Slope stability analysis method and system based on multi-source perception

Abstract

This invention relates to the field of multi-source data analysis and processing technology, specifically to a slope stability analysis method and system based on multi-source sensing. The method includes: collecting slope monitoring data, generating a multi-source time-mapping sample set, constructing a multi-source time reference function, and using the multi-source time reference function to reconstruct the time series of the slope monitoring data; constructing event nodes according to preset event scale rules, constructing cross-scale causal constraint edges, generating multi-layer event subgraphs according to event scales, and constructing a cross-scale event sequence reconstruction map; generating an event chain sequence, selecting a set of target paths according to constraint satisfaction criteria and encoding them as a real event chain sequence; performing structural analysis and parameter statistics on the real event chain sequence, and generating slope time series analysis results. This invention achieves slope stability analysis by reconstructing multi-source monitoring data using a unified time reference and restoring the real event chain order based on a cross-scale event map.

Description

Technical Field

[0001] This invention relates to the field of multi-source data analysis and processing technology, specifically to a slope stability analysis method and system based on multi-source sensing. Background Technology

[0002] In the field of slope engineering safety monitoring, various types of monitoring channels are often deployed to obtain deformation data, pore water pressure data, acoustic emission activity data, and strain change data of the slope's interior and surface. Various monitoring devices operate with different acquisition frequencies, different triggering mechanisms, and different data upload methods, resulting in multi-source heterogeneous monitoring data. In actual monitoring, engineers usually need to analyze the evolution process inside the slope and judge potential unstable behaviors based on these multi-source monitoring data. Therefore, it is required that the monitoring data maintain consistency in the time dimension and that the temporal structure between events can accurately reflect the real change process inside the slope.

[0003] Under long-term operation of multi-source monitoring systems, there are two problems: First, the time markers of different monitoring channels drift to varying degrees, resulting in structural temporal distortion in the monitoring records. Second, due to the differences in duration, rate of change, and observation granularity of micro, meso, and macro changes within the slope, cross-scale events may be reversed in sequence or misplaced in position after being recorded, causing temporal reconstruction of the actual event chain in the monitoring data. Summary of the Invention

[0004] The purpose of this invention is to provide a slope stability analysis method and system based on multi-source sensing, so as to solve the two problems mentioned in the background art, which are structural temporal distortions in the monitoring records on the time axis caused by the varying degrees of drift of the acquisition clock of the multi-source monitoring channels.

[0005] To achieve the above objectives, the technical solution of the present invention is: a slope stability analysis method based on multi-source sensing, comprising:

[0006] S1. Collect slope monitoring data and extract the time relationship between different monitoring channels to generate a multi-source time mapping sample set. Perform fitting operation on the multi-source time mapping sample set to construct a multi-source time reference function. Use the multi-source time reference function to perform reference time series reconstruction mapping on the time identifier of the slope monitoring data to obtain the corrected time series dataset.

[0007] Among them, the multi-source time reference function is a time mapping function structure used to characterize the mapping relationship between the time identifiers of different monitoring channels and the unified time reference domain;

[0008] S2. Generate an event candidate set based on the corrected time series dataset, construct event nodes according to the preset event scale rules, analyze and calculate the time difference and scale relationship between event nodes, construct cross-scale causal constraint edges between event nodes, generate multi-level event subgraphs according to event scale and establish cross-level mapping relationships, and construct a cross-scale event sequence reconstruction graph.

[0009] Among them, event nodes are dynamic state segment coding labels determined according to preset event scale rules, used to distinguish different event categories; multi-layer event subgraphs are layer structures divided according to event scale, including micro-event subgraphs, meso-event subgraphs, and macro-event subgraphs; cross-scale event sequence reconstruction map is a data graph structure composed of event nodes, cross-scale causal constraint edges, and multi-layer event subgraphs, and supports dynamic structure updates, used to reflect the multi-scale temporal relationship of events within the slope;

[0010] S3. Based on the cross-scale event sequence, reconstruct the graph to generate an event chain sequence, perform path search on event nodes and cross-scale causal constraint relationships, select the target path set according to the constraint satisfaction criterion and encode it as a real event chain sequence;

[0011] Among them, the event chain sequence is an ordered combination of event nodes selected from the cross-scale event sequence reconstructed graph;

[0012] S4. Perform structural analysis and parameter statistics on the real event chain sequence to generate and output the slope time series analysis results.

[0013] Preferably, in step S1, the multi-source time reference function is a time mapping function structure composed of the mapping relationship between the time identifiers of different monitoring channels and a unified time reference domain. It is used to reconstruct the time identifiers of each monitoring channel in the unified time reference domain to form a corrected time series dataset. The construction method of the multi-source time reference function includes: generating a multi-source time mapping sample set using the time relationship between different monitoring channels, and performing function fitting on the mapping relationship between the time identifiers of monitoring channels and the unified time reference domain based on the multi-source time mapping sample set to form a multi-source time reference function for reconstructing and mapping the time identifiers of slope monitoring data to a reference time series.

[0014] Preferably, the structure of the multi-source time reference function is a time mapping function structure composed of a time relationship vector set, a mapping coefficient set, and a time reference domain index set; wherein, the time relationship vector set is a set of time difference vectors formed according to the time identifier correspondence between different monitoring channels; the mapping coefficient set is a set of coefficient parameters obtained by fitting the multi-source time mapping sample set; and the time reference domain index set is a set of index data used to identify each time position in a unified time reference domain; the method of the multi-source time reference function for reconstructing the time identifier of slope monitoring data into a time series includes: determining the mapping position of the time identifier of the monitoring channel in a unified time reference domain based on the time relationship vector set, performing unified mapping processing on the time identifiers of different monitoring channels based on the mapping coefficient set, and generating the mapped time identifiers based on the time reference domain index set to form a corrected time series dataset.

[0015] Preferably, in step S2, the event node is a dynamic state segment encoding unit generated based on the monitored change segments recorded in the event candidate set and combined with preset event scale rules. This unit is used to distinguish different event categories and is used to identify and classify local change segments appearing in the monitoring data during multi-scale event analysis. The structure of the event node specifically includes: a time period field, a scale label field, a state segment description field, and a node index field. The state segment description field can be regenerated according to the new change segment when the monitoring data change characteristics are adjusted, which is used to dynamically update the attributes of the event node. The scale label specifically includes micro-level, meso-level, and macro-level categories.

[0016] Preferably, in step S2, the cross-scale causal constraint edge is a causal connection unit generated based on the time difference, scale relationship, and event triggering relationship between event nodes. It is used to constrain the temporal dependency relationship between different event nodes and to establish cross-scale causal associations between event nodes in the cross-scale event sequence reconstruction graph. The specific structure of the cross-scale causal constraint edge is as follows: causal direction field, time difference field, scale association field, and constraint parameter field. The causal direction field is used to identify the causal order between event nodes, and the scale association field is used to identify the cross-layer dependency relationship between events of different scales, so as to show the cross-scale causal constraint attribute of the cross-scale causal constraint edge in the cross-scale event sequence reconstruction graph.

[0017] Preferably, in step S2, the multi-layer event subgraph is a layer structure formed by dividing event nodes into different layers according to their scale labels. This structure is used to hierarchically organize and manage event nodes of different scale categories during event analysis. The multi-layer event subgraph includes micro-event subgraphs, meso-event subgraphs, and macro-event subgraphs. The structure of the micro-event subgraph is used to record event nodes with the scale label of micro-category and their event relationships generated within the micro-scale range. The structure of the meso-event subgraph is used to record event nodes with the scale label of meso-category and their event relationships generated within the meso-scale range. The structure of the macro-event subgraph is used to record event nodes with the scale label of macro-category and their event relationships generated within the macro-scale range. Each event subgraph includes a node set field, an internal relationship field, and a layer index field for referencing node positions in cross-layer structures.

[0018] Preferably, in step S2, the cross-layer mapping relationship is established based on the hierarchical relationship between event nodes, and is used to locate associated nodes between different event subgraphs, so as to establish cross-scale event association relationships between multi-level event subgraphs. The specific method for constructing the cross-scale event sequence reconstruction map is as follows: event nodes are added to micro-event subgraphs, meso-event subgraphs and macro-event subgraphs according to their scale categories to form a layer structure. Then, cross-scale causal constraint edges are generated according to the time difference, scale relationship and event triggering relationship between event nodes, and the cross-scale causal constraint edges are added to the corresponding layer structure. On this basis, a cross-layer mapping relationship is established according to the hierarchical relationship between the scale categories of event nodes and the mapping relationship is recorded in the cross-layer mapping structure. The cross-scale event sequence reconstruction map is formed by sequentially combining event nodes, multi-level event subgraphs, cross-scale causal constraint edges and cross-layer mapping relationships.

[0019] Preferably, in step S3, the event chain sequence is an ordered combination of event nodes selected from the graph based on the causal connection relationship between event nodes in the reconstructed graph according to the cross-scale event sequence and the connection order of the cross-scale causal constraint edges. It is used to represent the temporal arrangement relationship between event nodes under a unified time reference domain. The constraint satisfaction criterion is specifically a judgment rule for judging and filtering candidate paths based on the causal direction field, time difference field, scale association field and constraint parameter field recorded in the cross-scale causal constraint edges. It is used to select the path that simultaneously satisfies the causal direction consistency, time difference range consistency, scale association matching and constraint parameter requirements from multiple candidate event connection paths as a valid event chain sequence.

[0020] On the other hand, the present invention provides a slope stability analysis system based on multi-source sensing, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the slope stability analysis method based on multi-source sensing described above.

[0021] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects:

[0022] 1. In this invention, a multi-source time reference function is established based on the reconstruction of the time relationship between multi-source monitoring channels. This allows records from different monitoring devices to be arranged in the same time reference domain, so that the monitoring data have a consistent time axis basis in subsequent analysis. This avoids the time sequence disorder caused by acquisition clock offset or recording delay, and enables slope monitoring data to be analyzed in a unified time sequence.

[0023] 2. In this invention, by constructing a cross-scale event sequence reconstruction map composed of event nodes, cross-scale causal constraint edges, and multi-layer event subgraphs, the temporal relationship of events at different scales in the monitoring data is combined, filtered, and reconstructed, so that the event sequence that was originally misplaced due to scale differences can be restored, and the event chain from micro-changes to macro-trends inside the slope can be expressed in a continuous and orderly structure. Attached Figure Description

[0024] Figure 1 This is a flowchart of an embodiment of the present invention. Detailed Implementation

[0025] Example 1, as Figure 1 As shown, the slope stability analysis method based on multi-source sensing proposed in this invention has the following specific implementation steps:

[0026] S1. Collect slope monitoring data and extract the time relationship between different monitoring channels to generate a multi-source time mapping sample set. Perform fitting operation on the multi-source time mapping sample set to construct a multi-source time reference function. Use the multi-source time reference function to perform reference time series reconstruction mapping on the time identifier of the slope monitoring data to obtain the corrected time series dataset.

[0027] Among them, the multi-source time reference function is a time mapping function structure used to characterize the mapping relationship between the time identifiers of different monitoring channels and the unified time reference domain;

[0028] S2. Generate an event candidate set based on the corrected time series dataset, construct event nodes according to the preset event scale rules, analyze and calculate the time difference and scale relationship between event nodes, construct cross-scale causal constraint edges between event nodes, generate multi-level event subgraphs according to event scale and establish cross-level mapping relationships, and construct a cross-scale event sequence reconstruction graph.

[0029] Among them, event nodes are dynamic state segment coding labels determined according to preset event scale rules, used to distinguish different event categories; multi-layer event subgraphs are layer structures divided according to event scale, including micro-event subgraphs, meso-event subgraphs, and macro-event subgraphs; cross-scale event sequence reconstruction map is a data graph structure composed of event nodes, cross-scale causal constraint edges, and multi-layer event subgraphs, and supports dynamic structure updates, used to reflect the multi-scale temporal relationship of events within the slope;

[0030] S3. Based on the cross-scale event sequence, reconstruct the graph to generate an event chain sequence, perform path search on event nodes and cross-scale causal constraint relationships, select the target path set according to the constraint satisfaction criterion and encode it as a real event chain sequence;

[0031] Among them, the event chain sequence is an ordered combination of event nodes selected from the cross-scale event sequence reconstructed graph;

[0032] S4. Perform structural analysis and parameter statistics on the real event chain sequence to generate and output the slope time series analysis results.

[0033] In this embodiment S1, slope monitoring data refers to the set of raw observation data acquired by various monitoring channels deployed on the slope structure to be analyzed during the monitoring period. The slope monitoring data includes observation values ​​recorded by monitoring channel, acquisition time identifiers, and necessary equipment operating status fields. Among them, the observation values ​​may include, but are not limited to, displacement observation values, pore water pressure observation values, acoustic emission event parameters, strain observation values, and relevant parameters of the slope surface or deep environment. All kinds of observation values ​​form corresponding monitoring records in time series form, which are continuously collected and uploaded by different monitoring channels during operation to form complete slope monitoring data.

[0034] In this embodiment S1, the multi-source acquisition of slope monitoring data can be implemented using a multi-channel synchronous acquisition architecture, an asynchronous acquisition architecture, or a hybrid acquisition architecture. In the synchronous acquisition architecture, each monitoring channel performs the acquisition and reporting of monitoring data at the same time through a unified trigger signal. In the asynchronous acquisition architecture, each monitoring channel performs independent acquisition and uploads data at its own set sampling frequency. In the hybrid acquisition architecture, some monitoring channels use the synchronous trigger mode while others use the independent sampling mode. During the data access phase, the system uniformly records the channel identifier and time identifier of each monitoring record, so that the resulting slope monitoring data has distinguishability and indexability.

[0035] In this embodiment S1, different monitoring channels refer to monitoring units used to observe different physical quantities of the slope or different monitoring areas. Each monitoring channel consists of an independent monitoring device or monitoring module and collects corresponding physical quantities in a fixed observation method. Different monitoring channels may include, but are not limited to, displacement monitoring channels, pore water pressure monitoring channels, acoustic emission monitoring channels, strain monitoring channels, temperature and humidity monitoring channels, and slope underground structure monitoring channels. Each monitoring channel undertakes different observation tasks in the slope structure and generates monitoring records that have physical meaning and are interconnected.

[0036] In this embodiment S1, the time relationship between different monitoring channels refers to the relative time position relationship between the times when different monitoring channels complete data acquisition within the same monitoring cycle. This relative time position is determined by the time stamp recorded by the clock inside the monitoring channel. Due to the use of different device clocks, sampling frequencies, or reporting mechanisms, there may be fixed offsets, dynamic offsets, or long-term drift phenomena between the time stamps of different monitoring channels. When constructing a multi-source time mapping sample set, the system uses the synchronous or near-synchronous acquisition behavior of each monitoring channel during the monitoring process to extract these relative time information, so that the time relationship between the original time stamps of different monitoring channels can be quantified and used for fitting the subsequent time mapping function.

[0037] In this embodiment S1, the multi-source time mapping sample set is a set of samples extracted from slope monitoring data to represent the correspondence between the original time identifiers of different monitoring channels and the time identifiers of the unified time reference domain. Each sample record in the sample set includes a monitoring channel identifier, an original time identifier, and a unified time reference domain time identifier that corresponds to or is approximately corresponding to the original time identifier. The multi-source time mapping sample set is used to model the time offset, time drift, and asynchronous acquisition behavior of different monitoring channels and to construct the mapping basis under the unified time reference domain.

[0038] In this embodiment S1, the data structure of the multi-source time mapping sample set can adopt a triplet record structure or a key-value pair structure. In one optional structure, the multi-source time mapping sample set includes several records. Each record has a monitoring channel identifier field, an original time identifier field, and a unified time reference domain time identifier field. Each field can be stored using an integer index, a floating-point timestamp, or a standard time string. In another optional structure, a key-value mapping structure can also be adopted, with the monitoring channel identifier as the key and the original time identifier sequence and the reference time identifier sequence as the value, so that the time correspondence can be quickly retrieved and fitted when constructing the multi-source time reference function.

[0039] In this embodiment S1, the multi-source time reference function is a time mapping function structure composed of the mapping relationship between the time identifiers of different monitoring channels and a unified time reference domain. It is used to reconstruct the time identifiers of each monitoring channel in the unified time reference domain to form a corrected time series dataset. The construction method of the multi-source time reference function includes: generating a multi-source time mapping sample set using the time relationship between different monitoring channels, and performing function fitting on the mapping relationship between the time identifiers of monitoring channels and the unified time reference domain based on the multi-source time mapping sample set to form a multi-source time reference function for reconstructing and mapping the time identifiers of slope monitoring data to a reference time series.

[0040] In this embodiment S1, the time identifier of different monitoring channels refers to the time field recorded by various monitoring channels deployed on the target slope when a data acquisition or data upload is completed. Different monitoring channels may include displacement monitoring channels, pore water pressure monitoring channels, acoustic emission monitoring channels, and strain monitoring channels, etc. Each monitoring channel carries a time identifier corresponding to the acquisition behavior when acquiring raw monitoring data. This time identifier can be the acquisition time tag recorded by the local clock of the data acquisition device, the sampling time index tag written after the data acquisition is completed, or the reception time tag recorded by the host when the data is uploaded to the slope monitoring host. In specific implementation, the time identifiers of different monitoring channels can be stored in a unified timestamp format or in a combination format of "date + hour, minute, second", but they all have a monotonically increasing time series characteristic, which is used to identify the order of different monitoring records of the same monitoring channel in the time dimension.

[0041] In this embodiment S1, the unified time reference domain refers to a unified time scale used to align and reconstruct the time identifiers of different monitoring channels. In this embodiment, the unified time reference domain can be constructed using the system clock of the slope monitoring host or using a standard time system calibrated by the time synchronization service. The unified time reference domain can be represented as a continuous time axis. Starting from the start time of slope monitoring, the time axis is divided into a series of reference time indices according to a fixed time step. Each reference time index corresponds to a unique unified time identifier, which is used as the target mapping space for the time identifiers of multi-source monitoring channels, so that the acquisition time of different monitoring channels is comparable and arrangable within the unified time reference domain.

[0042] In this embodiment S1, the mapping relationship between the time identifiers of different monitoring channels and the unified time reference domain refers to the one-to-one or approximate correspondence established between the original time identifier of each monitoring channel and the reference time index on the unified time reference domain. This mapping relationship is obtained in this embodiment by constructing a multi-source time mapping sample set. The multi-source time mapping sample set includes several sample records, each of which contains at least the original time identifier of a monitoring channel and a unified time reference domain time identifier that physically corresponds to or is as close as possible to the original time identifier in the acquisition process. For example, a time synchronization phase can be set at the initial stage of system operation, causing each monitoring channel to trigger a time synchronization phase simultaneously. In each acquisition operation, the slope monitoring host records the time identifier on the unified time reference domain and pairs it with the local time identifier reported by each monitoring channel to form a raw-reference time sample pair. Alternatively, during continuous monitoring, mechanisms such as periodic synchronization heartbeat messages and fixed-cycle acquisition tasks can be used to extract the raw time identifier and unified time reference domain time identifier generated by multiple rounds of synchronous trigger acquisition to supplement the multi-source time mapping sample set. The multi-source time mapping sample set formed in the above manner contains the correspondence between the raw time identifier and the unified time reference domain time identifier of different monitoring channels at different monitoring times, which constitutes the observation sample of the mapping relationship between the monitoring channel time identifier and the unified time reference domain.

[0043] In this embodiment S1, function fitting refers to modeling the mapping relationship between the monitoring channel time markers and the unified time reference domain based on a multi-source time mapping sample set. This allows a parameterized time mapping function structure to approximately represent the correspondence between the original time markers and the unified time reference domain time markers. In the implementation process, function fitting includes three parts: determining the structural form of the time mapping function, determining the types and number of parameters to be determined, and solving for the time mapping function parameters based on the multi-source time mapping sample set. The structural form of the time mapping function can be a single-channel independent mapping structure or a multi-channel joint mapping structure. The parameters to be determined can include offset parameters, scaling parameters, nonlinear distortion parameters, and slowly varying parameters used to describe long-term drift trends. By solving for the above parameters, the fitting error of the time mapping function on the multi-source time mapping sample set satisfies the preset convergence criterion, thereby obtaining a multi-source time reference function that can be used for subsequent reference time series reconstruction mapping.

[0044] In this embodiment S1, the mapping relationship between the monitoring channel time marker and the unified time reference domain is fitted using a function based on a multi-source time mapping sample set. This can be achieved using various existing data fitting and parameter estimation techniques in the field. In one optional implementation, the least squares method can be used to solve for the parameters of the time mapping function. The original time marker and the unified time reference domain time marker of each sample record in the multi-source time mapping sample set are used as input-output pairs. An error function is constructed with the difference between the output of the time mapping function and the unified time reference domain time marker in the sample as the objective. The set of parameters that minimizes the sum of squares of this error function over all samples is then solved. In another optional implementation, a piecewise linear fitting method can be used. The monitoring time interval is divided into several time sub-intervals, and a set of local time mapping parameters is independently fitted within each time sub-interval. To accommodate the non-uniform drift behavior of monitoring channels at different time periods, another alternative implementation method is to use spline interpolation fitting to represent the time correspondence in the multi-source time mapping sample set as a time mapping curve constrained by control points. The relationship between the original time identifier and the unified time reference domain time identifier is fitted by adjusting the position of the control points. In another alternative implementation method, a function fitting technique based on a regression model can be used, such as a time mapping scheme based on a multinomial regression model, a kernel regression model, or a machine learning regression model. The original time identifier is used as the input feature, and the unified time reference domain time identifier is used as the output target. The parameters of the time mapping function are obtained through training, and after training, the time mapping function is used as a multi-source time reference function to reconstruct the time series mapping of the time identifier of subsequent slope monitoring data.

[0045] In this embodiment S1, after the multi-source time reference function is constructed, when reconstructing and mapping the original time identifiers of each monitoring channel in the slope monitoring data, the original time identifier of each monitoring record can be input into the multi-source time reference function. The multi-source time reference function outputs the mapped time identifier of the monitoring record under the unified time reference domain, and replaces the original time identifier with the mapped time identifier to form a corrected time series dataset, so that the monitoring records of different monitoring channels have a consistent time index basis under the unified time reference domain.

[0046] In this embodiment S1, the structure of the multi-source time reference function is a time mapping function structure composed of a time relationship vector set, a mapping coefficient set, and a time reference domain index set. The time relationship vector set is a set of time difference vectors formed based on the time identifier correspondence between different monitoring channels. Each time difference vector represents the original time difference data of a pair of monitoring channels at the corresponding acquisition time, and records the time difference value and its corresponding monitoring channel index information in vector form, used to construct the basic data structure of the time relationship in the multi-source time reference function. The mapping coefficient set is a set of coefficient parameters obtained based on the fitting operation of the multi-source time mapping sample set. Each mapping coefficient characterizes the mapping ratio, offset, or combination relationship of the monitoring channel time identifier in a unified time reference domain, and is used in parameter form to generate the time mapping in the multi-source time reference function. The structure enables the time identifiers of each monitoring channel to form a consistent time series within a unified time reference domain. The time reference domain index set is a set of index data used to identify each time position within the unified time reference domain. Each index corresponds to the position number of the time identifier after the baseline time series reconstruction mapping and is used to sequentially identify the unified time reference domain in the results generated by the multi-source time reference function, thereby constructing the time series order of the corrected time series dataset. The method of the multi-source time reference function for reconstructing the baseline time series of the time identifiers of slope monitoring data includes: determining the mapping position of the time identifiers of the monitoring channels in the unified time reference domain based on the time relationship vector set; performing unified mapping processing on the time identifiers of different monitoring channels based on the mapping coefficient set; and generating the mapped time identifiers based on the time reference domain index set to form the corrected time series dataset.

[0047] In this embodiment S1, the multi-source time reference function is composed of a time relationship vector set, a mapping coefficient set, and a time reference domain index set. The time relationship vector set is used to record the relative order between the original time identifiers of different monitoring channels, the mapping coefficient set is used to record the parameters required in the time correction process, and the time reference domain index set is used to record the index sequence in the unified time reference domain, so that the multi-source time reference function can output the corresponding corrected time identifier in the unified time reference domain after inputting the original time identifier.

[0048] In this embodiment S1, the multi-source time reference function can combine the above three sets in a way that includes a preset structure combination method, a segmented structure combination method, a relationship curve combination method, and a combination method based on fitting rules. In a preset structure combination method, the records in the time relationship vector set and the parameters in the mapping coefficient set are combined in a fixed order to form a computable structure, and the output time identifier is determined by referencing the time reference domain index set. In a segmented structure combination method, the time relationship vector set can be divided into several time periods according to the changes in the monitoring period, and a corresponding mapping coefficient subset is configured for each time period, so that the multi-source time reference function can apply different parameter combinations to complete time correction in different time periods. In a relationship curve combination method, the records in the time relationship vector set can be formed into a continuous relationship description structure according to their changing trends, and the relationship form is adjusted by referencing the mapping coefficient set in this structure, so that the unified time reference domain index set can determine the output time identifier accordingly. In a combination method based on fitting rules, the parameters in the mapping coefficient set can be determined by means of sample matching or error minimization, so that the original time identifier in the time relationship vector set and the reference time identifier in the unified time reference domain index set form the closest correspondence.

[0049] In this embodiment S1, the multi-source time reference function can take several implementable forms. In one implementable form, the time mapping structure can adopt a fixed parameter form, processing the original time identifiers in the time relationship vector set through a pre-determined combination of mapping coefficients and locating the corresponding index in the unified time reference domain index set. In another implementable form, the mapping structure can adopt a time-varying segmented parameter form, dividing the monitoring period into multiple time intervals and using different combinations of mapping coefficients in each time interval, so that the mapping structure can adapt to the non-uniform time drift generated by the monitoring equipment during long-term operation. In yet another implementable form, the mapping structure can adopt a control point-based form. One approach involves selecting a representative set of records from the time relation vector set as control points and constructing a relation description structure based on the changing patterns between these control points. This structure is then adjusted using a mapping coefficient set to create a continuous structure of output time identifiers on a unified time reference domain index set. Another approach uses a self-adjusting form based on sample fitting. This involves using the corresponding records in the time relation vector set and the unified time reference domain index set as fitting samples. By adjusting the parameters in the mapping coefficient set, the difference between the mapped time identifier and the reference time identifier is gradually reduced and meets preset criteria, thus forming a time mapping structure that automatically adjusts parameters according to sample changes.

[0050] In this embodiment S1, in another feasible form, the multi-source time reference function can adopt a learning structure dominated by the changing trend of sample data. By recording the changing patterns of the time relationship vector set in different monitoring periods and adjusting the parameters in the mapping coefficient set accordingly, the multi-source time reference function can be adaptive in long-term operation and can generate a corrected time identifier that meets the monitoring requirements on a unified time reference domain index set. In all the above forms, the multi-source time reference function takes the original time identifier recorded by the time relationship vector set as input, the parameter combination in the mapping coefficient set as the calculation basis, and the unified time reference domain index set as the output reference structure to realize the reconstruction of the benchmark time series of slope monitoring data under a unified time reference domain.

[0051] In this embodiment S2, the event node is a dynamic state segment encoding unit generated based on the monitored change segments recorded in the event candidate set and combined with preset event scale rules. It is used to identify and classify local change segments appearing in the monitoring data during multi-scale event analysis. The structure of the event node specifically includes: a time period field, a scale label field, a state segment description field, and a node index field. The state segment description field can regenerate a corresponding description based on the new change segment when the monitoring data change characteristics are adjusted, which is used to dynamically update the attributes of the event node. The scale label specifically includes micro-level, meso-level, and macro-level categories.

[0052] In this embodiment S2, generating an event candidate set based on the calibration time series dataset refers to performing continuous analysis, trend analysis, and abrupt change detection on the monitoring records of different monitoring channels in the calibration time series dataset, so that monitoring segments that show obvious changes in the time series are identified as candidate events. In the change detection, commonly used methods in the field, such as rate of change analysis, time series segmentation analysis, interval difference analysis, and abrupt change point identification, can be used to form monitoring change segments that are in the same monitoring channel and have similar change trends within a certain time range. The monitoring change segments are the basic building blocks of the event candidate set.

[0053] In this embodiment S2, the preset event scale rule refers to the judgment rule that classifies candidate events into different scale categories according to their change intensity, duration or spatial influence range based on the possible range of action or change characteristics of different monitoring quantities in slope monitoring. These scale categories may include micro event scale for characterizing local subtle changes, meso event scale for characterizing medium-range changes, and macro event scale for characterizing overall trends or important changes. The preset event scale rule determines the scale category to which the candidate event should belong through threshold conditions, duration conditions and change morphology conditions.

[0054] In this embodiment S2, the monitoring change segment recorded in the event candidate set refers to the monitoring segment that is adjacent in time and has a continuous change trend extracted from the corrected time series dataset. The monitoring change segment includes the start time, end time and monitoring change characteristics within the segment, so that it can be used as the basic data unit for subsequent generation of event nodes.

[0055] In this embodiment S2, the dynamic state segment encoding unit is an encoding structure used to structurally describe the change characteristics of the monitored change segment within its scale range. The data structure of the encoding unit includes an information unit for recording the start and end time of the change segment, a category unit for recording the scale category to which the change segment belongs, a description unit for recording the internal change trend of the change segment, and an index unit for recording the reference position of the encoding unit in the subsequent graph structure, so that each dynamic state segment encoding unit can serve as an event node.

[0056] In this embodiment S2, the identification and classification of local change segments appearing in the monitoring data is achieved by combining the monitoring change segments recorded in the event candidate set with preset event scale rules to determine the scale category of the change segment, and by using the time distribution and change pattern within the change segment to generate a dynamic state segment coding unit, so that each event node has clear change segment attributes and scale attributes after generation, enabling it to be used to represent monitoring change segments at a specific scale during multi-scale event analysis.

[0057] In this embodiment S2, the structure of an event node includes a time period field, a scale label field, a state segment description field, and a node index field. The time period field is used to record the start and end times of the monitoring change segment corresponding to the event node, indicating the position of the event on the analysis time axis. The scale label field is used to record the scale category to which the event node belongs, and is used for hierarchical division in multi-level event subgraphs. The state segment description field is used to record the change trend or key indicator changes within the monitoring change segment, and is used to reflect the internal change characteristics of the event node. The node index field is used to record the reference position of the event node in the cross-scale event sequence reconstruction map.

[0058] In this embodiment S2, the dynamic update of event nodes refers to the system recalculating the change trend based on the new monitoring change segment and updating the state segment description field and time period field in the event node when the calibration time series dataset adds new monitoring records or existing monitoring records change in subsequent monitoring cycles. This enables the event node to reflect the latest changes in the monitoring data while maintaining its scale category or re-determining it according to preset event scale rules. This allows the event node to have sustainable adjustment capabilities throughout the multi-scale event analysis process and maintain consistency when updating the cross-scale event sequence reconstruction map.

[0059] In this embodiment S2, the cross-scale causal constraint edge is a causal connection unit generated based on the time difference, scale relationship, and event triggering relationship between event nodes. It is used to constrain the temporal dependency relationship between different event nodes and to establish cross-scale causal associations between event nodes in the cross-scale event sequence reconstruction graph. The specific structure of the cross-scale causal constraint edge is as follows: causal direction field, time difference field, scale association field, and constraint parameter field. The causal direction field is used to identify the causal order between event nodes, and the scale association field is used to identify the cross-layer dependency relationship between events of different scales, so as to show the cross-scale causal constraint attribute of the cross-scale causal constraint edge in the cross-scale event sequence reconstruction graph.

[0060] In this embodiment S2, the time difference between event nodes refers to the relative order between the start and end times recorded in the time period field corresponding to the two event nodes. The time difference is used to represent the temporal distance between the two event nodes in the corrected unified time reference domain, enabling the system to determine whether the preceding event node is earlier or later than the other event node in time and to provide a basis for establishing causal relationships. The scale relationship between event nodes refers to the hierarchical relationship between the scale label fields recorded by the two event nodes, used to determine whether different event nodes are in the same scale, adjacent scale, or cross-scale state, enabling the system to identify the relative position of event nodes in the multi-layer event structure and determine the establishment method of cross-scale causal constraints. The event triggering relationship refers to the influence relationship between events inferred from the change characteristics and change order recorded in the event node state segment description field, used to determine whether the change of the preceding event node provides a triggering condition for the change of the following event node, enabling the system to identify potential causal relationships in the event node structure and to generate cross-scale causal constraint edges.

[0061] In this embodiment S2, the causal connection unit used to constrain the temporal dependency between different event nodes refers to a constraint structure determined comprehensively based on the time difference, scale relationship and event triggering relationship between event nodes. This structure is used to indicate that a certain event node has a temporal dependency or triggering effect on another event node in multi-scale event analysis, so that the cross-scale event sequence reconstruction map can form event connections with causal order.

[0062] In this embodiment S2, in the structure of the cross-scale causal constraint edge, the causal direction field is a structural field used to record the causal order between event nodes. By identifying the directional relationship from the previous event node to the next event node, the event connections in the graph have a clear causal direction, which is used to ensure that the reconstructed event sequence meets the temporal consistency. The time difference field is a field used to record the temporal distance between two event nodes in the cross-scale causal constraint edge. By recording the relative time position between event nodes, the system can use it to determine the temporal closeness between events and to filter event sequences that meet the constraint conditions during the path search process. The scale association field is used to record the cross-scale... The field representing the relationship between the scale labels of two event nodes in a causal constraint edge, by characterizing types such as "same-scale association," "adjacent-scale association," or "cross-scale association," enables the system to establish cross-level dependencies between multi-level event subgraphs and make the reconstructed graph of cross-scale event sequences hierarchically related. The constraint parameter field is used to record constraint indicators such as constraint strength, triggering conditions, or association stability of cross-scale causal constraint edges. By recording parameters representing the degree of causal association between event nodes in this field, the system can select edges that meet causal conditions during the event path search process and generate event chain sequences with temporal rationality.

[0063] In this embodiment S2, by using the causal direction field, time difference field, scale correlation field, and constraint parameter field to form cross-scale causal constraint edges, the cross-scale event sequence reconstruction graph can reflect the causal relationship, temporal dependency, and cross-scale hierarchical relationship between events, providing basic structural support for the subsequent generation of event chain sequences and path search.

[0064] In this embodiment S2, the multi-layer event subgraph is a layer structure formed by dividing event nodes into different layers according to the scale label to which the event nodes belong. It is used to hierarchically organize and manage event nodes of different scale categories during event analysis. The multi-layer event subgraph includes micro-event subgraphs, meso-event subgraphs, and macro-event subgraphs. The structure of the micro-event subgraph is used to record event nodes with the scale label of micro-category and the event relationships generated within the micro-scale range. The structure of the meso-event subgraph is used to record event nodes with the scale label of meso-category and the event relationships generated within the meso-scale range. The structure of the macro-event subgraph is used to record event nodes with the scale label of macro-category and the event relationships generated within the macro-scale range. Each event subgraph includes a node set field, an internal relationship field, and a layer index field for referencing node positions in cross-layer structures.

[0065] In this embodiment S2, the scale label of the event node is a classification field generated according to preset event scale rules to identify the scale category of the event node. This classification field identifies the event node as micro-scale, meso-scale, or macro-scale by recording information such as the change characteristics, duration, and range of the corresponding monitored change segment. It is used to classify the event nodes hierarchically when constructing multi-layer event subgraphs. The layer structure formed after classifying the event nodes into different layers refers to the system adding the event nodes to the micro-event subgraph, meso-event subgraph, and macro-event subgraph according to the scale label after identifying the event node. This ensures that each subgraph contains only event nodes of the same scale category. By recording the node set field, internal correlation field, and layer index field in each subgraph, a layer structure with independent hierarchical attributes is formed, so that the event nodes of each scale category have clear hierarchical boundaries in subsequent analysis.

[0066] In this embodiment S2, hierarchical organization and management of event nodes of different scale categories means that after the multi-level event subgraph is constructed, the system groups and manages the event nodes in different layers according to the scale label to which the node belongs. This allows micro-level event nodes to be correlated in the micro-level event subgraph, meso-level event nodes to be correlated in the meso-level event subgraph, and macro-level event nodes to be correlated in the macro-level event subgraph. Within the same layer, internal correlations are established using attributes such as time difference and continuity of change trends between event nodes, so that each event subgraph can independently perform operations such as morphological recognition, node connection, and local path search.

[0067] In this embodiment S2, the micro-event subgraph is used to record event nodes with the scale label "micro". These event nodes typically correspond to subtle changes in the monitoring data. Its structure includes a node set field recording the set of micro-event nodes, an internal correlation field recording the temporal sequence of the micro-event nodes, and a layer index field indicating the subgraph's position in a multi-layer structure, enabling the micro-event subgraph to represent the temporal correlation characteristics of minute change segments. The meso-event subgraph is used to record event nodes with the scale label "meso". These event nodes typically correspond to change segments in the monitoring data that have a certain degree of persistence or localized impact. Its structure includes a node set field recording the set of meso-event nodes. The meso-level event subgraph includes a point set field, an internal correlation field recording the logical connections between meso-level event nodes, and a layer index field recording the subgraph's position in a multi-layered structure. This allows the meso-level event subgraph to represent the temporal structure and correlation patterns of intermediate-level events. The macro-level event subgraph records event nodes with a macro scale label. These event nodes typically correspond to segments of change in the monitoring data that have an overall trend or a relatively long duration. Its structure includes a node set field recording the set of macro-level event nodes, an internal correlation field recording the temporal connections between macro-level event nodes, and a layer index field recording the subgraph's position in a multi-layered structure. This allows the macro-level event subgraph to reflect the temporal structure of large-scale events.

[0068] In this embodiment S2, the cross-layer mapping relationship is established based on the hierarchical relationship between event nodes. It is used to locate the associated nodes between different event subgraphs to establish cross-scale event association relationships between multi-level event subgraphs. The specific method for constructing the cross-scale event sequence reconstruction map is as follows: event nodes are added to micro-event subgraphs, meso-event subgraphs and macro-event subgraphs according to their scale categories to form a layer structure. Then, cross-scale causal constraint edges are generated according to the time difference, scale relationship and event triggering relationship between event nodes and added to the corresponding layer structure. On this basis, a cross-layer mapping relationship is established according to the hierarchical relationship between the scale categories of event nodes and the mapping relationship is recorded in the cross-layer mapping structure. The cross-scale event sequence reconstruction map is formed by sequentially combining event nodes, multi-level event subgraphs, cross-scale causal constraint edges and cross-layer mapping relationships.

[0069] In this embodiment S2, the cross-layer mapping relationship is a mapping structure established based on the hierarchical relationship between different scale categories of event nodes, used to locate associated nodes between different event subgraphs. This mapping structure records the layer index, node index, and associated attributes of the event nodes that are related to each other in different event subgraphs, enabling the system to retrieve other nodes with cross-scale associations to the current node based on the cross-layer mapping relationship when performing cross-scale event analysis. Establishing cross-scale event association relationships between multi-level event subgraphs is achieved by combining the time difference, scale relationship, and triggering relationship between event nodes to form cross-scale causal constraint edges, and determining the event nodes that the cross-scale causal constraint edges should connect based on the cross-layer mapping relationship. This allows cross-scale event association relationships to be correctly established between micro-level event subgraphs, meso-level event subgraphs, and macro-level event subgraphs, thereby forming a multi-level cross-scale event connection, enabling the cross-scale event sequence reconstruction map to form an overall event structure with hierarchical associations based on the multi-level graph structure.

[0070] In this embodiment S3, the event chain sequence is an ordered combination of event nodes selected from the graph based on the causal connection relationship between event nodes in the reconstructed graph according to the cross-scale event sequence and the connection order of the cross-scale causal constraint edges. It is used to represent the temporal arrangement relationship between event nodes under a unified time reference domain. The constraint satisfaction criterion is specifically a judgment rule for judging and filtering candidate paths based on the causal direction field, time difference field, scale association field and constraint parameter field recorded in the cross-scale causal constraint edges. It is used to select the path that simultaneously satisfies the causal direction consistency, time difference range consistency, scale association matching and constraint parameter requirements from multiple candidate event connection paths as a valid event chain sequence.

[0071] In this embodiment S3, generating an event chain sequence based on the cross-scale event sequence reconstruction graph refers to performing path search on event nodes and their cross-scale causal constraint edges in the cross-scale event sequence reconstruction graph, and combining event nodes that satisfy the constraint satisfaction criteria according to their causal order in the graph to form an event chain sequence. In the path search, commonly used graph traversal techniques, sequential search techniques, and hierarchical search techniques based on hierarchical graph structures can be used to enable the system to identify possible event connection paths from the graph based on the connection method of cross-scale causal constraint edges.

[0072] In this embodiment S3, path search for event nodes and cross-scale causal constraints refers to starting from an event node in the graph as the starting node, and gradually visiting the next event node with a causal connection relationship with the current node based on the causal direction field recorded in the cross-scale causal constraint edge. During the visit, the time difference field, scale association field, and constraint parameter field are checked to see if they meet the path extension conditions. In the path search, depth-first search, breadth-first search, or hierarchical first search can be used to enable the system to gradually expand the path until a complete candidate path is formed while maintaining the causal order.

[0073] In this embodiment S3, the causal direction consistency in the constraint satisfaction criteria means that the causal direction fields of all cross-scale causal constraint edges in the candidate path must maintain the same direction, so that the node arrangement order in the event chain will not have reverse connections. The time difference range consistency means that the time difference fields recorded in the cross-scale causal constraint edges must be within a preset reasonable range, so that adjacent events in the event chain have adjacent or continuous time relationships. The scale association matching means that the scale labels between adjacent event nodes in the path must be consistent with the scale association fields recorded in the cross-scale causal constraint edges, so that the cross-scale event chain has the correct scale hierarchy structure. The constraint parameter requirements mean that the constraint parameter fields recorded in the cross-scale causal constraint edges must meet the preset parameter conditions, so that the event nodes in the path have an association degree that conforms to the causal connection rules.

[0074] In this embodiment S3, the effective event chain sequence refers to the ordered combination of event nodes whose constituent nodes satisfy the requirements of causal direction consistency, time difference range consistency, scale correlation matching, and constraint parameters after path search. This event chain sequence is used to represent the temporal connection relationship between cross-scale events in the monitoring data under a unified time reference domain.

[0075] In this embodiment S4, structural parsing of the event chain sequence refers to structural analysis of the sequential relationships, scale hierarchy relationships, and causal connections of the event nodes in the event chain sequence. This enables the system to identify key nodes, key connection segments, and event hierarchy change structures in the event chain sequence. Chain-based structural analysis, hierarchical relationship parsing, and sequential relationship organization techniques can be used in the structural parsing process to convert the event chain sequence into a structured representation suitable for parameter statistics and pattern recognition. Parameter statistics of the event chain sequence refer to statistical analysis of the duration, number of event nodes, number of event hierarchy changes, number of cross-scale connections, and constraint feature distribution of the event chain sequence based on the event node time period field, scale label field, and time difference field and constraint parameter field of the cross-scale causal constraint edge records. This allows the system to extract parameters reflecting the overall trend of slope change or local instability characteristics from the event chain sequence. Chain-based parameter statistics, interval statistics, and hierarchical statistics techniques can be used in the parameter statistics to give the statistical results a structured form that can be used for time series analysis.

[0076] In this embodiment S4, the slope time series analysis results specifically include the structural analysis results of the event chain sequence, the parameter statistical results, the identification information of key event nodes in the event chain sequence, the event hierarchy change structure information, the time series span information of the event chain, and the event connection characteristic information.

[0077] Example 2: The slope stability analysis system based on multi-source sensing proposed in this invention is applied to the slope stability analysis method based on multi-source sensing proposed in Example 1. It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the slope stability analysis method based on multi-source sensing in Example 1.

[0078] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A slope stability analysis method based on multi-source sensing, characterized in that, Includes the following steps: S1. Collect slope monitoring data and extract the time relationship between different monitoring channels to generate a multi-source time mapping sample set. Perform fitting operation on the multi-source time mapping sample set to construct a multi-source time reference function. Use the multi-source time reference function to perform reference time series reconstruction mapping on the time identifier of the slope monitoring data to obtain the corrected time series dataset. Among them, the multi-source time reference function is a time mapping function structure used to characterize the mapping relationship between the time markers of different monitoring channels and the unified time reference domain; the different monitoring channels include displacement monitoring channel, pore water pressure monitoring channel, acoustic emission monitoring channel and strain monitoring channel; S2. Generate a candidate set of events based on the corrected time series dataset, construct event nodes according to the preset event scale rules, analyze and calculate the time difference and scale relationship between event nodes, construct cross-scale causal constraint edges between event nodes, generate multi-level event subgraphs according to event scale and establish cross-level mapping relationships, and construct a cross-scale event sequence reconstruction graph. The specific method for constructing a cross-scale event sequence reconstruction graph is as follows: Event nodes are added to micro-event subgraphs, meso-event subgraphs, and macro-event subgraphs according to their scale categories to form a layer structure. Then, cross-scale causal constraint edges are generated based on the time difference, scale relationship, and event triggering relationship between event nodes, and these edges are added to the corresponding layer structure. On this basis, a cross-layer mapping relationship is established based on the hierarchical relationship between the scale categories to which event nodes belong, and this mapping relationship is recorded in the cross-layer mapping structure. By sequentially combining event nodes, multi-layer event subgraphs, cross-scale causal constraint edges, and cross-layer mapping relationships, a cross-scale event sequence reconstruction graph is formed. Among them, event nodes are dynamic state segment coding labels determined according to preset event scale rules, used to distinguish different event categories; multi-layer event subgraphs are layer structures divided according to event scale, including micro-event subgraphs, meso-event subgraphs, and macro-event subgraphs; cross-scale event sequence reconstruction map is a data graph structure composed of event nodes, cross-scale causal constraint edges, and multi-layer event subgraphs, and supports dynamic structure updates, used to reflect the multi-scale temporal relationship of events within the slope; S3. Based on the cross-scale event sequence, reconstruct the graph to generate an event chain sequence, perform path search on event nodes and cross-scale causal constraint relationships, select the target path set according to the constraint satisfaction criterion and encode it as a real event chain sequence; Among them, the event chain sequence is an ordered combination of event nodes selected from the cross-scale event sequence reconstructed graph; S4. Perform structural analysis and parameter statistics on the real event chain sequence to generate and output the slope time series analysis results.

2. The slope stability analysis method based on multi-source sensing according to claim 1, characterized in that: In S1, the multi-source time reference function is a time mapping function structure composed of the mapping relationship between the time identifiers of different monitoring channels and the unified time reference domain. It is used to reconstruct the time identifiers of each monitoring channel in the unified time reference domain to form a corrected time series dataset. The construction method of the multi-source time reference function includes: generating a multi-source time mapping sample set by utilizing the time relationship between different monitoring channels, and performing function fitting on the mapping relationship between the time identifier of the monitoring channel and the unified time reference domain based on the multi-source time mapping sample set, so as to form a multi-source time reference function for reconstructing the time sequence mapping of the time identifier of the slope monitoring data.

3. The slope stability analysis method based on multi-source sensing according to claim 2, characterized in that: The structure of the multi-source time reference function is a time mapping function structure composed of a time relationship vector set, a mapping coefficient set, and a time reference domain index set. The time relationship vector set is a set of time difference vectors formed based on the time identifier correspondence between different monitoring channels; the mapping coefficient set is a set of coefficient parameters obtained by fitting a multi-source time mapping sample set; and the time reference domain index set is a set of index data used to identify each time position in a unified time reference domain. The method for reconstructing the time series of slope monitoring data using the multi-source time reference function includes: determining the mapping position of the monitoring channel's time identifier in a unified time reference domain based on the time relationship vector set; performing unified mapping processing on the time identifiers of different monitoring channels based on the mapping coefficient set; and generating mapped time identifiers based on the time reference domain index set to form a corrected time series dataset.

4. The slope stability analysis method based on multi-source sensing according to claim 3, characterized in that: In step S2, the event node is a dynamic state segment encoding unit generated based on the monitored change segments recorded in the event candidate set and combined with preset event scale rules. It is used to identify and classify local change segments appearing in the monitoring data during multi-scale event analysis. The structure of the event node specifically includes: a time period field, a scale label field, a state segment description field, and a node index field. Among them, the state segment description field can regenerate the corresponding description based on the new change segment when the monitoring data change characteristics are adjusted, which is used to dynamically update the attributes of the event node. The scale label specifically includes micro-level, meso-level, and macro-level categories.

5. The slope stability analysis method based on multi-source sensing according to claim 4, characterized in that: In S2, the cross-scale causal constraint edge is a causal connection unit generated based on the time difference, scale relationship, and event triggering relationship between event nodes. It is used to constrain the temporal dependency relationship between different event nodes and to establish cross-scale causal associations between event nodes in the cross-scale event sequence reconstruction graph. The specific structure of the cross-scale causal constraint edge is as follows: causal direction field, time difference field, scale association field, and constraint parameter field. The causal direction field is used to identify the causal order between event nodes, and the scale association field is used to identify the cross-layer dependency relationship between events of different scales, so as to show the cross-scale causal constraint attribute of the cross-scale causal constraint edge in the cross-scale event sequence reconstruction graph.

6. The slope stability analysis method based on multi-source sensing according to claim 5, characterized in that: In S2, the multi-layer event subgraph is a layer structure formed by dividing event nodes into different layers according to the scale label to which the event nodes belong. It is used to hierarchically organize and manage event nodes of different scale categories during the event analysis process. The multi-layer event subgraph includes micro-event subgraphs, meso-event subgraphs, and macro-event subgraphs. The structure of the micro-event subgraph is used to record event nodes with the scale label of micro-category and the event relationships generated within the micro-scale. The structure of the meso-event subgraph is used to record event nodes with the scale label of meso-category and the event relationships generated within the meso-scale. The structure of the macro-event subgraph is used to record event nodes with the scale label of macro-category and the event relationships generated within the macro-scale. Each event subgraph includes a node set field, an internal relationship field, and a layer index field used to reference the node position in the cross-layer structure.

7. The slope stability analysis method based on multi-source sensing according to claim 6, characterized in that: In S2, the cross-layer mapping relationship is established based on the hierarchical relationship between event nodes. It is used to locate the mapping structure of associated nodes between different event subgraphs, so as to establish cross-scale event association relationships between multi-level event subgraphs.

8. The slope stability analysis method based on multi-source sensing according to claim 7, characterized in that: In S3, the event chain sequence is an ordered combination of event nodes selected from the graph based on the causal connection relationship between event nodes in the reconstructed graph according to the cross-scale event sequence and the connection order of the cross-scale causal constraint edges. It is used to represent the temporal arrangement relationship between event nodes under a unified time reference domain. The constraint satisfaction criterion is a judgment rule that judges and filters candidate paths based on the causal direction field, time difference field, scale association field and constraint parameter field recorded in the cross-scale causal constraint edges. It is used to select the path that simultaneously satisfies the causal direction consistency, time difference range consistency, scale association matching and constraint parameter requirements from multiple candidate event connection paths as a valid event chain sequence.

9. A slope stability analysis system based on multi-source sensing, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes a computer program to implement the slope stability analysis method based on multi-source sensing as described in any one of claims 1-8.

Images