults: When multiple sensors trigger alarms simultaneously, traditional systems often have difficulty determining the true root cause and specific location of the faults. For example, when multiple vibration sensors alarm simultaneously, it may be caused by external construction or changes in the internal structure of the pipe gallery. The lack of effective correlation analysis makes it difficult for operation and maintenance personnel to quickly troubleshoot and solve problems, increasing the maintenance cost and emergency response time.

[0006] 3. Susceptibility to environmental interference and high false alarm rate: The environment in the underground pipe gallery is complex and changeable, and factors such as temperature, humidity, and electromagnetic interference may affect the readings of sensors, resulting in frequent false alarms. Traditional threshold setting often has difficulty in balancing sensitivity and robustness. Too low a threshold is prone to false alarms, while too high a threshold may miss real abnormal situations.

[0007] To solve the above problems, some attempts have been made in the industry. For example, more complex rule engines are introduced to comprehensively judge by combining the readings of multiple sensors; or simple data correlation analysis is adopted to determine whether there are abnormal associations between adjacent sensors. However, these methods are still difficult to effectively process large-scale, multi-source heterogeneous monitoring data, and lack the ability to mine the event evolution logic and deep causal relationships. In addition, for underground utility tunnels with different structures and operating characteristics, these methods often require a large amount of customized development and parameter adjustment, and it is difficult to achieve generalization and rapid deployment. Summary of the Invention

[0008] The present invention provides an intelligent data fusion and dynamic warning system for an underground utility tunnel, and its main purpose is to solve the problems of the existing underground utility tunnel monitoring technology lacking in-depth analysis of event time sequence and causal relationships, resulting in untimely warning, difficult fault location, high false alarm rate, etc.

[0009] To achieve the above object, an intelligent data fusion and dynamic warning system for an underground utility tunnel provided by the present invention includes: a data acquisition module for receiving in real time the monitoring data of a plurality of sensors disposed in the underground utility tunnel, and the monitoring data includes timestamp information;

[0010] A time sequence causal chain synchronization module for dynamically constructing a causal time sequence chain between the monitoring data of the plurality of sensors according to the predefined physical topology structure of the underground utility tunnel, and restoring discrete monitoring data points into a continuous event stream to identify the causal root of abnormal events, wherein the causal time sequence chain represents the chronological order of the monitoring data in time and the causal relationship between the data;

[0011] A lightweight causal pruning module for predefined a causal backbone path based on the physical topology structure, the causal backbone path includes the causal relationship of key data nodes, and maintaining a lightweight time sequence index of the causal backbone path under normal conditions, and when abnormal data is detected, activating the causal time sequence chain synchronization module to construct a complete causal time sequence chain based on an event trigger mechanism, wherein the event trigger mechanism is judged by the following formula:

[0012]

[0013] Wherein, S i (t) represents the monitoring data of sensor i at the current moment t, represents the average value of the monitoring data of sensor i within the time window t-Δt, and θ i represents the preset abnormal threshold of sensor i;

[0014] The dynamic warning module is used to determine whether an abnormal event occurs based on the causal timing chain analysis results output by the lightweight causal pruner module, and to issue a warning message when an abnormal event is determined to occur, wherein the warning message includes the root cause location of the abnormal event.

[0015] As a preferred embodiment of the present invention, the timing causal chain synchronization module, in the process of constructing the causal timing chain, further includes: a micro-timestamp deviation calibration unit, which is used to analyze the timestamp information of the monitoring data received by the data acquisition module from different sensors. If it is detected that the deviation between the timestamps exceeds a preset threshold, the monitoring data is timestamp-calibrated based on the timing alignment algorithm to ensure the precise alignment of the monitoring data participating in the construction of the causal timing chain in the time dimension, thereby avoiding misjudgment of causal relationships due to slight time differences.

[0016] As a preferred embodiment of the present invention, the lightweight causal pruner module further includes: a dynamic causal relationship adjustment unit, which is used to analyze the causal relationship strength between each data node in the causal timing chain based on a preset causal strength evaluation model after the causal timing chain synchronization module constructs a complete causal timing chain; if the causal relationship strength is lower than a preset threshold, the causal timing chain is dynamically pruned to remove nodes with weak causal relationships to reduce the risk of misjudgment and optimize the early warning efficiency.

[0017] As a preferred embodiment of the present invention, when analyzing the causal timing chain, the dynamic warning module detects a causal chain that satisfies the following conditions:

[0018]

[0019] Among them, A j represents the abnormality degree of the jth data node in the causal time series chain, w j represents the weight coefficient of the jth data node, which is determined according to the importance of the data node in the physical topology structure, n represents the total number of data nodes in the causal timing chain, and Θ represents a preset comprehensive abnormality threshold.

[0020] As a preferred embodiment of the present invention, the warning information generated by the dynamic warning module includes the specific location where the abnormal event occurs, and the specific location is determined based on the sensor position corresponding to the abnormal data node in the causal timing chain and the physical topology structure.

[0021] As a preferred embodiment of the present invention, it further includes a semantic synchronization engine and a dynamic threshold evolution module. The semantic synchronization engine is used to resolve the semantic conflicts between the monitoring data of different sensors received by the data acquisition module. The output of the semantic synchronization engine is connected to the input of the time-series causal chain synchronization module to jointly construct a holographic data fusion including physical meaning and time dimension. The dynamic threshold evolution module is used to dynamically adjust the preset threshold for judging abnormal events in the dynamic early warning module according to the causal context information provided by the time-series causal chain synchronization module.

[0022] As a preferred embodiment of the present invention, when the dynamic early warning module issues a warning message including the root cause location of an abnormal event, it further includes: a causal path confidence evaluation unit, which is used to calculate the confidence of the root cause location of the abnormal event based on the propagation delay and signal attenuation degree of each causal relationship in the causal time-series chain constructed by the time-series causal chain synchronization module, and attach the confidence score to the warning message to assist the operation and maintenance personnel in judging the reliability of the abnormal root cause. Among them, the propagation delay and signal attenuation degree are obtained based on the statistical analysis of the historical operation data of the underground utility tunnel.

[0023] As a preferred embodiment of the present invention, the sensors collected by the data acquisition module include vibration sensors, temperature sensors, and pressure sensors.

[0024] As a preferred embodiment of the present invention, the time-series causal chain synchronization module determines the time-series relationship and potential causal relationship between data by analyzing the timestamp information of the monitoring data of different sensors and the connection relationship indicated by the physical topology structure.

[0025] As a preferred embodiment of the present invention, the predefined causal backbone path in the lightweight causal pruning module is determined based on the pipeline connection relationship of the underground utility tunnel, and only the data nodes located on the backbone path are synchronized in time series.

[0026] Compared with the problems described in the background art, the beneficial effects of the present invention are:

[0027] 1. By constructing a dynamic time-series causal chain for cross-sensor data, it is no longer limited to isolated instantaneous data points, and can capture the complete evolution process of events and potential causal associations. For example, it can identify the complete chain from a small vibration gradually leading to a temperature anomaly and finally causing a pressure mutation, so as to achieve earlier and more insightful warnings before the failure really occurs, avoiding the lag and misjudgment that may be caused by traditional systems relying solely on threshold matching.

[0028] 2. By adopting a lightweight causal pruning device, the prior knowledge of the inherent physical topology structure of the utility tunnel is utilized to pre-define the main path of data analysis, and the full-link analysis is only activated when necessary, greatly reducing the computing burden of the system under normal conditions, enabling complex data fusion and causal analysis to operate efficiently on edge computing devices with limited resources, and avoiding excessive dependence on high-cost cloud computing resources.

[0029] 3. Aiming at the problems of environmental noise and sensor drift, through causal chain analysis with physical topology constraints, the present invention sets up a natural barrier for data analysis, which can automatically filter out interference signals irrelevant to the actual operation state of the utility tunnel. For example, the system will prioritize the correlation of sensor data on adjacent pipelines and automatically weaken the influence of vibration signals far from key equipment, thus greatly improving the signal-to-noise ratio of early warning information and enabling edge operation and maintenance personnel to focus more on abnormal situations that really need attention.

[0030] 4. Taking the physical topology structure of the utility tunnel as the basis for defining causal relationships and supporting direct import from BIM models, when facing underground utility tunnels with different structures, it shows strong adaptability and replicability. It can quickly deploy and apply the intelligent early warning function without re-performing complex model training or parameter tuning for each utility tunnel, greatly reducing the technical threshold and implementation cost of the intelligent upgrade of the utility tunnel, accelerating the popularization of smart utility tunnels, and accurately identifying the real cause of abnormal events and capturing early signs missed by traditional systems due to ignoring temporal correlations through micro time stamp deviation calibration and dynamic causal chain construction based on physical topology, thus winning valuable time for timely disposal. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is the pruning flow chart of the optimized causal time sequence chain of the present invention;

[0032] Figure 2 It is the schematic diagram of data interaction between modules of the present invention;

[0033] Figure 3 It is the overall architecture diagram of the intelligent data fusion and dynamic early warning system for underground utility tunnels of the present invention;

[0034] Figure 4 It is the schematic diagram of time sequence index of key nodes of the present invention.

[0035] The implementation, functional features and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0036] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0037] An embodiment of the present application provides an intelligent data fusion and dynamic warning system for an underground integrated pipe gallery, including: a data acquisition module, configured to receive in real time the monitoring data of a plurality of sensors disposed in the underground integrated pipe gallery, and the monitoring data includes timestamp information;

[0038] A temporal causal chain synchronization module, configured to dynamically construct a causal temporal chain between the monitoring data of the plurality of sensors according to the predefined physical topology structure of the underground integrated pipe gallery, and restore discrete monitoring data points into a continuous event stream to identify the causal root of an abnormal event, where the causal temporal chain represents the chronological order of the monitoring data in time and the causal relationship between the data;

[0039] A lightweight causal pruning module, configured to predefined a causal backbone path based on the physical topology structure, the causal backbone path includes the causal relationship of key data nodes, and maintain a lightweight temporal index of the causal backbone path under normal conditions, and when abnormal data is detected, activate the causal temporal chain synchronization module to construct a complete causal temporal chain based on an event trigger mechanism, where the event trigger mechanism is judged by the following formula:

[0040]

[0041] where, S i (t) represents the monitoring data of sensor i at the current moment t, represents the average value of the monitoring data of sensor i within the time window t-Δt, and θ i represents the preset abnormal threshold of sensor i;

[0042] A dynamic warning module, configured to judge whether an abnormal event occurs based on the causal temporal chain analysis result output by the lightweight causal pruning module, and send a warning message when it is determined that an abnormal event occurs, and the warning message includes the root location of the abnormal event.

[0043] As a preferred embodiment of the present invention, in the process of constructing the causal temporal chain, the temporal causal chain synchronization module further includes: a micro timestamp deviation calibration unit, configured to analyze the timestamp information of the monitoring data received by the data acquisition module from different sensors, and if it is detected that the deviation between the timestamps exceeds a preset threshold, perform timestamp calibration on the monitoring data based on a temporal alignment algorithm to ensure the precise alignment of the monitoring data participating in the construction of the causal temporal chain in the time dimension, so as to avoid misjudgment of the causal relationship caused by a small time difference.

[0044] As a preferred embodiment of the present invention, the lightweight causal pruning module further includes: a dynamic causal relationship adjustment unit, configured to, after the causal time series chain synchronization module constructs a complete causal time series chain, analyze the causal relationship strength between each data node in the causal time series chain based on a preset causal strength evaluation model. If the causal relationship strength is lower than a preset threshold, perform dynamic pruning on the causal time series chain, removing nodes with weak causal relationships to reduce the risk of misjudgment and optimize the warning efficiency.

[0045] As a preferred embodiment of the present invention, when analyzing the causal time series chain, the dynamic warning module, if detecting a causal chain that satisfies the following conditions:

[0046]

[0047] wherein, A j represents the degree of abnormality of the j-th data node in the causal time series chain, w j represents the weight coefficient of the j-th data node, which is determined according to the importance of the data node in the physical topology structure, n represents the total number of data nodes in the causal time series chain, and Θ represents a preset comprehensive abnormality threshold.

[0048] As a preferred embodiment of the present invention, the warning information generated by the dynamic warning module includes the specific location where the abnormal event occurs, and the specific location is determined based on the sensor location corresponding to the abnormal data node in the causal time series chain and the physical topology structure.

[0049] As a preferred embodiment of the present invention, it further includes a semantic synchronization engine and a dynamic threshold evolution module. The semantic synchronization engine is used to solve the semantic conflicts between the monitoring data of different sensors received by the data acquisition module. The output of the semantic synchronization engine is connected to the input of the time series causal chain synchronization module to jointly construct a holographic data fusion including physical meaning and time dimension. The dynamic threshold evolution module is used to dynamically adjust the preset threshold for judging abnormal events in the dynamic warning module according to the causal context information provided by the time series causal chain synchronization module.

[0050] As a preferred embodiment of the present invention, when the dynamic warning module issues a warning information including the root cause location of the abnormal event, it further includes: a causal path confidence evaluation unit, configured to calculate the confidence of the root cause location of the abnormal event based on the propagation delay and signal attenuation degree of each causal relationship in the causal time series chain constructed by the time series causal chain synchronization module, and attach the confidence score to the warning information to assist the operation and maintenance personnel in judging the reliability of the abnormal root cause. Among them, the propagation delay and signal attenuation degree are obtained based on the statistical analysis of historical tunnel operation data.

[0051] As a preferred embodiment of the present invention, the sensors collected by the data acquisition module include vibration sensors, temperature sensors, and pressure sensors; the timing causal chain synchronization module determines the timing relationship and potential causal relationship between data by analyzing the timestamp information of the monitoring data of different sensors and the connection relationship indicated by the physical topology; the predefined causal backbone path in the lightweight causal pruning module is determined based on the pipeline connection relationship of the underground utility tunnel, and only the data nodes located on the backbone path are synchronized in time sequence, which all belong to the extended implementation manners known to those of ordinary skill in the art.

[0052] Embodiment 1: This embodiment specifically describes the specific process of early warning using the intelligent data fusion and dynamic warning system of the underground utility tunnel. For example, the system first receives the monitoring data of various sensors deployed in the underground utility tunnel in real time through the data acquisition module. These sensors may include, but are not limited to: temperature sensors, humidity sensors, pressure sensors, gas concentration sensors, vibration sensors, liquid level sensors, etc. Each piece of monitoring data contains accurate timestamp information, recording the specific moment when the data is generated. In order to construct the causal timing chain between data, the system pre-stores the physical topology information of the underground utility tunnel. This topology can be understood as a detailed network diagram, where each sensor or monitoring point is regarded as a node, and the physical connections such as pipelines, cable trays, and ventilation ducts are regarded as the edges between nodes. In addition, the topology information may also include attributes such as the physical distance between nodes, connection types, and importance levels during the operation of the utility tunnel. For example, the pressure sensor on the main water supply pipeline may be considered more important than the sensor monitoring the ambient temperature. The physical topology information can be pre-configured manually or automatically generated by importing the BIM (Building Information Model) data of the utility tunnel.

[0053] The timing causal chain synchronization module uses the received monitoring data with timestamps and combines the predefined physical topology to dynamically construct the causal timing chain between data. The construction process can be understood as follows: when the data of a sensor changes, the system will search for other sensors directly or indirectly connected to it according to the physical topology information and analyze their change relationships in time. For example, if an abnormal vibration is detected by a vibration sensor in a certain area, the system will search for the pressure sensor and flow sensor on the adjacent pipeline according to the physical topology information. If the values of these sensors also change accordingly subsequently (for example, the pressure drops and the flow is abnormal), the system will initially judge that there is a potential causal relationship between these data. And the timestamp information is crucial in judging the causal relationship. The system will analyze the sequence of events, and generally believes that the earlier event may be the cause of the later event.

[0054] To reduce the computational burden of the system during normal operation, the system predefines causal backbone paths based on the physical topology. These backbone paths are usually the most critical physical connections in the pipe gallery or the paths most prone to chain reactions. During normal operation of the system, the lightweight causal pruning module mainly focuses on data changes on these backbone paths. When the data acquisition module receives new monitoring data, the system first determines whether the data triggers a preset event trigger condition. For example, for sensor i, the monitoring data S i (t) at the current time t and its average value S i (t - Δt) within the past time window t - Δt exceed the preset anomaly threshold θ i , that is, it satisfies: S i (t) - S i (t - Δt) > θ i . Its function is to quickly identify sensor nodes that may have anomalies. Once the data of a certain sensor triggers this condition, the lightweight causal pruning module will be activated, and the system will no longer only focus on the predefined causal backbone paths, but will start to construct or activate a more complete causal time series chain related to the triggered sensor. This includes analyzing the data of more indirectly connected sensors and the time series relationships within a longer time range.

[0055] Based on the analysis results of the causal time series chain output by the lightweight causal pruning module, the dynamic warning module determines whether an abnormal event occurs and issues a warning message. The basis for judgment can be a comprehensive analysis of the anomaly levels of each data node in the entire causal chain and their mutual relationships. For example, the system can define an anomaly level A j for each data node in the causal time series chain. This anomaly level can be calculated based on factors such as the deviation degree of the current value of the node from the historical normal range and the change rate. In addition, considering the different importance of different sensors in the operation of the pipe gallery, the system assigns a weight coefficient w j to each data node, and this weight coefficient can be determined according to its position and importance level in the physical topology. For example, sensors located at key nodes of the main pipeline may have a higher weight.

[0056] The dynamic warning module will comprehensively consider the abnormal conditions of the entire causal chain. For example, it makes judgments in the following ways:

[0057]

[0058] Among them, n is the total number of data nodes in the causal time sequence chain, and Θ is a preset comprehensive anomaly threshold. The logic expressed by this formula is that when the sum of the weighted anomaly degrees of each node in the entire causal chain exceeds a certain threshold, the system will determine that an abnormal event requiring early warning has occurred and send out corresponding early warning information. In this way, the system can not only monitor the anomalies of individual sensors, but more importantly, analyze the propagation and correlation of abnormal events in the entire utility tunnel system, so as to more accurately identify potential risks and provide more insightful early warnings, avoiding false alarms or missed alarms that may be caused by traditional systems relying solely on a single threshold judgment. At the same time, using the physical topology structure for causal analysis also provides an important basis for locating the root cause of faults.

[0059] Embodiment 2: Refer to Figure 1 , which is the pruning flowchart of the optimized causal time sequence chain of the present invention. It starts with receiving the complete causal time sequence chain, and then the system evaluates the causal strength between adjacent data nodes and judges whether the causal strength is lower than the threshold?; if the judgment result is yes, then prune this weak causal relationship, if the judgment result is no, then retain this causal relationship; then, the system judges whether all causal relationships have been evaluated, if the judgment result is no, then return to continue evaluating the causal strength between adjacent data nodes, if the judgment result is yes, then output the optimized causal time sequence chain, and the process ends. Figure 2 is the schematic diagram of data interaction between modules of the intelligent data fusion and dynamic early warning system for underground utility tunnels proposed by the present invention. The figure shows the data flow relationship between the data acquisition module, the lightweight causal pruning module (LCP), and the time sequence causal chain synchronization module (TCCS). First, the data acquisition module receives the monitoring data and sends the monitoring data to the time sequence causal chain synchronization module (TCCS). After detecting abnormal data (meeting the trigger condition), the time sequence causal chain synchronization module (TCCS) sends a request to activate and construct a complete causal time sequence chain to the lightweight causal pruning module (LCP), and then the lightweight causal pruning module (LCP) returns the constructed complete causal time sequence chain to the time sequence causal chain synchronization module (TCCS). Figure 3 is the overall architecture diagram of the intelligent data fusion and dynamic early warning system for underground utility tunnels proposed by the present invention. This system is mainly composed of four core modules, namely the data acquisition module, the time sequence causal chain synchronization module (TCCS), the lightweight causal pruning module (LCP), and the dynamic early warning module. These modules work together to achieve the comprehensive monitoring and intelligent early warning functions of the operation status of underground utility tunnels. Figure 4 is the schematic diagram of the time sequence index of key nodes. The figure shows sensor A (key node), sensor B (key node), and sensor C (key node), indicating that the system will maintain the time sequence relationship from sensor A to sensor B and then to sensor C under normal conditions for lightweight causal analysis.

[0060] Example 3: After the data acquisition module obtains the monitoring data containing timestamps, the time-series causal chain synchronization module first performs micro-timestamp deviation calibration. Specifically, to address the possible clock drift or synchronization errors of different sensors, the following method is adopted in this example: Assume that a high-precision time synchronization server is deployed in the underground integrated pipe gallery, and each sensor periodically synchronizes time with this server. For the timestamp TS i (t) of the data collected by sensor i at time t, the deviation δ std (t) between it and the standard time T i (t) can be estimated in the following way: During the synchronization period, record the multiple synchronization time differences between sensor i and the time synchronization server, and conduct statistical analysis, such as calculating the average deviation and the deviation change rate. Based on these statistical information, a time-varying deviation model δ i (t) = f(t, parameters) can be established, where parameters include the average deviation, drift rate, etc. The calibrated timestamp TS' i (t) can be expressed as: TS' i (t) = TS i (t) - δ i (t). In this way, the time deviation between the data of different sensors can be minimized to lay a foundation for constructing an accurate time-series causal chain.

[0061] In the event trigger mechanism, for the determination of the time window Δt in the formula for quickly identifying the sensor nodes that may have anomalies: The size of the time window Δt should be adaptively adjusted according to the characteristics of different sensor types and monitoring parameters. For example, for temperature sensors, whose changes are usually relatively slow, Δt can be set relatively long (e.g., 5 - 10 minutes); while for gas concentration sensors, whose changes may be relatively rapid, Δt should be set shorter (e.g., 1 - 3 minutes). Specifically, based on historical data analysis of the fluctuation frequency and periodicity of the monitoring values of different sensors, select the best time window that can effectively capture abnormal changes and is not easily affected by noise. For example, the autocorrelation of the monitoring data can be calculated, and the time lag at which the autocorrelation significantly decreases can be selected as the reference value of Δt.

[0062] Determination of the anomaly threshold θ i : The preset anomaly threshold θ i can be set based on the statistical analysis of historical data. For sensor i, collect its monitoring data in the normal working state, and calculate its average value μ i and standard deviation σ i . The anomaly threshold θ i can be set as a multiple based on the standard deviation. For example: θi = k·σ i , where k is an adjustable parameter that can be adjusted according to the actual application scenario and the sensitivity to anomalies. For example, if it is necessary to detect potential risks earlier, k can be set to a smaller value; if false alarms are to be reduced, the value of k can be appropriately increased. In addition, more complex statistical methods can also be adopted, such as the percentile-based threshold setting method, to better adapt to the distribution characteristics of different types of data, which all belong to the extended implementation methods known to those of ordinary skill in the art.

[0063] In the dynamic warning module, in the formula for comprehensively evaluating the anomaly status of the causal time sequence chain, the anomaly degree A j is calculated as follows: The anomaly degree A j of node j can be quantified in various ways. This embodiment provides a method based on the standard deviation: Assume that the monitored value of node j at the current moment is V j (t), its average value over a past period of time is μ j , and the standard deviation is σ j , then its anomaly degree A j can be calculated as: This value represents the degree to which the current monitored value deviates from the normal range, and the larger the value, the higher the anomaly degree. And when determining the weight coefficient w j : The weight coefficient w j is used to reflect the importance of node j in the physical topology structure. The importance of a node can be evaluated from multiple dimensions. For example, for critical equipment nodes: Sensor nodes monitoring critical equipment (such as current sensors on the main power supply line, operation status sensors of important ventilation equipment) should have a higher weight; Topological location: Sensor nodes located at critical positions in the pipe gallery (such as nodes near the main road, nodes in areas prone to accidents) should also have a higher weight; Historical failure frequency: The higher the frequency of failures of this node or its nearby nodes in history, the higher its weight should be; The specific value of the weight coefficient w j can be determined by methods such as historical data analysis or graph theory-based centrality algorithms. For example, a graph representing the physical topology structure of the pipe gallery can be constructed, with nodes representing sensors and edges representing connection relationships, and then indicators such as degree centrality, betweenness centrality, or eigenvector centrality of each node are calculated, and these indicators are normalized and used as the weight coefficient, which all belong to the extended implementation methods known to those of ordinary skill in the art.

[0064] The semantic synchronization engine is responsible for semantically correlating and fusing data from different sensors. For example, when temperature sensors and smoke sensors in a certain area show anomalies simultaneously, the semantic synchronization engine can identify that this may be a fire event and generate a warning containing richer semantic information. The implementation method can be based on a predefined rule library or knowledge graph, which stores potential risk events represented by abnormal combinations of different types of sensor data and their severity levels. When the monitored data meets the rules or matches the patterns in the knowledge graph, the corresponding semantic warning can be triggered.

[0065] Dynamic threshold evolution module: The comprehensive anomaly threshold Θ should not be fixed, but should be able to be dynamically adjusted according to the operating status of the utility tunnel, historical data, and changes in the external environment. For example, during the peak operation period of the utility tunnel or under adverse weather conditions, the value of Θ can be appropriately reduced to improve the sensitivity of the system; while during the stable operation period or normal environment, the value of Θ can be appropriately increased to reduce false alarms. The method of dynamic adjustment can be based on techniques such as time series analysis and machine learning prediction. For example, establishing a prediction model of the threshold changing over time using historical data belongs to the extended implementation methods known to those of ordinary skill in the art.

[0066] Example 4: During the data acquisition process, different sensors may have clock drift or different synchronization errors, affecting the timestamp accuracy of the data. Therefore, in the time-series causal chain synchronization module, in addition to regular data synchronization, a micro-timestamp deviation calibration mechanism needs to be added to ensure the alignment accuracy of the timestamps of the data collected by different sensors. The specific steps are as follows: Each sensor (assuming sensor i) calibrates with a standard time synchronization server when collecting data in real time. For each synchronization, the system will record the deviation ΔT i between sensor i and the standard time, and obtain the deviation value ΔT i (t).

[0067] Assume that each sensor has a model based on historical deviations for predicting future timestamp errors. The model obtains the deviation value prediction function through historical data analysis: ΔT i (t) = f(t, parameters), where f(t, parameters) is the prediction model, and parameters include historical deviation data and the drift rate. The calibrated timestamp TS' i (t) is calculated by the following formula: TS' i (t) = TS i (t) - ΔT i (t). Through this calibration process, the system can effectively eliminate the small time differences between sensors, ensure data synchronization accuracy, and avoid misjudgment of causal relationships caused by time deviations.

[0068] To improve the accuracy of data fusion, especially in a multi-sensor environment, it is necessary to optimize the current anomaly triggering mechanism to ensure that the causal relationship between the data points triggered by the sensor and the surrounding data points can be accurately associated. The specific optimization steps are as follows. When the monitoring data of sensor S i (t) changes, it is first compared with the historical data of the sensor. The following formula is used to determine whether there is an anomaly: |S i (t) - S i (t - Δt)| > θ i , where S i (t) represents the sensor data at the current moment, and S i (t - Δt) is the average value within the time window, and θ i is the threshold of the sensor. If this condition is met, the anomaly judgment is triggered. In a multi-sensor system, if an anomaly is triggered by a certain sensor, the relevant sensors around it also need to perform a timing analysis to construct a causal timing chain. To avoid an excessive computational burden, the full-link causal timing chain synchronization is only performed when there are significant anomalies. In addition, based on the physical topology structure, through a predefined causal backbone path, the system focuses on the anomaly data of the key sensors in the pipe gallery, thereby reducing the computational burden under normal conditions and only activating the construction of the complete causal chain when necessary.

[0069] In the dynamic early warning module of the causal chain, when adopting a mechanism to comprehensively evaluate the anomaly situation of the causal chain nodes, it is necessary to optimize the allocation of weight coefficients and the calculation method of the anomaly degree. The specific optimization steps are as follows. When constructing the causal timing chain, for each data node A j , we calculate the anomaly degree of this node through the following formula: where V j (t) is the monitoring data of node j at the current moment, μ j is the average value of this node under normal conditions, and σ j is the standard deviation of this node. Through this standardized deviation calculation method, the anomaly degree can be more accurately evaluated. The determination of the weight coefficient w j is set according to the position and importance of each node in the physical topology structure. For example, for the sensors near key equipment, a higher weight should be given. The adjustment of the weight coefficient w j can be carried out in the following way:

[0070] w j = f(importance, distance to critical point, historical fault rate),

[0071] Among them, importance is the importance of the sensor during the operation of the pipe gallery, distance to critical point is the distance from the sensor to critical equipment or pipelines, and historical fault rate is the historical fault frequency of the sensor.

[0072] In the dynamic early warning module, the setting of the threshold is crucial for the detection of abnormal events. Therefore, in addition to the static threshold, a dynamic threshold evolution mechanism should also be introduced to dynamically adjust the threshold based on real-time data and environmental changes. During the operation of the system, the comprehensive threshold Θ for abnormal detection is dynamically adjusted according to the current operation state, environmental conditions, and historical data of the pipe gallery. For example, in response to certain emergencies, such as severe weather or equipment operation failures, the system will adjust the threshold through time series analysis, historical data, and external environmental changes to ensure that the system maintains high sensitivity at critical moments. Specifically, the dynamic threshold can be adjusted through the following steps: analyzing historical data and using machine learning models to predict future operation states; adjusting the comprehensive threshold Θ based on the prediction model. If the system is in a high-risk state (such as high temperature, heavy equipment load, etc.), the threshold is dynamically reduced to enhance the abnormal sensitivity of the system. At the same time, in order to ensure that the system has strong real-time processing capabilities and robustness, all optimization measures should be combined with edge computing technology for preliminary processing at the sensor end, and only important information is transmitted to the cloud or central server for in-depth analysis, reducing the computational burden of the system and improving the system response speed, which all belong to the extended implementation methods known to those of ordinary skill in the art.

[0073] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. 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. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An intelligent data fusion and dynamic warning system for an underground integrated pipe gallery, characterized in that Including: A data acquisition module, configured to receive in real time the monitoring data of multiple sensors deployed in an underground utility tunnel, where the monitoring data includes timestamp information; A temporal causal chain synchronization module, configured to dynamically construct a causal temporal chain between the monitoring data of the multiple sensors according to the predefined physical topology of the underground utility tunnel, and restore discrete monitoring data points into a continuous event stream to identify the causal root of abnormal events. Wherein, the causal temporal chain characterizes the chronological order of the monitoring data in time and the causal relationship between the data; A lightweight causal pruning module, configured to predefined a causal backbone path based on the physical topology, where the causal backbone path includes the causal relationship of key data nodes, and maintain a lightweight temporal index of the causal backbone path under normal conditions, and when abnormal data is detected, activate the causal temporal chain synchronization module to construct a complete causal temporal chain based on an event trigger mechanism. Wherein, the event trigger mechanism is judged by the following formula: Among them, S i (t) represents the monitoring data of sensor i at the current moment t, represents the average value of the monitoring data of sensor i within the time window t - Δt, and θ i represents the preset abnormal threshold of sensor i; A dynamic warning module, configured to judge whether an abnormal event occurs based on the analysis result of the causal temporal chain output by the lightweight causal pruning module, and send a warning message when it is determined that an abnormal event occurs, where the warning message includes the root location of the abnormal event.

2. The intelligent data fusion and dynamic early warning system for an underground integrated pipe gallery according to claim 1, wherein During the process of constructing the causal temporal chain, the temporal causal chain synchronization module further includes: a micro timestamp deviation calibration unit, configured to analyze the timestamp information of the monitoring data received by the data acquisition module from different sensors. If it is detected that the deviation between the timestamps exceeds a preset threshold, the monitoring data is timestamp calibrated based on a temporal alignment algorithm.

3. The intelligent data fusion and dynamic early warning system for the underground integrated pipe gallery according to claim 2, characterized in that, The lightweight causal pruning module further includes: a dynamic causal relationship adjustment unit, configured to analyze the causal relationship strength between the data nodes in the causal temporal chain based on a preset causal strength evaluation model after the causal temporal chain synchronization module constructs a complete causal temporal chain. If the causal relationship strength is lower than a preset threshold, the causal temporal chain is dynamically pruned to remove nodes with weak causal relationships.

4. The intelligent data fusion and dynamic early warning system for the underground integrated pipe gallery according to claim 1, wherein When analyzing the causal temporal chain, the dynamic warning module, if it detects a causal chain that meets the following conditions: Among them, A j represents the degree of abnormality of the j-th data node in the causal time sequence chain, and w j represents the weight coefficient of the j-th data node, which is determined according to the importance of the data node in the physical topology structure. n represents the total number of data nodes in the causal time sequence chain, and Θ represents a preset comprehensive abnormality threshold.

5. The intelligent data fusion and dynamic early warning system for an underground integrated pipe gallery according to claim 4, wherein The warning message generated by the dynamic warning module includes the specific location where the abnormal event occurs, and the specific location is determined based on the sensor location corresponding to the abnormal data node in the causal temporal chain and the physical topology.

6. The intelligent data fusion and dynamic early warning system for an underground integrated pipe gallery as claimed in claim 1, wherein It further includes a semantic synchronization engine and a dynamic threshold evolution module. The semantic synchronization engine is used to solve the semantic conflicts between the monitoring data of different sensors received by the data acquisition module. The output of the semantic synchronization engine is connected to the input of the temporal causal chain synchronization module to jointly construct a holographic data fusion including physical meaning and time dimension. The dynamic threshold evolution module is used to dynamically adjust the preset threshold for judging abnormal events in the dynamic warning module according to the causal context information provided by the temporal causal chain synchronization module.

7. The intelligent data fusion and dynamic warning system for an underground integrated pipe gallery according to claim 6, characterized in that, When the dynamic warning module issues a warning message containing the root cause location of an abnormal event, it further includes: a causal path confidence evaluation unit, which is used to calculate the confidence of the root cause location of the abnormal event based on the propagation delay and signal attenuation degree of each causal relationship in the causal time sequence chain constructed by the time sequence causal chain synchronization module, and attach the confidence score to the warning message to assist the operation and maintenance personnel in judging the reliability of the abnormal root cause. Among them, the propagation delay and signal attenuation degree are obtained based on the statistical analysis of the historical operation data of the utility tunnel.

8. The intelligent data fusion and dynamic early warning system for an underground integrated pipe gallery according to claim 1, wherein, The sensors collected by the data collection module include vibration sensors, temperature sensors, and pressure sensors.

9. The intelligent data fusion and dynamic early warning system for the underground integrated pipe gallery according to claim 1, characterized in that, The time sequence causal chain synchronization module determines the time sequence relationship and potential causal relationship between data by analyzing the timestamp information of the monitoring data of different sensors and the connection relationship indicated by the physical topology structure.

10. The intelligent data fusion and dynamic early warning system for an underground integrated pipe gallery according to claim 1, characterized in that, The predefined causal backbone path in the lightweight causal pruning module is determined based on the pipeline connection relationship of the underground integrated utility tunnel, and only the data nodes located on the backbone path are synchronized in time sequence.

Images