Metro tunnel blasting harmful effect peak value prediction system based on multi-source data

Abstract

The present application relates to the technical field of blasting engineering, specifically is the subway tunnel blasting harmful effect peak value prediction system based on multi-source data, including: the duration of response blasting effect, obtain harmful effect data at different subway tunnel structure; according to the continuous distribution degree and dispersion trend of harmful effect data, delimit the regional unit and regional identification corresponding to harmful effect data;Peak value characteristic analysis is carried out on harmful effect data, the change trend of peak value in each regional unit is calculated, and the peak value interval of each regional unit is delimited;The regional unit is converted into an early warning node, the peak value interval of each early warning node is associated and identified, and the association between the early warning nodes is determined;The association between the early warning nodes is mapped to each blasting point, and the peak value mapping relationship under each blasting is determined, so that the early warning activity degree of different regions is determined. The processing efficiency and scene pertinence of peak value prediction are improved.

Description

Technical Field

[0001] This invention relates to the field of blasting engineering technology, specifically a peak prediction system for harmful effects of blasting in subway tunnels based on multi-source data. Background Technology

[0002] In railway and subway tunnel blasting projects, the harmful effects of blasting are generally divided into rock mass vibration and air shock waves. The peak values ​​of these two effects directly determine the safety of the tunnel structure and are core factors leading to problems such as tunnel lining cracking, track deformation, and structural damage. Therefore, predicting the peak blast value is a prerequisite for the safety management of subway tunnel blasting.

[0003] For example, Chinese Patent Publication No. CN113837440A discloses a method, device, electronic device, and medium for predicting blasting effects; wherein, the method includes: acquiring a dataset corresponding to the blasting parameters of the blasting site; inputting the samples contained in the test set of the dataset into a pre-trained random forest (RF) blasting effect prediction model to obtain the corresponding output results; and determining the blasting effect prediction result of the blasting site based on the output results.

[0004] For example, Chinese Patent Publication No. CN118194203A discloses a blasting intelligent control method based on collaborative management. This method includes analyzing the operating status of target equipment based on the reliability values ​​of sub-regional parameters to provide basic information for subsequent data processing and management; analyzing the abnormal causes of target equipment by constructing a fault tree and quantitatively analyzing the accident rate of abnormal target equipment to determine the degree of deviation; obtaining fault risk assessment standards by judging the degree of deviation and analyzing the accident probability of abnormal target equipment; generating control instructions by comparing on-site detection results and fault risk assessment standards; and locating faulty equipment to determine the cause of deviation.

[0005] In existing technologies, blasting effects are predicted through blasting parameters, and data is selected based on the importance of the blasting parameters; or blasting data is detected and analyzed through fault tree analysis. However, existing technologies tend to focus on the analysis and processing of multiple blasting tasks, ignoring the differences in the continuous distribution, dispersion trend, and peak representation of the actual space. This results in a one-dimensional and one-sided analysis of the data processing, leading to problems such as scene lag in forecast analysis and isolated early warning modes. Summary of the Invention

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a peak prediction system for harmful effects of blasting in subway tunnels based on multi-source data, comprising: a data acquisition module, used to acquire harmful effect data at different subway tunnel structures in response to the duration of the blasting effect; wherein, the harmful effect data includes vibration effect data and impact effect data.

[0007] The region division module is used to delineate the regional units and region identifiers corresponding to the harmful effect data based on the degree of continuous distribution and dispersion trend of the harmful effect data.

[0008] The parameter analysis module is used to perform peak characteristic analysis on harmful effect data based on the regional range and regional identifier of the regional unit, calculate the changing trend of the peak value in each regional unit, and delineate the peak value interval of each regional unit.

[0009] The peak identification module is used to convert regional units into early warning nodes, identify the correlation between the peak ranges of each early warning node, and determine the correlation between the early warning nodes.

[0010] The index evaluation module is used to map the correlation between early warning nodes to each blasting point, and to determine the level of early warning activity in different areas based on the peak mapping relationship under each blast.

[0011] The beneficial effects of this invention are as follows: First, by responding to the duration of the blasting effect, this invention obtains harmful effect data of vibration and impact effects at different structural locations in subway tunnels; second, based on the continuous distribution and dispersion trend of the harmful effect data, it delineates regional units and regional identifiers; simultaneously, guided by the range and identifiers of the regional units, it conducts peak characteristic analysis on the harmful effect data, delineating the peak range of each regional unit; finally, it transforms the regional units into early warning nodes, and through the correlation and combination with blasting points, it clarifies the early warning activity level of different regions. This avoids the lack of dimensions in the current scenario and simultaneously records the relative range and related trends of the regional units, improving the comprehensiveness, accuracy, and linkage of peak forecasts; finally, the mapped data directly binds different blasting conditions with actual effects, improving the processing efficiency in various scenarios.

[0012] Second, this invention constructs stage numbers based on the input vibration and impact effect data, uses these stage numbers to define the boundaries of each blasting task, determines the area location corresponding to a single blasting analysis, and filters out valid harmful effect data. This improves the targeting and efficiency of current data processing. Simultaneously, based on the set stage numbers, it calls the index pairing information between the stage numbers and blasting tasks to form index number groups; data verification is performed on each index number group to check the jump boundaries under each stage, and the number boundaries are adjusted based on the jump boundaries. Through the one-to-one correspondence between stage numbers and blasting tasks, the entire process is traceable, avoiding problems such as misnumbering and incorrect data attribution, further ensuring the accuracy and reliability of the data source.

[0013] Third, this invention divides the hazardous effect data by location, extracting the corresponding location interval for each data set; it extracts features from the hazardous effect data to generate a continuous risk surface; based on the gradient and distribution pattern of the isohyets in the continuous risk surface, it verifies the degree of continuity and dispersion trend of the data, dividing the location interval into multiple regional units. This intuitively displays the impact range of blasting hazardous effects within the linear space of the tunnel, ensuring that each regional unit corresponds to the same planned risk characteristics, laying a data foundation for subsequent processing.

[0014] IV. This invention captures peak features by using the scope and identifier of regional units as a guide and following a continuous time window; it filters data based on the value range of peak features to construct a peak sequence; it calculates the peak difference of the peak sequence over a continuous time period, using the peak change rate as the peak change trend; it judges in real time whether the peak change rate meets the steady-state condition according to the analysis period, and if so, it filters data according to the peak change rate to determine the steady-state time period, which is used as the current feature dimension for processing, and finally delineates the peak interval of the regional unit. This ensures the spatial specificity of peak analysis and improves the accuracy and anti-interference ability of peak forecasting.

[0015] Fifth, this invention sets regional units as early warning nodes and assigns node identifiers based on the range and identifier of the regional units. For each early warning node, when harmful effect data is collected, it traverses other adjacent early warning nodes to mine the correlation between early warning nodes. This realizes the mining of effect correlations between adjacent regions, provides a data foundation for coordinated early warning, and improves the processing efficiency of peak forecast. Attached Figure Description

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0017] Figure 1 This is a system framework diagram of a subway tunnel blasting harmful effect peak prediction system based on multi-source data.

[0018] Figure 2 This is a flowchart illustrating the regional division module of a subway tunnel blasting hazard effect peak prediction system based on multi-source data;

[0019] Figure 3 This is a flowchart illustrating the parameter analysis module of a subway tunnel blasting harmful effect peak prediction system based on multi-source data. Detailed Implementation

[0020] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. Where specific techniques or conditions are not specified in the embodiments, they shall be performed in accordance with the techniques or conditions described in the literature in the art or in accordance with the product manual.

[0021] See Figure 1 A peak forecasting system for harmful effects of blasting in subway tunnels based on multi-source data includes: a data acquisition module, a region division module, a parameter analysis module, a peak identification module, and an index evaluation module. The output of the data acquisition module is connected to the region division module, the output of the region division module is connected to the parameter analysis module, the output of the parameter analysis module is connected to the peak identification module, and the output of the peak identification module is connected to the index evaluation module.

[0022] The data acquisition module is used to respond to the duration of the blasting effect and acquire harmful effect data at different subway tunnel structures; among which, the harmful effect data includes vibration effect data and impact effect data.

[0023] The region division module is used to delineate the regional units and region identifiers corresponding to the harmful effect data based on the degree of continuous distribution and dispersion trend of the harmful effect data.

[0024] The parameter analysis module is used to perform peak characteristic analysis on harmful effect data based on the regional range and regional identifier of the regional unit, calculate the changing trend of the peak value in each regional unit, and delineate the peak value interval of each regional unit.

[0025] The peak identification module is used to convert regional units into early warning nodes, identify the correlation between the peak ranges of each early warning node, and determine the correlation between the early warning nodes.

[0026] The index evaluation module is used to map the correlation between early warning nodes to each blasting point, and to determine the level of early warning activity in different areas based on the peak mapping relationship under each blast.

[0027] The vibration effect data consists of particle vibration velocity, vibration acceleration, and vibration duration, representing waveform data of explosion vibration; the impact effect data consists of overpressure peak value, impulse, and barometric pressure duration, representing the relative air pressure generated after the explosion.

[0028] Specifically, the peak overpressure is the maximum pressure difference between the air shock wave generated by the explosion and the ambient atmospheric pressure during its propagation; the impulse is the maximum value of the integral of overpressure over time during the positive pressure action time of the shock wave; and the positive pressure action time is the maximum value of the time it takes for the shock wave overpressure to rise to its maximum value and decay to the ambient atmospheric pressure, which directly determines the magnitude of the impulse.

[0029] In the current scenario, the data acquisition module is used to gradually deploy pressure sensors and vibration sensors along the blasting operation area closest to the subway tunnel. Each vibration sensor is placed on the tunnel structure, and each pressure sensor is 1.2m-1.5m above the ground to avoid interference from ground reflection waves. The pressure sensors receive the pressure signals generated by the explosion and measure transient pressure changes. The vibration sensors convert the ground vibration signals into electrical signals and determine the relative impact of the explosion by the peak values ​​of the relative frequencies. The collected data is then aggregated into multiple peak intervals containing characteristic peaks according to the acquisition time of the multi-sensor synchronous triggering.

[0030] In one embodiment of the present invention, the data acquisition module is used to divide the subway tunnel into a grid network, associate the input harmful effect data according to the input stage and spatial location, and label the output vibration effect data and impact effect data according to specific numbers and data jump change patterns, thereby clarifying the data subject of the current scene processing.

[0031] Specifically, the data acquisition module is implemented by: constructing stage numbers for the collected harmful effect data based on the input vibration effect data and impact effect data; the stage numbers are used to record the processing time and location of the blasting task.

[0032] Specifically, the stage numbers correspond to the stages of the blasting process (such as pre-splitting, loosening, main blasting, hazard removal, and slag removal), and each stage is defined with a corresponding number of blasts.

[0033] Each blasting task is numbered and demarcated using stage numbers to determine the location of the area involved in each blasting analysis, and the output harmful effect data is determined according to the activation status of each area.

[0034] The location of this area represents the area corresponding to the current stage. Each area is marked with an activation status identifier according to whether it has been blasted, is awaiting blasting, or does not require blasting. Hazardous effect data is generated under the corresponding identifier to characterize the activation status of each area. The output hazardous effect data includes the gridded coordinates corresponding to the area location and the stage description corresponding to the stage number.

[0035] Furthermore, when numbering the vibration effect data and impact effect data into stages, in addition to numbering the data according to the relative stages of each blasting task, it is also necessary to determine the relative jump boundary under each stage and verify the processing direction of the current blasting task.

[0036] When using stage numbers to demarcate each blasting task, the implementation method further includes: based on the set stage number, calling the index pairing information between the stage number and the blasting task to form an index number group. The index pairing information represents the pairing between the stage number and the blasting task, and these are arranged in the order of the stage numbers to form the index number group.

[0037] For each index number group, perform data verification, check the jump boundaries under each stage, and adjust the numbering boundaries based on the jump boundaries.

[0038] Among them, the jump boundary represents the time interval where a certain parameter is missing, discontinuous, or the parameter changes significantly in adjacent stages; the numbering boundary indicates that discontinuous, missing, and significantly changing data parts need to be removed according to the time interval to obtain a relatively stable data distribution, thereby forming harmful effect data with stage numbers; the missing and discontinuous scenarios are specifically when a certain borehole or a group of boreholes is not detonated as designed, the stage numbers are continuous but there is no blasting signal in a certain segment, or the vibration waveform has a discontinuity, resulting in the output data being discontinuous; the data represents the abnormal situation of fault / misfire, and will be directly used as the output content of the abnormal scenario, and will not be used as the main body of the subsequent peak analysis.

[0039] As for the significant changes in parameters, which mostly represent the actual time points of the blasting mission, the jump start point and jump end point will be set according to the time points of the corresponding stage numbers, and the corresponding data will be used as the main body of the subsequent peak analysis according to the stage corresponding to the jump boundary.

[0040] Furthermore, the jump boundary represents the time interval between adjacent blasting stages, from when the harmful effects of the previous stage decay to below three standard deviations of the environmental baseline, to when the peak effect of the next stage begins to jump; among them, the criterion for determining a significant change in parameters between adjacent stages is that the peak change rate within an adjacent 10ms window is greater than or equal to 200%.

[0041] In one embodiment of the present invention, the region division module is used to define the spatial concentration of blasting tasks at each stage, quantify the execution of each task according to the degree of continuous distribution and dispersion trend of the concentration of harmful effect data, and form location-related regional units.

[0042] like Figure 2 As shown, the implementation of the region division module includes: dividing the input harmful effect data into positions and extracting the position interval corresponding to each group of harmful effect data; wherein, the position interval represents the data interval of each group of harmful effect position combinations.

[0043] Feature extraction is performed on the components of the harmful effect data to generate a continuous risk surface. The continuous risk surface represents the data distribution pattern divided by contour lines after data interpolation. The generation process of the continuous risk surface is adjusted according to different data components.

[0044] Specifically, in scenarios where the harmful effects data consist of vibration and impact data, there are six different dimensions of data. If the current identification content is only one dimension, the contour lines of the continuous risk surface can be set directly through Kriging space interpolation. If there are six dimensions, the corresponding continuous risk surfaces can be generated for each dimension, or the data of the three dimensions can be summarized in sequence according to the inclusion form of vibration and impact effects. For example, after normalizing the data of all dimensions, a weighted sum is performed, and the continuous risk surface is generated based on the weighted sum value. The weights are set according to the proportion of the normalized value of each dimension to the total weight.

[0045] Based on the gradient and distribution of mean lines in the continuous risk surface, the degree of continuous distribution and dispersion trend of harmful effect data are checked, and the analysis location interval is divided into multiple regional units.

[0046] Among them, the degree of continuous distribution describes the sparsity and smoothness of the distribution of contour lines under continuous distribution according to the gradient characteristics of contour lines; the dispersion trend determines whether the contour lines are concentrated or widely distributed in space according to the form of convergence closure and continuous distribution dispersion, thereby clarifying the spatial distribution of harmful effect data.

[0047] Specifically, considering the degree of continuous distribution, a scenario with sparse contour lines and small gradients in space indicates a high degree of continuous distribution, where the risk changes slowly in space without drastic abrupt changes; similarly, dense contour lines and large gradients indicate a low degree of continuous distribution, with local abrupt changes or breaks.

[0048] Specifically, considering the dispersion trend, a clear separation between closed and dispersed contour lines in the risk surface is considered a low dispersion trend, with risk areas clustered together; similarly, a fragmented scenario without clear closure or separation indicates a high dispersion trend, with risk areas widely distributed and without a clear core.

[0049] Furthermore, the dispersion trend is divided into relative values ​​based on whether or not it is closed, and the degree of continuous distribution is determined by retrieving historical data according to the gradient of the contour lines. Based on the median value of the historical data in the corresponding scenario, it is divided into low continuity and high continuity. In other words, the sparsity of the contour lines in the current space is quantified by the average value of the gradient magnitude of the current risk surface.

[0050] Based on the above, and according to the degree of continuity and the trend of dispersion, regional units are divided into combinations of high continuity + low dispersion, high continuity + high dispersion, low continuity + low dispersion, and low continuity + high dispersion.

[0051] Among them, high continuity + low dispersion indicates that the energy decays uniformly and is concentrated in one area, requiring local protection of this area to verify the relative peak value of the harmful effect; high continuity + high dispersion indicates that the energy is uniform but widely dispersed, requiring determination of whether there are multiple overlapping segments in this area, and using the characteristics of its peak value to determine whether the total explosive charge should be reduced subsequently; low continuity + low dispersion indicates the existence of a clear abrupt change boundary, but the area after the abrupt change is concentrated, and the boundary of this area may be a geological fault or a man-made blast wall, which can be used as a boundary to isolate the risk; low continuity + high dispersion indicates multiple abrupt changes in space and fragmented risk, and the number of measuring points may be insufficient or there may be multiple independent blast sources, requiring densification of measuring points in this area and re-division of the area and peak value analysis.

[0052] Furthermore, the area identifier is used for spatial statistics of area units and data interpretation during the blasting phase, to illustrate the data combination form of the divided area units.

[0053] Specifically, the implementation of region identifiers in the region division module includes: parsing the data structure of the region unit and extracting the blasting task stage, spatial location, and temporal location from the region unit's stored fields.

[0054] Based on the parameter changes of the blasting task stage, spatial location, and temporal location corresponding to the stored fields, all related data information is searched, and the data information is segmented and identified sequentially as the output area identifier.

[0055] The regional identifier will search for all data corresponding to the blasting mission stage, spatial location, and temporal location, statistically analyze this data, and set corresponding semantic description identifiers to form regional identifiers for different regional units, spatial location segments, temporal location segments, and blasting mission stage segments, so as to represent the peak value of each location within a specific time interval.

[0056] In one embodiment of the present invention, the parameter analysis module constructs structured peak features for each regional unit based on the characteristic peak expression form of each regional unit, and combines them into the output peak interval in the form of data intervals; then the peak identification module uses the structured peak features between peak intervals to perform correlation analysis on different regional units, transforming the overall peak transmission form into scheduling path and resource constraints, thereby clarifying the peak detection scenario of multi-dimensional data.

[0057] like Figure 3As shown, the implementation of peak intervals in the parameter analysis module includes: using the regional range and regional identifier of the regional unit as a guide, and capturing peak characteristics according to a continuous time window.

[0058] Among them, the peak feature is obtained by processing the relevant waveforms of vibration effect data and impact effect data, identifying each characteristic peak in the form of a continuous time window, and selecting the currently captured peak feature according to the number, area and symmetry of the corresponding characteristic peak as the analysis criteria, so as to characterize the representative events under the peak division, improve the accuracy of peak selection, and thus improve the effect of subsequent peak analysis.

[0059] Specifically, the captured peak features include: the number of characteristic peaks within a continuous time window, the mean of the asymmetry of all characteristic peaks, and the sum of the peak areas of all characteristic peaks. These data are used as the peak features for verification.

[0060] The degree of asymmetry is set using an asymmetry factor, which is the ratio of the trailing edge width to the leading edge width at 10% peak height, reflecting the tailing and leading edge extension of the waveform. Secondly, the 10% peak height is the ratio of the falling edge width to the rising edge width of the characteristic peak relative to the peak apex. The rising edge is the time interval from the peak-valley baseline to the peak apex, and the falling edge is the time interval from the peak apex to the returning peak-valley baseline. This asymmetry factor is used to explain the relative situation of the explosion vibration waveform and the impact waveform represented by the vibration effect, in order to explain the relative characteristics of their characteristic peaks.

[0061] Furthermore, feature filtering is performed based on the value range of the peak features, and the filtered peak features are used to construct a peak sequence.

[0062] In the feature selection process, upper and lower thresholds are set for the number of feature peaks to exclude windows with excessively sparse or dense signals. When the upper threshold for the number of feature peaks is exceeded, feature peaks that meet the upper threshold are selected in reverse order based on their height, from largest to smallest. The mean asymmetry is required to be within the range [0.2, 5.0], and abnormal distortion peaks outside this range are removed. Finally, the peak area needs to be set to a minimum sum of peak areas, ensuring sufficient signal strength within the continuous time window.

[0063] Specifically, the upper and lower limits of the number of feature peaks and the minimum value of the sum of peak areas are set by using quantiles of historical data in specific scenarios. For example, the upper and lower limits of the number of feature peaks are selected from the median value of the historical data in ascending order and the quartile at the 75th percentile, respectively, and the sum of peak areas is selected from the median value of the historical data as the minimum value to be distinguished here, so as to determine the effectiveness of peak feature selection.

[0064] It should be noted that after the peak sequence is filtered, it is combined into an output peak sequence according to the peak value corresponding to each feature peak, and the number of feature peaks, the mean of the degree of asymmetry, and the peak area of ​​the feature peaks corresponding to the current peak feature are marked.

[0065] Calculate the peak difference of the peak sequence over a continuous time period, and regard the rate of change of its peak as the trend of peak change.

[0066] Based on the current data analysis time period, determine in real time whether the peak change rate meets the steady-state condition. If the steady-state condition is met, perform data filtering according to the peak change rate to determine the time period that meets the condition.

[0067] By considering the time period that is satisfied as the current feature dimension, the peak range of the regional unit is determined.

[0068] Among them, steady-state conditions are used for real-time detection of blasting vibrations and detection and processing of shock wave overpressure; when the peak change rate is used as the main body of trend analysis, its peak change rate often shows a non-monotonic trend; when setting steady-state conditions, on the one hand, according to the data type being analyzed, the standard deviation and mean of the peak value are calculated respectively to check the fluctuation range of the change; on the other hand, the peak change rate is used as the calculation basis to check the relative change rate of adjacent time periods; according to the scenarios corresponding to the current vibration effect data and shock effect data, the thresholds corresponding to the vibration effect and shock effect are extracted from the database respectively, and the time period that meets the steady-state conditions is recorded.

[0069] When the current harmful effects include vibration duration and positive pressure duration, peak analysis of adjacent time periods is used to determine the overall fluctuation range and relative rate of change for these two parameters; other data are analyzed directly based on the statistical data of adjacent characteristic peaks.

[0070] Specifically, a threshold for the rate of change is set based on the median value of historical data. Secondly, according to the data type being analyzed, the median value of historical data corresponding to each parameter is found, and a threshold for the fluctuation range to be verified for each parameter is selected. The selection method is as follows: particle vibration velocity is selected with a range less than or equal to 15% of the mean; vibration acceleration is selected with a standard deviation less than or equal to 10% of the mean; vibration duration is selected with a range less than or equal to 15% of the mean; overpressure peak value is selected with a coefficient of variation less than or equal to 0.15; impulse is selected with a ratio of range to mean less than or equal to 0.2; and barotropic action time is selected with a ratio of range to mean less than or equal to 0.25. Based on these data values, the main body of parameter analysis for each dimension is obtained. The values ​​described here are illustrative representations of the median value of historical data; the specific values ​​are adjusted according to the actual blasting scenario.

[0071] In one embodiment of the present invention, the peak recognition module is used to introduce the geographical location of the subway tunnel, match the regional units with the scene described during the division, and clarify the correlation between the peak intervals of each region.

[0072] The peak identification module is implemented by setting the regional unit as an early warning node based on the regional range and regional identifier of the regional unit, and assigning a node identifier to the early warning node; wherein, the node identifier of the early warning node is set based on the data combination of the peak interval of the regional unit, and the peak value of the blasting task under the corresponding process stage is recorded synchronously.

[0073] For each warning node, when harmful effect data is collected, other adjacent warning nodes are traversed to explore the correlation between warning nodes; the explored warning nodes are marked at specific locations in the subway tunnel and pushed to the index evaluation module in a synchronized manner to determine the location of the peak distribution under each blast, and thus determine the activity level of warnings in each area.

[0074] Furthermore, when mining the correlation between early warning nodes, the implementation method includes: for each other early warning node traversed, using nearest neighbor difference consistency, constraining the peak range of other early warning nodes to ensure that the values ​​of adjacent early warning nodes are related.

[0075] Among them, the nearest neighbor difference consistency is used to indicate that the relative difference between the current warning node and other adjacent warning nodes does not exceed the allowable construction safety threshold, which is determined according to the project construction scenario.

[0076] To satisfy the nearest neighbor difference consistency, the current warning node is combined with other warning nodes, and the similarity of each pair of adjacent warning nodes is calculated by converting the peak interval into a vector; the similarity of warning nodes is calculated by cosine similarity.

[0077] If the nearest neighbor difference consistency is not met, it indicates that there are hidden geological changes or sensor failures between the two early warning nodes, requiring designs such as composite borehole delay.

[0078] The association between warning nodes is configured based on the similarity between each pair of adjacent warning nodes. The similarity of each pair of adjacent warning nodes defines the association state of the adjacent nodes according to its value, including: a strong association with a similarity greater than or equal to 0.7, where the peak intervals are highly similar, and the corresponding regional units exhibit similar continuous distribution, dispersion trends, and peak distribution; a moderate association with a similarity between 0.4 and 0.7, indicating that the regional units have some similarity, but parameters such as peak value and continuous distribution differ, making them susceptible to adjustments in local parameters during blasting operations; and a weak association with a similarity < 0.4, representing significant overall differences between two consecutive regional units, where spatial discontinuities are likely in terms of continuous distribution, peak distribution, etc., and abnormal parameter identification abrupt changes.

[0079] In one embodiment of the present invention, the index evaluation module is used to synchronize the correlation of the early warning node to the blasting point of the blasting task, so as to clarify the backtracking data and process description between peak changes and the blasting task execution process.

[0080] Among them, the peak mapping relationship represents the mapping relationship between the data combination form of the warning node and the configuration of the blasting point under a specific association relationship. It represents the relative mapping between the peak value identified by the warning node and the corresponding blasting point after the association relationship of the current warning node is synchronized to the blasting point.

[0081] Furthermore, the implementation of the index evaluation module includes: when the relationship between adjacent warning nodes changes, viewing the warning nodes that are input each time, and quantifying the warning activity level of adjacent warning nodes based on the input frequency of the warning nodes.

[0082] The change in the correlation means that the adjacent warning point pairs may change due to the execution of the blasting task, or the adjustment of blasting parameters, cumulative damage, etc., which will cause changes in the content of weak correlation, strong correlation, etc. In this case, it is necessary to first determine whether the adjacent warning nodes need to be warned according to the construction safety threshold, and quantify the frequency of the recording of the adjacent warning point pairs before and after the change in the correlation, so as to determine the relative change of the coupling between the two, and then focus on the synchronous update status of each parameter under the blasting environment.

[0083] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered within the protection scope of the present invention.

Claims

1. A peak value prediction system for harmful effects of blasting in subway tunnels based on multi-source data, characterized in that, include: The data acquisition module is used to respond to the duration of the blasting effect and acquire harmful effect data at different subway tunnel structures; among which, the harmful effect data includes vibration effect data and impact effect data; The region division module is used to delineate the regional units and region identifiers corresponding to the harmful effect data based on the degree of continuous distribution and dispersion trend of the harmful effect data; The parameter analysis module is used to perform peak characteristic analysis on harmful effect data based on the regional range and regional identification of the regional unit, calculate the changing trend of the peak value in each regional unit, and delineate the peak value interval of each regional unit. The peak identification module is used to convert regional units into early warning nodes, identify the correlation between the peak ranges of each early warning node, and determine the correlation between the early warning nodes. The index evaluation module is used to map the correlation between early warning nodes to each blasting point, and to determine the level of early warning activity in different areas based on the peak mapping relationship under each blast.

2. The peak value prediction system for harmful effects of blasting in subway tunnels based on multi-source data according to claim 1, characterized in that, The methods for implementing harmful effect data in the data acquisition module include: Based on the recorded vibration effect data and impact effect data, a stage numbering system is constructed for the collected harmful effect data. Each blasting task is numbered and demarcated using stage numbers to determine the location of the area involved in each blasting analysis, and the output harmful effect data is determined according to the activation status of each area.

3. The peak value prediction system for harmful effects of subway tunnel blasting based on multi-source data according to claim 2, characterized in that, When using stage numbers to demarcate each demolition task, the implementation methods also include: Based on the set stage number, the index pairing information between the stage number and the demolition task is called to form an index number group; For each index number group, perform data verification, check the jump boundaries under each stage, and adjust the numbering boundaries based on the jump boundaries.

4. The peak value prediction system for harmful effects of blasting in subway tunnels based on multi-source data according to claim 1, characterized in that, The implementation methods of region units in the region division module include: The entered harmful effect data is divided into locations, and the location interval corresponding to each group of harmful effect data is extracted; Feature extraction is performed on the components of the harmful effects data to generate a continuous risk surface; Based on the gradient and distribution of mean lines in the continuous risk surface, the degree of continuous distribution and dispersion trend of harmful effect data are checked, and the analysis location interval is divided into multiple regional units.

5. The peak value prediction system for harmful effects of blasting in subway tunnels based on multi-source data according to claim 1, characterized in that, The implementation methods of region identifiers in the region division module include: Analyze the data structure of the region unit and extract the explosive task stage, spatial location, and temporal location from the region unit's stored fields; Based on the parameter changes of the blasting task stage, spatial location, and temporal location corresponding to the stored fields, all related data information is searched, and the data information is segmented and identified sequentially as the output area identifier.

6. The peak value prediction system for harmful effects of blasting in subway tunnels based on multi-source data according to claim 1, characterized in that, The implementation methods for peak intervals in the parameter analysis module include: Using the regional scope and regional identifier of the regional unit as a guide, peak characteristics are captured according to a continuous time window; Feature filtering is performed based on the value range of the peak features, and the filtered peak features are used to construct a peak sequence. Calculate the peak difference of the peak sequence over a continuous time period, and regard the rate of change of its peak as the trend of peak change; Based on the current data analysis time period, determine in real time whether the peak change rate meets the steady-state condition. If the steady-state condition is met, perform data filtering according to the peak change rate to determine the time period that meets the condition. By considering the time period that is satisfied as the current feature dimension, the peak range of the regional unit is determined.

7. The peak value prediction system for harmful effects of subway tunnel blasting based on multi-source data according to claim 6, characterized in that, Peak characteristics captured according to continuous time windows include: The number of characteristic peaks within a continuous time window, the mean of the asymmetry of all characteristic peaks, and the sum of the peak areas of all characteristic peaks are used as the peak features for verification.

8. The peak value prediction system for harmful effects of blasting in subway tunnels based on multi-source data according to claim 1, characterized in that, The methods for implementing the correlation between early warning nodes in the peak identification module include: Based on the regional range and regional identifier of the regional unit, the regional unit is set as an early warning node, and a node identifier is assigned to the early warning node; For each warning node, when harmful effect data is collected, the system iterates through other adjacent warning nodes to explore the correlation between them.

9. The peak value prediction system for harmful effects of blasting in subway tunnels based on multi-source data according to claim 8, characterized in that, When mining the relationships between early warning nodes, the implementation methods include: For each of the other early warning nodes traversed, the nearest neighbor difference consistency is used to constrain the peak range of the other early warning nodes; When satisfying the nearest neighbor difference consistency, the current warning node is combined with other warning nodes, and the similarity of each pair of adjacent warning nodes is calculated by converting the peak interval into a vector. The association between warning nodes is configured by the similarity between each pair of adjacent warning nodes.

10. The peak value prediction system for harmful effects of blasting in subway tunnels based on multi-source data according to claim 1, characterized in that, The implementation methods of the index evaluation module include: When the relationship between adjacent warning nodes changes, examine the warning nodes that are entered each time, and quantify the warning activity level of adjacent warning nodes based on the input frequency of the warning nodes.

Images