A Big Data-Based Method and System for Disaster Source Tracing and Analysis in Hard Rock Mining

Abstract

This invention discloses a method and system for tracing and analyzing the source of disasters in hard rock mining based on big data, belonging to the field of mining data processing technology. The method specifically includes: acquiring multi-source data of the hard rock mining operation area; constructing a mining disaster analysis dataset based on the multi-source data; constructing a fracture-induced disaster prior field reflecting the evolution of surrounding rock fractures based on the mining disaster analysis dataset; identifying suspected disaster source areas; constructing a disaster tracing relationship model for the suspected disaster source areas; and performing inversion inference on the disaster tracing relationship model to obtain the disaster source location, triggering factors, and triggering time. This application can reduce ambiguities in tracing caused by inconsistent time and unclear spatial references in multi-source data, providing clear data basis and operational paths for determining the disaster source area, triggering factors, and triggering time. It is applicable to hard rock mining scenarios where the causes of disasters are complex and the boundaries of responsibility are difficult to define under deep hard rock mining conditions.

Description

Technical Field

[0001] This invention belongs to the field of mining data processing technology, specifically a method and system for tracing and analyzing disaster sources in hard rock mining based on big data. Background Technology

[0002] With the large-scale development of deep hard rock mineral resources, the mining environment is characterized by high ground stress, high structural complexity, and strong superposition of disturbances. Disasters such as rock bursts, spalling, and roof falls are characterized by their suddenness and wide impact. These disasters are often caused by the long-term evolution of surrounding rock fissures and the combined effects of multiple operational disturbance factors. Their occurrence mechanism has significant spatiotemporal coupling and concealment, and a single monitoring method is insufficient to fully characterize the disaster formation process.

[0003] In existing technologies, a large amount of multi-source heterogeneous data, such as images, waveforms, event records, and geological models, has been gradually accumulated during the hard rock mining process. The main analytical methods for hard rock mining disasters include microseismic location analysis, surrounding rock stress monitoring, empirical criterion analysis, and risk assessment methods based on historical statistics. These methods mostly focus on early warning before disasters occur or on determining the location afterward. They are difficult to reconstruct the fracture evolution, disturbance events, and their causal relationships during the disaster formation process, and therefore it is difficult to achieve source correlation analysis of the disaster source area location, triggering factors, and triggering time. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes a method and system for tracing and analyzing disaster sources in hard rock mining based on big data. Under the condition of multi-source big data, the evolution process of surrounding rock fissures is correlated with operational disturbances and disaster results through modeling, thereby enabling traceable analysis of the location of disaster source areas, triggering factors, and triggering times.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] Big data-based methods for tracing and analyzing the sources of disasters in hard rock mining include:

[0007] Acquire multi-source data of the hard rock mining operation area, and construct a mining disaster analysis dataset based on the multi-source data. The multi-source data includes underground mine image or video data, microseismic or acoustic emission data, drilling and blasting construction parameter data, and geological and mining progress data.

[0008] Based on the mining disaster analysis dataset, a fracture-induced disaster prior field reflecting the evolution state of surrounding rock fracture is constructed to identify potential disaster source areas;

[0009] For the suspected source areas of the disaster, a disaster source tracing relationship model is constructed, and the disaster source tracing relationship model is inverted to obtain the source area location, triggering factors and triggering time of the disaster.

[0010] Specifically, the set of analysis objects for the hard rock mining operation area is determined, and a unified object identifier is assigned to each unit. The set of analysis objects includes stope space units, roadway topology units, operation event units, and monitoring point units.

[0011] Multi-source data is obtained based on the object identifier. The multi-source data is encapsulated into event fragments according to a preset sampling window. Each event fragment is written with a corresponding object identifier and acquisition time identifier. The multi-source data includes underground mine image or video data, microseismic or acoustic emission data, drilling and blasting construction parameter data, and geological and mining progress data. The event fragments include underground mine image or video event fragments, microseismic or acoustic emission event fragments, drilling and blasting construction parameter event fragments, and geological and mining progress event fragments.

[0012] One type of data from the multi-source data is selected as the reference time scale source; time scale drift correction and time segment resampling are performed on the remaining multi-source data so that each event segment forms a time series under the same reference time scale.

[0013] Extract the fracture skeleton, block boundary discontinuity segments, and texture abrupt change segments from images or videos in the time series, and combine them into fracture fingerprint entries for mining space units.

[0014] The crack fingerprint entries and event fragments are associated and cataloged according to the object identifier to form a mining disaster analysis dataset.

[0015] Specifically, the mining disaster analysis dataset is sliced ​​according to a preset time window, and the hard rock mining area is divided into multiple spatial units. For each spatial unit, a corresponding feature container is established in each time window. The feature container is used to carry the fracture fingerprint entries and event fragments associated with the spatial unit.

[0016] The crack fingerprint entries within each spatial unit are arranged in chronological order to generate a crack fingerprint sequence.

[0017] Using the fissure fingerprint sequence as the anchoring object, event segments associated with the same spatial unit are aligned to the fissure fingerprint sequence according to their temporal proximity. When event segments overlap or have inconsistent indications within the same time window, the overlap or inconsistency is marked and resolved according to a preset priority rule to obtain a resolved fused sequence.

[0018] Based on the resolved fusion sequence, each spatial unit is assigned a corresponding prior field value, and a topological adjacency relationship is introduced between adjacent spatial units to form a prior field grid. The prior field value is determined by the co-occurrence relationship between the topological change segment of the crack fingerprint sequence and the microseismic or acoustic emission event segment after time alignment.

[0019] In the prior field grid, a set of spatial units that satisfy a preset connectivity rule is identified, and the set of spatial units is encapsulated as a suspected disaster source area.

[0020] Specifically, based on the resolved fusion sequence, each spatial unit is assigned a corresponding prior field value, and a prior field grid is formed by introducing topological adjacency relationships between adjacent spatial units, including:

[0021] For each spatial unit, the resolved fusion sequence is segmented according to temporal continuity, and a corresponding surrounding rock state identifier is enumerated for each segment. The surrounding rock state identifier is determined by the change segment of the fracture fingerprint sequence and the event segment aligned with it.

[0022] Within the same spatial unit, the segments are sorted according to the chronological order of the surrounding rock state indicators, and a state trajectory reflecting the state migration order within the spatial unit is generated. The state trajectory serves as the internal evolution description benchmark of the spatial unit.

[0023] The state trajectory is mapped to the corresponding prior field value, and the prior field value is bound to the corresponding spatial unit. Different state trajectories correspond to different prior field value intervals, and multiple prior field value records are allowed to exist in the same spatial unit in different time windows.

[0024] The adjacency pairs between each spatial unit are determined, and the propagation relationship of the prior field value is established between adjacent spatial units. The prior field values ​​of adjacent spatial units are synchronously adjusted according to the time consistency and topological reachability rules to form a prior field grid containing spatial units and their adjacency relationships.

[0025] Specifically, identifying a set of spatial units that satisfy a preset connectivity rule within the prior field grid, and encapsulating this set of spatial units as a suspected disaster source area, includes:

[0026] Based on the spatial cells and their topological adjacency relationships in the prior field grid, preset connectivity rules are generated, and rule identifiers are assigned to each connectivity rule.

[0027] Within each time window, spatial units that meet the preset prior field value conditions are selected from the prior field grid as seed units, and the seed units are associated with their corresponding allocation rule identifiers to form a seed set for connectivity expansion.

[0028] Starting with the seed set, perform connectivity expansion in the prior field grid according to the preset connectivity rules corresponding to the rule identifier to obtain a set of spatial cells;

[0029] The spatial unit set is encapsulated as a suspected disaster source area.

[0030] Specifically, for the suspected disaster source area, a disaster source tracing relationship model is constructed, and the disaster source tracing relationship model is inverted to obtain the disaster source area location, triggering factors, and triggering time, including:

[0031] For each suspected disaster source area, its corresponding spatial boundary and time range are determined, and crack fingerprint sequences and event fragments falling within the spatial boundary and time range are extracted from the mining disaster analysis dataset to form a source area evidence set;

[0032] In the source region evidence set, a disaster outcome anchor point is determined, and a retrospective time window is generated based on the disaster outcome anchor point. The retrospective time window is then divided into multiple retrospective levels in chronological order.

[0033] Within each retrospective level, the crack fingerprint sequence and event fragments are mapped to event nodes, and directed relationships are set between all event nodes to construct a disaster source tracing relationship model;

[0034] Inversion inference constraints are generated for the disaster source tracing relationship model. Under the condition of satisfying the inversion inference constraints, candidate source tracing paths are enumerated backward from the disaster result anchor point, and a corresponding path identifier and path evidence list are generated for each candidate source tracing path.

[0035] Based on the aforementioned path evidence list, candidate source tracing paths are adjudicated to determine the target source tracing path, and the source area location, triggering factors, and triggering time of the disaster are output based on the target source tracing path.

[0036] Specifically, within each backtracking level, the crack fingerprint sequence and event fragments are mapped to event nodes, and directed relationships are established between all event nodes to construct a disaster source tracing relationship model, including:

[0037] Within each backtracking level, the gap fingerprint sequence falling into that backtracking level and various event fragments are merged according to the object identifier;

[0038] The fracture fingerprint sequence is segmented into change segments and mapped to fracture event nodes. The drilling and blasting construction parameter event segments and geological and mining progress event segments are mapped to disturbance event nodes. The microseismic or acoustic emission event segments are mapped to response event nodes. The disaster result anchor points are mapped to result event nodes, forming a cross-modal consistent event node set.

[0039] Based on the set of event nodes, a candidate set of directed relationships is generated according to preset rules;

[0040] The candidate set of directed relations is assembled into directed relations between event nodes, and loop relations, directional conflict relations and duplicate relations are pruned. Based on the event node set and its directed relations obtained after pruning, a disaster source tracing relation model is constructed.

[0041] Specifically, inversion inference constraints are generated for the disaster source tracing model. Under the condition that the inversion inference constraints are satisfied, candidate source tracing paths are enumerated backward from the disaster result anchor point, and a corresponding path identifier and path evidence list are generated for each candidate source tracing path, including:

[0042] Based on the source type identifier and backtracking level identifier of the event node, a set of constraint templates for inversion inference constraints is set, and the set of constraint templates is instantiated on the disaster source tracing relationship model to obtain the inversion inference constraint set;

[0043] Starting from the aforementioned disaster outcome anchor point, a backtracking search boundary is set;

[0044] Within the backtracking search boundary, a reverse traversal is performed from the disaster result anchor point along the directed relationship in the disaster source tracing relationship model. During each traversal expansion, the candidate expansion edge and candidate expansion node are verified by the inversion inference constraint set. The traversal branches that pass the verification are used to form candidate source tracing paths.

[0045] For each candidate tracing path, a path identifier is generated, and based on each candidate tracing path, the index of the event node reference fragments it passes through and the corresponding constraint verification records are summarized to generate a path evidence list.

[0046] Specifically, candidate source tracing paths are adjudicated based on the aforementioned path evidence list to determine the target source tracing path. Based on the target source tracing path, the source area location, triggering factors, and triggering time of the disaster are output, including:

[0047] The path evidence lists for each candidate tracing path are merged according to evidence type, and the corresponding event node identifier and reference fragment index are written for each type of evidence after merging to form an evidence vector;

[0048] A set of adjudication rules is generated based on the evidence vector description. Candidate tracing paths that do not meet any adjudication rule are filtered out according to the set of adjudication rules to obtain a set of retained candidates.

[0049] The reserved candidate set is adjudicated according to a preset hierarchical priority, the target tracing path is determined from the reserved candidate set, and an adjudication identifier is written for the target tracing path;

[0050] Based on the target tracing path, the source elements are extracted, and the source area location, triggering factors, and triggering time of the disaster are output.

[0051] The big data-based hard rock mining disaster source tracing and analysis system is used to implement the big data-based hard rock mining disaster source tracing and analysis method, including: a data acquisition module, a suspected source area determination module, and a source tracing and analysis module;

[0052] The data acquisition module is used to acquire multi-source data of the hard rock mining operation area and construct a mining disaster analysis dataset based on the multi-source data. The multi-source data includes underground mine image or video data, microseismic or acoustic emission data, drilling and blasting construction parameter data, and geological and mining progress data.

[0053] The suspected source area determination module is used to construct a fracture disaster-inducing prior field reflecting the evolution state of surrounding rock fracture based on the mining disaster analysis dataset, and to determine the suspected source area of ​​the disaster.

[0054] The source tracing analysis module is used to construct a disaster source tracing relationship model for the suspected disaster source area, and to perform inversion inference on the disaster source tracing relationship model to obtain the disaster source area location, triggering factors and triggering time.

[0055] Compared with the prior art, the beneficial effects of the present invention are:

[0056] This invention proposes a method and system for tracing the source of disasters in hard rock mining based on big data. It unifies the spatiotemporal modeling of multi-source heterogeneous big data generated during hard rock mining operations, using image fracture fingerprints as the core and integrating information such as microseismic activity, construction events, and geological progress to construct a fracture-induced disaster prior field and further form a disaster source relationship model. Under the conditions of satisfying temporal causality and spatial topological constraints, it inversely infers the disaster formation process, thereby achieving source analysis of the disaster source area location, triggering factors, and triggering time. Compared with existing technologies that rely mainly on single monitoring data or empirical criteria, this method can reduce ambiguities in source tracing caused by inconsistent time and unclear spatial references in multi-source data. It provides clear data basis and operational path for determining the disaster source area, triggering factors, and triggering time, and is applicable to hard rock mining scenarios where the causes of disasters are complex and the boundaries of responsibility are difficult to define under deep hard rock mining conditions. Attached Figure Description

[0057] Figure 1 Flowchart of the big data-based hard rock mining disaster source tracing analysis method provided by this invention;

[0058] Figure 2 The architecture diagram of the hard rock mining disaster source tracing and analysis system based on big data provided by this invention. Detailed Implementation

[0059] The present application will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0060] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0061] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. In addition, the terms "first," "second," and "third" used in this application do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0062] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0063] Example 1

[0064] Please see Figure 1 The present invention provides an embodiment of a method comprising the following specific steps:

[0065] Step S1: Obtain multi-source data of the hard rock mining operation area, and construct a mining disaster analysis dataset based on the multi-source data. The multi-source data includes underground mine image or video data, microseismic or acoustic emission data, drilling and blasting construction parameter data, and geological and mining progress data.

[0066] The specific steps of step S1 are as follows:

[0067] Step S101: Determine the set of analysis objects for the hard rock mining operation area and assign a unified object identifier to each unit. The set of analysis objects includes stope space units, roadway topology units, operation event units, and monitoring point units.

[0068] In this embodiment, addressing the challenges of complex spatial structures, frequent operational activities, and diverse monitoring data sources in hard rock mining areas, the following approach is first adopted: Based on the actual mining project, elements within the operational area that can independently participate in the disaster formation process are abstracted and categorized into four sets of analytical objects: stope spatial units, roadway topology units, operational event units, and monitoring point units. Stope spatial units carry information on surrounding rock conditions and fracture evolution; roadway topology units characterize the reachability relationships between spatial units; operational event units describe construction disturbance behaviors with clearly defined occurrence times; and monitoring point units bind data collection sources. Furthermore, by assigning a unified and unique object identifier to each analytical object, image data, event record data, and monitoring data originally scattered across different data sources are mapped to the same object identifier system. This allows subsequent analysis of time alignment, spatial correlation, and causal relationships of multi-source data to be conducted around the object identifier.

[0069] Step S102: Obtain multi-source data based on the object identifier, encapsulate the multi-source data into event fragments according to a preset sampling window, and write the corresponding object identifier and acquisition time identifier for each event fragment. The multi-source data includes underground mine image or video data, microseismic or acoustic emission data, drilling and blasting construction parameter data, and geological and mining progress data. The event fragments include underground mine image or video event fragments, microseismic or acoustic emission event fragments, drilling and blasting construction parameter event fragments, and geological and mining progress event fragments.

[0070] In this embodiment, based on the object identification system established in step S101, multi-source data corresponding to each object identifier is retrieved from different data sources, and continuous and discrete data are processed into events according to a unified sampling logic. Specifically, for underground mine image or video data, a fixed time span or work stage boundary is used as a sampling window to segment continuous images and form image or video event segments with clear start and end times; for microseismic or acoustic emission data, a signal trigger interval or preset time interval is used as a sampling window to encapsulate continuous waveform data into microseismic or acoustic emission event segments; for drilling and blasting construction parameter data and geological and mining progress data, they are organized into corresponding event segments according to work records or progress nodes; subsequently, each event segment is written with its associated object identifier and acquisition time identifier, so that the event segment has a clear orientation on both the time axis and the object level.

[0071] Step S103: Select one type of data from the multi-source data as the reference time scale source; perform time scale drift correction and time segment resampling on the remaining multi-source data so that each event segment forms a time series under the same reference time scale.

[0072] In this embodiment, to address the issue of clock inconsistencies in acquisition equipment, recording mechanisms, and storage methods among multi-source data, a data type with strong temporal continuity and stable acquisition is first selected from the multi-source data as a reference timescale source. Specifically, the time axis corresponding to underground mine images or video data can be selected as a unified reference. Subsequently, using the reference timescale source as a reference, timescale correction processing is performed on the event segments corresponding to the remaining multi-source data. This correction process corrects the overall offset and local drift segment by segment by comparing the recording time of each event segment with its relative order in the mining operation process. After completing the timescale correction, time segment resampling is performed on event segments whose time resolution is inconsistent with the reference timescale, so that all types of event segments can be mapped to a unified time scale. It should be noted that through the above processing, event segments that originally came from different sources and had different time recording precisions can form a time sequence with a clear chronological relationship under the same reference timescale.

[0073] Step S104: Extract the fracture skeleton, block boundary discontinuity segments, and texture abrupt change segments from the images or videos in the time series, and combine them into fracture fingerprint entries for the mining space unit.

[0074] In this embodiment, the surface morphology information of the surrounding rock contained in the time series of images or videos in the mine is analyzed frame by frame. First, the light and dark variation structure distributed linearly or branchedly along the rock surface is identified from the image and abstracted as a fracture skeleton. On this basis, the continuity analysis of the area around the fracture skeleton is performed to determine the location where the boundary between the surrounding rock blocks is misaligned, interrupted or abrupt, thereby extracting discontinuous segments of the block boundary. At the same time, the gray distribution and texture arrangement of local areas in the image are compared and analyzed to mark the areas that change significantly in adjacent time frames, forming texture abrupt segment. Subsequently, the fracture skeleton, discontinuous segments of block boundary and texture abrupt segment within the same mining spatial unit and having a continuous relationship in time are combined, and the corresponding spatial unit identifier and time identifier are written into the combination to form a fracture fingerprint entry.

[0075] It should be noted that the extraction of fracture fingerprint entries is not limited to a specific image processing algorithm. Its purpose is to extract structural descriptive information that can reflect the morphology and evolution characteristics of fractures in the surrounding rock from underground mine images or video data. In specific implementation, well-known image processing and computer vision methods in the field can be used to analyze images or video frames.

[0076] Step S105: Associate and catalog the fracture fingerprint entries and event fragments according to the object identifier to form a mining disaster analysis dataset.

[0077] In this embodiment, after extracting fracture fingerprint entries and unifying the time of multi-source event fragments, the fracture fingerprint entries and various event fragments are cataloged and integrated using object identifiers as the main association. Specifically, for each fracture fingerprint entry, based on its accompanying stope spatial unit identifier and time identifier, the event fragment set is searched for mine underground image or video event fragments, microseismic or acoustic emission event fragments, drilling and blasting construction parameter event fragments, and geological and mining progress event fragments with the same object identifier and overlapping or adjacent time ranges. These are recorded as related items in the same cataloging unit. For event fragments that cannot form a one-to-one correspondence within the same time window, their spatial affiliation is maintained through object identifiers, and their time pointers are stored independently. It should be noted that through the above-mentioned association cataloging method centered on object identifiers, the image features of surrounding rock fractures and multi-source operations and monitoring events within the same or adjacent time periods can be organized into unified data entries. This allows information from different sources during the mining process to form a complete record under the same spatial object and time context, thereby constructing a mining disaster analysis dataset for subsequent disaster source tracing analysis.

[0078] Step S2: Based on the mining disaster analysis dataset, construct a fracture-induced disaster prior field that reflects the evolution of surrounding rock fracture and determine the suspected disaster source area.

[0079] The specific steps of step S2 are as follows:

[0080] Step S201: Slice the mining disaster analysis dataset according to a preset time window, and divide the hard rock mining area into multiple spatial units. Establish a corresponding feature container for each spatial unit within each time window. The feature container is used to carry the fracture fingerprint entries and event fragments associated with the spatial unit.

[0081] In this embodiment, to ensure that the mining disaster analysis dataset can simultaneously reflect the temporal evolution and spatial distribution characteristics of the surrounding rock state, the mining disaster analysis dataset is first sliced ​​according to a preset time window. The time window can be set according to the mining operation rhythm or the monitoring data recording frequency, so that the data within the same time window has relatively consistent temporal semantics. At the same time, the hard rock stope is divided into multiple spatial units according to the roadway layout, stope boundary and spatial continuity to ensure that each spatial unit can cover a relatively stable surrounding rock area. Subsequently, a corresponding feature container is established for each spatial unit within each time window, and fracture fingerprint entries and various event fragments with the same spatial unit identifier and time identifier falling within the time window range are written into the feature container.

[0082] Step S202: Arrange the crack fingerprint entries in each spatial unit in chronological order to generate a crack fingerprint sequence.

[0083] In this embodiment, for the feature container established in step S201, the fracture fingerprint entries contained in each spatial unit within different time windows are organized. Specifically, based on the time identifier carried by the fracture fingerprint entries, the fracture fingerprint entries within the same spatial unit are sorted to form a continuous arrangement on the time axis; for fracture fingerprint entries with overlapping acquisition intervals or time identifiers, they are adjusted according to the chronological order of their source images or video frames to maintain the consistency of the time series; through the above processing, the originally discretely stored fracture fingerprint entries are transformed into a fracture fingerprint sequence reflecting the change of the surrounding rock surface state within the spatial unit over time.

[0084] Step S203: Using the gap fingerprint sequence as the anchoring object, align event segments associated with the same spatial unit to the gap fingerprint sequence according to their temporal proximity. When event segments overlap or have inconsistent indications within the same time window, mark and resolve the overlap or inconsistency according to a preset priority rule to obtain a resolved fusion sequence.

[0085] In this embodiment, to achieve a unified representation of multi-source events within the same spatial unit, the fracture fingerprint sequence is used as a time anchoring reference to align various event fragments associated with the spatial unit. Specifically, according to the time identifier carried by the event fragments, microseismic or acoustic emission event fragments, drilling and blasting construction parameter event fragments, and geological and mining progress event fragments are mapped to the corresponding time positions in the fracture fingerprint sequence based on their occurrence time, so that each event fragment forms a proximity or overlap relationship with the fracture fingerprint entry on the time axis. When multiple event fragments overlap in time or their indications are inconsistent within the same time window, the relevant event fragments are marked according to a preset priority rule, and the event fragments with higher priority are retained as the main record, while the remaining event fragments are treated as auxiliary records or temporarily masked. Through the above alignment, marking, and resolution processes, a resolved fusion sequence is formed.

[0086] Step S204: Based on the resolved fusion sequence, assign a corresponding prior field value to each spatial unit, and introduce topological adjacency relationship between adjacent spatial units to form a prior field grid, wherein the prior field value is jointly determined by the co-occurrence relationship between the topological change segment of the crack fingerprint sequence and the microseismic or acoustic emission event segment after time alignment.

[0087] The specific steps of step S204 are as follows:

[0088] Step S2041: For each spatial unit, the resolved fusion sequence is segmented according to temporal continuity, and a corresponding surrounding rock state identifier is enumerated for each segment. The surrounding rock state identifier is determined by the change segment of the fracture fingerprint sequence and the event segment aligned with it.

[0089] In this embodiment, for the resolved fusion sequence formed in each spatial unit, the continuity of the fusion sequence is first checked according to the time identifier, and the fusion sequence is divided into multiple continuous segments at the positions where there is obvious discontinuity in time, the event type changes, or the fracture fingerprint characteristics change in stages. Then, the fracture fingerprint sequence contained in each segment is identified for change segments, and feature segments reflecting the adjustment of fracture morphology, distribution or structure are extracted. Combined with microseismic or acoustic emission event segments, drilling and blasting construction parameter event segments and geological and mining progress event segments aligned in the same time period, the surrounding rock state of the segment is enumerated and identified.

[0090] Step S2042: Within the same spatial unit, the segments are sorted according to the chronological order of the surrounding rock state identifiers, and a state trajectory reflecting the state transition order within the spatial unit is generated. The state trajectory serves as the internal evolution description benchmark for the spatial unit.

[0091] In this embodiment, after segmenting the fusion sequence and enumerating the surrounding rock state identifiers, for each time segment within the same spatial unit, the surrounding rock state identifiers are sorted according to their corresponding time identifiers, so that each segment forms a clear sequential relationship on the time axis. Subsequently, the surrounding rock state identifiers arranged in chronological order are connected sequentially to generate a state trajectory describing the change process of the surrounding rock state from early to late within the spatial unit. It should be noted that this state trajectory is not a simple time list, but rather uses the surrounding rock state identifiers as nodes to reflect the staged state migration sequence under the influence of fracture evolution and related events within the same spatial unit, thereby forming a continuous descriptive benchmark for the evolution of the surrounding rock within the spatial unit.

[0092] Step S2043: Map the state trajectory to the corresponding prior field value and bind the prior field value to the corresponding spatial unit. Different state trajectories correspond to different prior field value intervals, and multiple prior field value records are allowed to exist in the same spatial unit in different time windows.

[0093] In this embodiment, after obtaining the state trajectory of a spatial unit, the state trajectory is used as a mapping basis and subjected to interval processing. That is, according to the arrangement order, duration, and change characteristics between adjacent states of the surrounding rock state identifiers in the state trajectory, a corresponding a priori field value interval is assigned to the state trajectory. Subsequently, the a priori field value interval is bound to the spatial unit that generated the state trajectory and recorded together with its corresponding time window identifier. The state trajectories formed by different spatial units or the same spatial unit in different time windows have differences in structure and order, so they can be mapped to different a priori field value intervals, thereby allowing the same spatial unit to have multiple a priori field value records in different time windows. Through this mapping and binding method, the surrounding rock evolution information that originally existed in the form of discrete state trajectories can be transformed into an a priori field representation that can be compared and propagated in the spatial and temporal dimensions.

[0094] It should be noted that the prior field value is not used to represent a specific physical quantity. Its role is to relatively characterize and compare the evolution state of surrounding rock fracture in different spatial units within different time windows. The determination of the prior field value is based on the structural characteristics of the state trajectory within the spatial unit, rather than a single instantaneous monitoring data. In the specific implementation process, the state trajectory is classified or intervalized according to the order of change, duration, and magnitude of change between adjacent states of the surrounding rock state identifiers in the state trajectory, and different levels or intervals are mapped to the corresponding prior field values. For example, a state trajectory in which the surrounding rock state is stable for a long time and the fracture changes slowly can be mapped to a prior field value in a lower interval; a state trajectory in which the surrounding rock state changes frequently or the fracture structure evolves rapidly can be mapped to a prior field value in a higher interval.

[0095] Step S2044: Determine the adjacency pairs between each spatial unit, establish the propagation association of prior field values ​​between adjacent spatial units, and synchronously adjust the prior field values ​​of adjacent spatial units according to the time consistency and topological reachability rules to form a prior field grid containing spatial units and their adjacency relationships.

[0096] In this embodiment, after binding the prior field values ​​of each spatial unit, the existence of direct adjacency or reachability between spatial units is determined based on the spatial division results of the mining area and the roadway topology, thereby identifying adjacency pairs between spatial units. Subsequently, a propagation association of prior field values ​​is established between each adjacency pair, enabling adjacent spatial units to form associative prior field value records in the time dimension. Specifically, for adjacent spatial units with prior field value records within the same or adjacent time windows, the correspondence of their prior field values ​​is calibrated according to the time consistency rule, and topological reachability rules such as roadway connection direction and layer relationship are combined to restrict or disconnect associations that do not meet the propagation conditions. Through the above adjacency determination, association establishment, and synchronous adjustment process, a prior field grid composed of multiple spatial units and their adjacency relationships is formed.

[0097] Step S205: Identify a set of spatial units that satisfy a preset connectivity rule in the prior field grid, and encapsulate the set of spatial units as a suspected disaster source area.

[0098] The specific steps of step S205 are as follows:

[0099] Step S2051: Based on the spatial units and their topological adjacency relationships in the prior field grid, generate preset connectivity rules and assign rule identifiers to each connectivity rule. The preset connectivity rules include adjacency connectivity rules along the topological path of the alley, projection connectivity rules across layers, and continuation connectivity rules across time windows.

[0100] In this embodiment, after obtaining the prior field grid, the partitioning results of spatial units in the prior field grid and their topological adjacency relationships are used as the basis to summarize and organize the connection methods that may form continuous influence relationships between spatial units, thereby generating preset connectivity rules. Specifically, for spatial units directly connected in the roadway topology, adjacency connectivity rules along the roadway topology path are defined to describe the continuous association of the surrounding rock state in the same roadway path direction; for spatial units located at different elevations or different layers but with a projection overlap relationship in the planar position, projection connectivity rules across layers are defined to describe the corresponding association of the surrounding rock state in the vertical direction; for prior field value records formed by the same spatial unit in adjacent time windows, continuation connectivity rules across time windows are defined to describe the continuity of the surrounding rock state in the time dimension; subsequently, rule identifiers are assigned to the above-mentioned connectivity rules respectively, so as to distinguish and call different rules in the subsequent connectivity expansion process.

[0101] Step S2052: Within each time window, select spatial units that meet the preset prior field value conditions from the prior field grid as seed units, and associate the seed units with their corresponding allocation rule identifiers to form a seed set for connectivity expansion.

[0102] In this embodiment, for each preset time window, the spatial units in the prior field grid that have prior field value records within that time window are traversed, and spatial units that meet the conditions are selected as seed units according to the preset prior field value conditions. The prior field value conditions are used to limit whether the state characteristics of the spatial unit within the corresponding time window have the basis for further connectivity analysis. Subsequently, each selected seed unit is associated with its spatial location and applicable connectivity rule identifier, so that the seed unit can clearly identify the connectivity mode it can adopt in the subsequent connectivity expansion process. Through the above filtering and association process, a seed set containing seed units and their corresponding rule identifiers is formed.

[0103] Step S2053: Starting from the seed set, perform connectivity expansion in the prior field grid according to the preset connectivity rules corresponding to the rule identifier to obtain a set of spatial cells. The connectivity expansion is limited to the spatial cells that satisfy the topological reachability constraint and the time consistency constraint, and a time bridging mark is established for cross-time window connections that occur during the expansion process.

[0104] In this embodiment, after obtaining the seed set, each seed unit is used as the starting point for connectivity expansion. Based on its associated rule identifier, adjacent spatial units that satisfy the corresponding preset connectivity rules are progressively searched in the prior field grid. During the expansion process, it is first determined whether the candidate spatial units satisfy the reachability conditions in the roadway topology, and at the same time, it is verified whether their prior field value records maintain temporal consistency with the current time window or adjacent time windows. Only spatial units that simultaneously satisfy the above constraints are included in the expansion scope. For connectivity cases involving cross-time windows during the expansion process, after confirming the existence of a continuity relationship within adjacent time windows, a time bridging marker is established for this connectivity relationship to identify the continuity of the spatial unit set in the time dimension. Through the above-mentioned constrained expansion process, a spatial unit set containing multiple spatial units can be progressively formed from the seed set.

[0105] Step S2054: Encapsulate the spatial unit set into a suspected disaster source region, and write a source region identifier, a list of included spatial units, a time bridging marker, and a source region boundary description for the suspected disaster source region. The source region boundary description is determined by the outer contour unit of the spatial unit set in the prior field grid and its topological adjacency disconnection position.

[0106] In this embodiment, after completing the connectivity expansion of the spatial unit set, the entire spatial unit set is regarded as a candidate analysis object and encapsulated to form a suspected disaster source region. Specifically, a unique source region identifier is assigned to the spatial unit set, and all spatial unit identifiers constituting the set and their corresponding time bridging markers are recorded together to characterize the continuity of the source region in the spatial and temporal dimensions. Subsequently, the distribution of the spatial unit set in the prior field grid is traversed, and spatial units located at the outer edge of the set and topologically disconnected from non-set spatial units are identified. These are used as outer contour units, and a source region boundary description is generated by combining the corresponding topological adjacency disconnection position. It should be noted that, through the above encapsulation method, the suspected disaster source region not only includes its internal spatial unit composition but also has clear boundary references and time association markers, thereby enabling the source region to serve as an independent analysis unit for subsequent disaster source tracing relationship modeling and inversion inference.

[0107] Step S3: For the suspected disaster source area, construct a disaster source relationship model, and perform inversion inference on the disaster source relationship model to obtain the disaster source area location, triggering factors and triggering time.

[0108] The specific steps of step S3 are as follows:

[0109] Step S301: For each suspected disaster source area, determine its corresponding spatial boundary and time range, and extract the crack fingerprint sequence and event fragments that fall within the spatial boundary and time range from the mining disaster analysis dataset to form a source area evidence set.

[0110] In this embodiment, for each suspected disaster source area that has been encapsulated, the spatial boundary range of the suspected disaster source area is first determined based on the list of spatial units corresponding to its source area identifier and the source area boundary description. At the same time, the time range covering the formation process of the source area is determined by combining its time bridging marker. Subsequently, the spatial boundary and time range are used as filtering conditions to search the catalog entries established in the mining disaster analysis dataset, extract the fracture fingerprint sequences and various event fragments whose spatial affiliation falls within the spatial boundary and whose time marker is within the time range, and aggregate them to form a source area evidence set.

[0111] Step S302: Determine the disaster result anchor point in the source area evidence set, generate a retrospective time window based on the disaster result anchor point, and divide the retrospective time window into multiple retrospective levels in chronological order. The disaster result anchor point is determined by at least one of the following: disaster record event fragment, production stoppage / alarm event fragment, or surrounding rock state change event fragment.

[0112] In this embodiment, after obtaining the source area evidence set corresponding to the suspected disaster source area, the event fragments in the evidence set are examined one by one. Event fragments that can directly indicate the disaster occurrence result are prioritized as disaster result anchor points. The event fragments specifically include at least one of the following: disaster record event fragments, production stoppage or alarm event fragments corresponding to the disaster occurrence time, and surrounding rock state abrupt change event fragments reflecting a significant change in the surrounding rock state in a short period of time. After determining the disaster result anchor point, the time marker corresponding to the anchor point is used as a reference to backtrack and generate a backtracking time window covering a period of time before the disaster occurred. According to the distance between the time and the disaster result anchor point, the backtracking time window is divided into multiple backtracking levels with a chronological order.

[0113] Step S303: Within each backtracking level, map the crack fingerprint sequence and event fragments to event nodes. Based on the chronological order, the association of the same object identifier, and the topological reachability of the tunnel, set directed relationships between all event nodes to construct a disaster source tracing relationship model.

[0114] The specific steps of step S303 are as follows:

[0115] Step S3031: Within each backtracking level, merge the gap fingerprint sequence and various event fragments falling into that backtracking level according to the object identifier.

[0116] In this embodiment, after the hierarchical division of the retrospective time window is completed, for each retrospective level, fracture fingerprint sequences and various event fragments whose time markers fall within the range of that retrospective level are screened, and then merged according to their object identifiers. Specifically, fracture fingerprint sequences with the same stope space unit identifier or operation event unit identifier are grouped with microseismic or acoustic emission event fragments, drilling and blasting construction parameter event fragments, and geological and mining progress event fragments into the same merging unit to ensure that relevant information of the same analysis object within that retrospective level is recorded centrally.

[0117] Step S3032: Segment the fracture fingerprint sequence into fracture event nodes according to the change segments, map the drilling and blasting construction parameter event segments and the geological and mining progress event segments into disturbance event nodes, map the microseismic or acoustic emission event segments into response event nodes, and map the disaster result anchor points into result event nodes. Write the source type identifier and reference segment index for each event node to form a cross-modal consistent event node set.

[0118] In this embodiment, after data merging within the same retrospective level is completed, a unified event node mapping process is performed on different types of data in the merged unit. Specifically, firstly, the fracture fingerprint sequence is analyzed segment by segment. Based on the location where the fracture morphology or distribution characteristics are adjusted, the fracture fingerprint sequence is divided into multiple change segments, and each change segment is mapped as a fracture event node to characterize the staged changes of the surrounding rock fractures within the retrospective level. Subsequently, drilling and blasting construction parameter event segments and geological and mining progress event segments are mapped as disturbance event nodes to describe operational behaviors or advancement nodes that may affect the state of the surrounding rock. At the same time, microseismic or acoustic emission event segments are mapped as response event nodes to characterize the response of the surrounding rock during disturbance or evolution. Event segments corresponding to disaster result anchor points are mapped as result event nodes. For each of the above types of event nodes, their source type identifier and reference segment index are written, so that each event node retains its pointing relationship to the original data.

[0119] Step S3033: Based on the event node set, generate a directed relation candidate set according to preset rules. The preset rules include at least: time sequence rules, adjacent rules within the same spatial unit, lane topology reachability rules, and cross-level inheritance rules. The cross-level inheritance rules are used to establish inheritance markers between event nodes in earlier backtracking levels and event nodes in later backtracking levels, forming a cross-level chain connection.

[0120] In this embodiment, after obtaining a cross-modal consistent set of event nodes, the set of event nodes is traversed to generate a candidate set of directed relationships for constructing tracing relationships. Specifically, firstly, based on the time identifier carried by the event nodes, directed relationship candidates under the time sequence rule are established for event nodes whose occurrence times are related; then, for event nodes with the same spatial unit identifier, directed relationship candidates under the adjacency rule are generated according to their temporal proximity relationship within the spatial unit; simultaneously, combined with the tunnel topology, directed relationship candidates under the tunnel topology reachability rule are generated for event nodes located in different spatial units but with a topologically reachable path; furthermore, between different backtracking levels, directed relationship candidates under the cross-level inheritance rule are established for event nodes in earlier backtracking levels and event nodes in later backtracking levels that have a temporal continuity relationship or an object identifier inheritance relationship, and an inheritance tag is written for this relationship.

[0121] Step S3034: Assemble the candidate set of directed relations into directed relations between event nodes, and prune the loop relations, directional conflict relations and duplicate relations that appear. Based on the event node set and its directed relations obtained after pruning, construct the disaster source tracing relation model.

[0122] In this embodiment, after generating a set of directed relation candidates, the directed relation candidates are assembled one by one into directed relations between event nodes according to their corresponding event nodes to form an initial relation structure. Subsequently, the initial relation structure is checked to identify loop relations, directional conflict relations between the same pair of event nodes, and duplicate relations pointing to the same event node. The above relations are then deleted or merged according to a preset pruning principle.

[0123] The disaster source tracing relationship model includes at least the following modules: event node construction module, relationship generation module, constraint modeling module, and source inference module. These modules are connected sequentially according to a preset data flow and control logic order to jointly complete the construction and inversion analysis of disaster source tracing relationships.

[0124] The event node construction module maps multi-source data involved in disaster source tracing analysis into a unified set of event nodes. Specifically, this module receives fracture fingerprint sequences and various event fragments from the mining disaster analysis dataset as input, and performs node processing based on the data source type and time attribute. In this embodiment, the fracture fingerprint sequence is segmented into fracture event nodes after being divided into change segments, drilling and blasting construction parameter event fragments and geological and mining progress event fragments are mapped as disturbance event nodes, microseismic or acoustic emission event fragments are mapped as response event nodes, and disaster record event fragments are mapped as result event nodes. Each event node contains at least a node type identifier, an object identifier, a time identifier, and a reference index to the original data. The relationship generation module is used to construct directed relationships between the event node sets. This module uses the time identifier and spatial affiliation of the event nodes. The model uses the roadway topology as the basis for relationship generation. Under the condition of satisfying the preset generation rules, it establishes directional associations between event nodes. The constraint modeling module is used to generate inversion inference constraints for the disaster source tracing relationship model to limit the search range and legality of the source tracing path. The inversion inference constraints include at least time irreversibility constraints, spatial topology reachability constraints, and disturbance-response matching constraints. Among them, time irreversibility constraints are used to limit the directed relationship between event nodes to satisfy the time sequence logic; spatial topology reachability constraints are used to limit the spatial units to which the event nodes belong to should satisfy the roadway topology or spatial adjacency relationship; disturbance-response matching constraints are used to limit the correspondence between disturbance event nodes and response event nodes in terms of time window and object identifier. The source tracing inference module is used to perform path inversion analysis on the disaster source tracing relationship model under the premise of satisfying the inversion inference constraints. This module takes the disaster result event node as the starting node, traverses the event node set in reverse along the directed relationship, and performs verification and pruning of candidate expansion paths according to the inversion inference constraints during the traversal process, thereby generating a set of candidate source tracing paths.

[0125] The input data for the disaster source tracing model includes: fracture fingerprint sequences and their variation segment identifiers; disturbance event data such as drilling and blasting construction parameter event segments, geological and mining progress event segments; response event data such as microseismic or acoustic emission event segments; disaster record or alarm event segments; time identifiers, spatial unit identifiers, and roadway topology data corresponding to event nodes; the output results of the disaster source tracing model include at least: one or more candidate source tracing paths; path identifiers and path evidence lists corresponding to each candidate source tracing path; the target source tracing path determined by adjudication; and the location of the disaster source area, triggering factors, and triggering time extracted based on the target source tracing path.

[0126] It should be noted that the disaster source tracing relationship model described in this invention does not rely on a black-box machine learning model that iteratively trains parameters. Its core lies in the structured modeling and constrained inversion inference process based on multi-source events. The model construction process mainly relies on the setting of event node generation rules, relationship generation rules, and inversion inference constraints, rather than training the model parameters through sample data. In some implementations, the rule thresholds or priority parameters can be empirically adjusted based on historical disaster data, but this adjustment process does not constitute the model training process and does not affect the basic structure and workflow of the disaster source tracing relationship model of this invention.

[0127] Step S304: Generate inversion inference constraints for the disaster source tracing relationship model. Under the condition of satisfying the inversion inference constraints, enumerate candidate source tracing paths backward from the disaster result anchor point, and generate a corresponding path identifier and path evidence list for each candidate source tracing path. The inversion inference constraints include at least: time irreversibility constraints, topology propagation constraints, and disturbance-response matching constraints.

[0128] The specific steps of step S304 are as follows:

[0129] Step S3041: Based on the source type identifier and backtracking level identifier of the event node, set a constraint template set for inversion inference constraints, and instantiate the disaster source tracing relationship model with the constraint template set to obtain the inversion inference constraint set. The constraint template set includes at least a time monotonic constraint template, a topological reachability constraint template, a disturbance-response pairing constraint template, and a cross-level inheritance constraint template.

[0130] In this embodiment, to ensure the disaster source tracing model possesses executable inversion inference conditions, the basic constraints that need to be satisfied during the source tracing process are first summarized and organized based on the source type identifier and backtracking level identifier carried by the event nodes, forming a set of constraint templates for inversion inference constraints. Specifically, a time monotonic constraint template is set for the arrangement of event nodes on the time axis to limit the order of event nodes in directed relationships; a topological reachability constraint template is set for the topological relationship between spatial units where event nodes are located to limit the spatial propagation path between event nodes; a perturbation-response pairing constraint template is set for the correspondence between perturbation event nodes and response event nodes to limit the matching method between them in terms of time and object identifiers; and a cross-level inheritance constraint template is set for the association of event nodes between different backtracking levels to limit the scope of use of inheritance markers. Subsequently, the set of constraint templates is applied to the disaster source tracing model, and the event nodes and their directed relationships in the model are instantiated one by one to generate an inversion inference constraint set corresponding to specific event nodes and relationships.

[0131] Step S3042: Starting from the disaster result anchor point, set a backtracking search boundary. The backtracking search boundary includes at least the maximum backtracking level, the range of spatial units that can be crossed, and the set of event node types that can be referenced. Assign a boundary identifier to the backtracking search boundary to constrain the subsequent path enumeration process.

[0132] In this embodiment, after generating the inversion inference constraint set, the event node corresponding to the disaster result anchor point is used as the starting position for the source tracing search, limiting the scope of subsequent backtracking analysis. Specifically, based on the time span of the suspected disaster source area and the backtracking level division results, a maximum backtracking level is set to avoid boundless expansion. Simultaneously, considering the spatial boundary of the suspected disaster source area and the roadway topology, a range of spatial units that can be traversed is set to limit the activity area of ​​the source tracing path in the spatial dimension. Furthermore, based on the semantic level required for source tracing analysis, a set of event node types that can be referenced is set, ensuring that the source tracing path expands only among crack event nodes, disturbance event nodes, response event nodes, and result event nodes. Subsequently, the aforementioned backtracking search boundary parameters are combined and assigned a unified boundary identifier for constraint verification of each expansion operation during subsequent path enumeration.

[0133] Step S3043: Within the backtracking search boundary, perform a reverse traversal from the disaster result anchor point along the directed relationship in the disaster source tracing relationship model, and verify the candidate expansion edge and candidate expansion node with the inversion inference constraint set during each traversal expansion. Form candidate source tracing paths for the traversal branches that pass the verification, and prune the traversal branches that fail the verification.

[0134] In this embodiment, with the backtracking search boundary set set and the inversion inference constraint set instantiated, the result event node corresponding to the disaster result anchor point is used as the starting point, and a reverse traversal is performed along the directed relationship in the disaster source tracing relationship model. Specifically, each incoming edge of the current node and its pointed upstream node are generated as candidate expansion edges and candidate expansion nodes. Before each expansion, the candidate expansion edges and candidate expansion nodes are checked item by item according to the inversion inference constraint set. The checks include at least: whether the time identifier of the candidate upstream node satisfies the time monotonicity constraint, whether the spatial unit where the candidate upstream node is located satisfies the topological reachability constraint, whether the combination of node types connected by the candidate edge satisfies the disturbance-response pairing constraint and the cross-level inheritance constraint, and at the same time confirming that the candidate expansion does not exceed the level, spatial range and node type set defined in the backtracking search boundary. For expansions that pass the checks, they are included in the current traversal branch and the node sequence and edge sequence traversed by the branch are recorded, so that the branch gradually forms a candidate source tracing path. For expansions that fail the checks, the traversal branch is truncated at the corresponding position and the forward expansion is stopped, thereby completing the pruning process.

[0135] Step S3044: Generate a path identifier for each candidate tracing path, and based on each candidate tracing path, summarize the event node reference fragment index and corresponding constraint verification record it passes through to generate a path evidence list. The path identifier includes at least a start node identifier, an end node identifier, a cross-level inheritance marker, and a time fragment identifier.

[0136] In this embodiment, after enumerating the candidate source tracing paths, each candidate source tracing path is uniquely identified and its evidence is organized. Specifically, based on the earliest event node and the event node corresponding to the disaster result anchor point in the candidate source tracing path, their node identifiers are extracted as the start node identifier and the end node identifier, respectively. At the same time, it is summarized whether the path contains cross-backtracking hierarchy inheritance relationships, and the corresponding inheritance markers are written into the path identifier. In addition, the time segment identifiers corresponding to the time range covered by the candidate source tracing path are also written in, thereby generating a path identifier that can reflect the path structure and time range. Subsequently, the event nodes traversed by each candidate source tracing path are traversed one by one, and the crack fingerprint entries or event segment indexes referenced by each event node are summarized. At the same time, the inversion inference constraint verification results corresponding to each extension step in the path generation process are recorded, and the above information is uniformly organized into a path evidence list.

[0137] It should be noted that the inversion inference process of the disaster source tracing relationship model is based on event nodes and their directed relationships. Candidate paths are screened through step-by-step backtracking combined with inversion inference constraints, which is a constraint search process performed on a directed graph structure. This process does not rely on a specific mathematical model or optimization algorithm and can be implemented using graph traversal and constraint verification methods commonly used in this field. Through the above method, the enumeration and adjudication process of disaster source tracing paths has clear operational steps and is executable; those skilled in the art can complete the corresponding implementation based on the contents described in the specification.

[0138] Step S305: Based on the path evidence list, adjudicate the candidate source tracing paths, determine the target source tracing path, and output the source area location, triggering factors, and triggering time of the disaster based on the target source tracing path.

[0139] The specific steps of step S305 are as follows:

[0140] Step S3051: Merge the path evidence lists of each candidate tracing path according to evidence type, and write the corresponding event node identifier and reference fragment index for each type of evidence after merging to form an evidence vector. The evidence types include at least crack fingerprint evidence, disturbance event evidence, response event evidence, cross-level inheritance evidence, and constraint verification evidence.

[0141] In this embodiment, after forming the path evidence list corresponding to the candidate tracing paths, the evidence information in the path evidence list is structured and organized. Specifically, for each candidate tracing path, the items in the path evidence list are first classified according to the source of the evidence and semantic attributes. Fracture fingerprint items are merged into fracture fingerprint evidence, drilling and blasting construction parameter event fragments and geological and mining progress event fragments are merged into disturbance event evidence, microseismic or acoustic emission event fragments are merged into response event evidence, relationship records containing inheritance markers are merged into cross-level inheritance evidence, and constraint verification records generated during the inversion inference process are merged into constraint verification evidence. Subsequently, the corresponding event node identifier and reference fragment index are written under each type of evidence, so that each type of evidence maintains a clear pointing relationship with the event node and the original data. Through the above merging and identifier writing process, the path evidence originally recorded in chronological order can be transformed into an evidence vector organized by evidence type.

[0142] Step S3052: Generate a set of adjudication rules based on the evidence vector description, and filter out candidate tracing paths that do not satisfy any adjudication rule according to the set of adjudication rules to obtain a set of retained candidates. The set of adjudication rules includes at least time chain closure rules, spatial chain closure rules, node type coverage rules, and evidence mutual exclusion rules.

[0143] In this embodiment, after obtaining the evidence vectors corresponding to each candidate tracing path, a set of adjudication rules for path adjudication is generated based on the time, space, and event type information reflected in the evidence vectors. Specifically, a time chain closure rule is generated by checking whether the time markers of event nodes in the evidence vector can form a continuous time pointing relationship between the starting node and the result node; a spatial chain closure rule is generated by checking whether the spatial units to which the event nodes in the evidence vector belong can form a continuous spatial association under the roadway topology or projection relationship; a node type coverage rule is generated by checking whether the evidence vector simultaneously contains key node types such as crack events, disturbance events, and response events; and an evidence mutual exclusion rule is generated by comparing whether there are mutually exclusive time or object pointing relationships between different evidence entries in the evidence vector. Subsequently, each candidate tracing path is verified one by one according to the set of adjudication rules, and candidate tracing paths that do not meet any adjudication rule are eliminated to obtain a retained candidate set.

[0144] Step S3053: The reserved candidate set is adjudicated according to a preset hierarchical priority. The target tracing path is determined from the reserved candidate set, and an adjudication identifier is written for the target tracing path. The hierarchical priority includes at least: first comparing the continuity of cross-level inheritance tags, then comparing the coverage of disturbance-response node pairs, and finally comparing the adjacency sequence of the gap event nodes in the backtracking level.

[0145] In this embodiment, after obtaining the reserved candidate set, the candidate tracing paths are adjudicated layer by layer according to a preset hierarchical priority. Specifically, firstly, the continuity of cross-level inheritance markers in each candidate tracing path is compared, and paths that can form a continuous inheritance relationship between multiple backtracking levels are preferentially retained to ensure the temporal continuity of the tracing process. On this basis, the coverage of disturbance event nodes and response event nodes in the remaining candidate tracing paths is further compared, and paths that can form a complete disturbance-response correspondence during the backtracking process are preferentially retained. Subsequently, for the candidate tracing paths that are still not distinguished, the adjacency sequence of the fracture event nodes in each backtracking level is compared, and paths that show a continuous evolution relationship of fracture events in time and space are preferentially selected. Through the above hierarchical priority adjudication process, a unique or a small number of target tracing paths are determined from the reserved candidate set, and an adjudication identifier is written for the target tracing paths.

[0146] Step S3054: Extract source tracing elements based on the target source tracing path, and output the source area location, triggering factors, and triggering time of the disaster. The source area location is determined by the spatial unit identifier corresponding to the starting node of the target source tracing path. The triggering factors are determined by the source type identifier of the disturbance event node in the target source tracing path and its combination relationship. The triggering time is determined by the time segment identifier corresponding to the disturbance event node.

[0147] In this embodiment, after the target source tracing path has been determined and written with the adjudication identifier, the target source tracing path is processed for element extraction. Specifically, the starting event node is traced back along the target source tracing path from the result event node to locate the starting event node in the path, and its corresponding spatial unit identifier is read. This spatial unit identifier is used as the basis for determining the location of the disaster source area. Subsequently, the disturbance event nodes included in the target source tracing path are summarized, and the composition of the disaster triggering factors is determined based on the source type identifier of each disturbance event node and its sequential combination relationship in the path. At the same time, the time segment identifier corresponding to the disturbance event node is read and used as the time reference for the disaster triggering time. It should be noted that by directly extracting the source tracing elements from the node identifiers and time identifiers of the target source tracing path, the output results of the disaster source area location, triggering factors, and triggering time have clear data sources and path support, thereby completing the standardized output of the disaster source tracing analysis results.

[0148] The preset priority rules, preset connectivity rules, and set of adjudication rules in this invention are all used to constrain and filter multiple solutions that occur during multi-source data fusion, spatial unit connectivity analysis, and source path adjudication. Their specific forms can be configured according to engineering application scenarios and are not limited to a single implementation method.

[0149] Example 2

[0150] Please see Figure 2 Another embodiment of the present invention provides: a hard rock mining disaster source tracing and analysis system based on big data, comprising: a data acquisition module, a suspected source area determination module, and a source tracing and analysis module;

[0151] The data acquisition module is used to acquire multi-source data of the hard rock mining operation area and construct a mining disaster analysis dataset based on the multi-source data. The multi-source data includes underground mine image or video data, microseismic or acoustic emission data, drilling and blasting construction parameter data, and geological and mining progress data.

[0152] The suspected source area determination module is used to construct a fracture disaster-inducing prior field reflecting the evolution state of surrounding rock fracture based on the mining disaster analysis dataset, and to determine the suspected source area of ​​the disaster.

[0153] The source tracing analysis module is used to construct a disaster source tracing relationship model for the suspected disaster source area, and to perform inversion inference on the disaster source tracing relationship model to obtain the disaster source area location, triggering factors and triggering time.

[0154] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0155] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for tracing and analyzing the sources of disasters in hard rock mining based on big data, characterized in that, include: Acquire multi-source data of the hard rock mining operation area, and construct a mining disaster analysis dataset based on the multi-source data. The multi-source data includes underground mine image or video data, microseismic or acoustic emission data, drilling and blasting construction parameter data, and geological and mining progress data. Based on the mining disaster analysis dataset, a fracture-induced disaster prior field reflecting the evolution state of surrounding rock fracture is constructed to identify potential disaster source areas; For the suspected source areas of the disaster, a disaster source tracing relationship model is constructed, and the disaster source tracing relationship model is inverted and inferred to obtain the source area location, triggering factors and triggering time of the disaster; Based on the aforementioned mining disaster analysis dataset, a fracture-prone prior field reflecting the evolution state of surrounding rock fracture is constructed to identify potential disaster source areas, including: The mining disaster analysis dataset is sliced ​​according to a preset time window, and the hard rock mining area is divided into multiple spatial units. A corresponding feature container is established for each spatial unit within each time window. The feature container is used to carry the fracture fingerprint entries and event fragments associated with the spatial unit. The crack fingerprint entries within each spatial unit are arranged in chronological order to generate a crack fingerprint sequence. Using the fissure fingerprint sequence as the anchoring object, event segments associated with the same spatial unit are aligned to the fissure fingerprint sequence according to their temporal proximity. When event segments overlap or have inconsistent indications within the same time window, the overlap or inconsistency is marked and resolved according to a preset priority rule to obtain a resolved fused sequence. Based on the resolved fusion sequence, each spatial unit is assigned a corresponding prior field value, and a topological adjacency relationship is introduced between adjacent spatial units to form a prior field grid. In the prior field grid, identify a set of spatial units that satisfy a preset connectivity rule, and encapsulate the set of spatial units as a suspected disaster source area; For the suspected disaster source areas, a disaster source tracing relationship model is constructed, and the model is inverted to obtain the disaster source area location, triggering factors, and triggering time, including: For each suspected disaster source area, its corresponding spatial boundary and time range are determined, and crack fingerprint sequences and event fragments falling within the spatial boundary and time range are extracted from the mining disaster analysis dataset to form a source area evidence set; In the source region evidence set, a disaster outcome anchor point is determined, and a retrospective time window is generated based on the disaster outcome anchor point. The retrospective time window is then divided into multiple retrospective levels in chronological order. Within each retrospective level, the crack fingerprint sequence and event fragments are mapped to event nodes, and directed relationships are set between all event nodes to construct a disaster source tracing relationship model; Inversion inference constraints are generated for the disaster source tracing relationship model. Under the condition of satisfying the inversion inference constraints, candidate source tracing paths are enumerated backward from the disaster result anchor point, and a corresponding path identifier and path evidence list are generated for each candidate source tracing path. Based on the aforementioned path evidence list, candidate source tracing paths are adjudicated to determine the target source tracing path, and the source area location, triggering factors, and triggering time of the disaster are output based on the target source tracing path.

2. The method for tracing and analyzing the source of disasters in hard rock mining based on big data as described in claim 1, characterized in that, The acquisition of multi-source data of the hard rock mining operation area, and the construction of a mining disaster analysis dataset based on the multi-source data, includes: A set of analysis objects is determined for the hard rock mining operation area, and a unified object identifier is assigned to each unit. The set of analysis objects includes stope space units, roadway topology units, operation event units, and monitoring point units. Multi-source data is obtained based on the object identifier. The multi-source data is encapsulated into event fragments according to a preset sampling window. Each event fragment is written with a corresponding object identifier and acquisition time identifier. The multi-source data includes underground mine image or video data, microseismic or acoustic emission data, drilling and blasting construction parameter data, and geological and mining progress data. The event fragments include underground mine image or video event fragments, microseismic or acoustic emission event fragments, drilling and blasting construction parameter event fragments, and geological and mining progress event fragments. One type of data from the multi-source data is selected as the reference time scale source; time scale drift correction and time segment resampling are performed on the remaining multi-source data so that each event segment forms a time series under the same reference time scale. Extract the fracture skeleton, block boundary discontinuity segments, and texture abrupt change segments from images or videos in the time series, and combine them into fracture fingerprint entries for mining space units. The crack fingerprint entries and event fragments are associated and cataloged according to the object identifier to form a mining disaster analysis dataset.

3. The method for tracing and analyzing the source of disasters in hard rock mining based on big data as described in claim 1, characterized in that, Based on the resolved fusion sequence, a corresponding prior field value is assigned to each spatial unit, and a prior field grid is formed by introducing topological adjacency relationships between adjacent spatial units, including: For each spatial unit, the resolved fusion sequence is segmented according to temporal continuity, and a corresponding surrounding rock state identifier is enumerated for each segment. The surrounding rock state identifier is determined by the change segment of the fracture fingerprint sequence and the event segment aligned with it. Within the same spatial unit, the segments are sorted according to the chronological order of the surrounding rock state indicators, and a state trajectory is generated. The state trajectory serves as the internal evolution description benchmark for the spatial unit. The state trajectory is mapped to the corresponding prior field value, and the prior field value is bound to the corresponding spatial unit; The adjacency pairs between each spatial unit are determined, and the propagation relationship of the prior field value is established between adjacent spatial units. The prior field values ​​of adjacent spatial units are synchronously adjusted according to the time consistency and topological reachability rules to form a prior field grid.

4. The method for tracing and analyzing the source of disasters in hard rock mining based on big data as described in claim 3, characterized in that, Identify a set of spatial cells that satisfy a preset connectivity rule in the prior field grid, and encapsulate the set of spatial cells as a suspected disaster source area, including: Based on the spatial cells and their topological adjacency relationships in the prior field grid, preset connectivity rules are generated, and rule identifiers are assigned to each connectivity rule. Within each time window, spatial units that meet the preset prior field value conditions are selected from the prior field grid as seed units, and the seed units are associated with their corresponding allocation rule identifiers to form a seed set. Starting with the seed set, perform connectivity expansion in the prior field grid according to the preset connectivity rules corresponding to the rule identifier to obtain a set of spatial cells; The spatial unit set is encapsulated as a suspected disaster source area.

5. The method for tracing and analyzing the source of disasters in hard rock mining based on big data as described in claim 1, characterized in that, Within each retrospective level, the crack fingerprint sequence and event fragments are mapped to event nodes, and directed relationships are established between all event nodes to construct a disaster source tracing relationship model, including: Within each backtracking level, the gap fingerprint sequence falling into that backtracking level and various event fragments are merged according to the object identifier; The fracture fingerprint sequence is segmented into change segments and mapped to fracture event nodes. The drilling and blasting construction parameter event segments and geological and mining progress event segments are mapped to disturbance event nodes. The microseismic or acoustic emission event segments are mapped to response event nodes. The disaster result anchor points are mapped to result event nodes, forming a cross-modal consistent event node set. Based on the set of event nodes, a candidate set of directed relationships is generated according to preset rules; The candidate set of directed relations is assembled into directed relations between event nodes, and loop relations, directional conflict relations and duplicate relations are pruned. Based on the event node set and its directed relations obtained after pruning, a disaster tracing relation model is constructed.

6. The method for tracing and analyzing the source of disasters in hard rock mining based on big data as described in claim 5, characterized in that, Inversion inference constraints are generated for the disaster source tracing model. Under the condition that the inversion inference constraints are satisfied, candidate source tracing paths are enumerated backward from the disaster result anchor point, and a corresponding path identifier and path evidence list are generated for each candidate source tracing path, including: Based on the source type identifier and backtracking level identifier of the event node, a set of constraint templates for inversion inference constraints is set, and the set of constraint templates is instantiated on the disaster source tracing relationship model to obtain the inversion inference constraint set; Starting from the aforementioned disaster outcome anchor point, a backtracking search boundary is set; Within the backtracking search boundary, a reverse traversal is performed from the disaster result anchor point along the directed relationship in the disaster source tracing relationship model. During each traversal expansion, the candidate expansion edge and candidate expansion node are verified by the inversion inference constraint set. The traversal branches that pass the verification are used to form candidate source tracing paths. For each candidate tracing path, a path identifier is generated, and based on each candidate tracing path, the index of the event node reference fragments it passes through and the corresponding constraint verification records are summarized to generate a list of path evidence.

7. The method for tracing and analyzing the source of disasters in hard rock mining based on big data as described in claim 6, characterized in that, Based on the aforementioned path evidence list, candidate source tracing paths are adjudicated to determine the target source tracing path. Based on the target source tracing path, the source area location, triggering factors, and triggering time of the disaster are output, including: The path evidence lists for each candidate tracing path are merged according to evidence type, and the corresponding event node identifier and reference fragment index are written for each type of evidence after merging to form an evidence vector; A set of adjudication rules is generated based on the evidence vector description. Candidate tracing paths that do not meet any adjudication rule are filtered out according to the set of adjudication rules to obtain a set of retained candidates. The reserved candidate set is adjudicated according to a preset hierarchical priority, the target tracing path is determined from the reserved candidate set, and an adjudication identifier is written for the target tracing path; Based on the target tracing path, the source elements are extracted, and the source area location, triggering factors, and triggering time of the disaster are output.

8. A big data-based hard rock mining disaster source tracing and analysis system, used to implement the big data-based hard rock mining disaster source tracing and analysis method according to any one of claims 1-7, characterized in that, include: Data acquisition module, suspicious source area identification module, and source tracing analysis module; The data acquisition module is used to acquire multi-source data of the hard rock mining operation area and construct a mining disaster analysis dataset based on the multi-source data. The multi-source data includes underground mine image or video data, microseismic or acoustic emission data, drilling and blasting construction parameter data, and geological and mining progress data. The suspected source area determination module is used to construct a fracture disaster-inducing prior field reflecting the evolution state of surrounding rock fracture based on the mining disaster analysis dataset, and to determine the suspected source area of ​​the disaster. The source tracing analysis module is used to construct a disaster source tracing relationship model for the suspected disaster source area, and to perform inversion inference on the disaster source tracing relationship model to obtain the disaster source area location, triggering factors and triggering time.

Images