Rock burst whole cycle cumulative, single early warning effect evaluation method and visualization method

Abstract

The application provides a rock burst whole cycle accumulation, single early warning effect evaluation method and visualization method, F1 scores are calculated for different rock burst grades respectively, and weights are given according to the damage degree of each rock burst grade, and the evaluation method gives different change sensitivities to each rock burst grade. This method scientifically evaluates the overall performance of the early warning system, especially in the case of unbalanced rock burst data. The design of independent calculation and weighting effectively avoids the deviation of traditional accuracy statistics, so as to avoid the high-frequency early warning results of low-grade rock burst from covering the early warning effect of the early warning system on high-grade rock burst.

Description

Technical Field

[0001] This invention relates to the field of rockburst early warning, and more specifically, to a method for evaluating the cumulative effect of rockburst throughout its entire cycle and for visualizing the effect of a single early warning. Background Technology

[0002] With the continuous development of underground resources and the rapid growth of the national economy, the country's demand for deep-earth engineering construction is increasing, and underground engineering is advancing into greater depths at an unprecedented pace. However, as construction progresses, the level of ground stress is constantly rising, and the geological environment is becoming increasingly complex. In complex geological and stress environments, rock masses are prone to fractures, leading to a decrease in rock mass strength. Under engineering disturbances, rock masses under high ground stress are easily induced to suddenly burst and erupt towards the free face, resulting in rockburst phenomena. This poses a severe challenge to the design, construction, and operation of deep underground engineering projects.

[0003] Currently, many scholars are conducting in-depth research in the field of rockburst early warning, proposing a series of rockburst early warning theories, methods, and technologies, making significant contributions to reducing the harm of rockburst hazards to engineering construction. In deep-earth engineering, effective monitoring, prediction, and early warning of rockburst hazards, along with the implementation of corresponding protective measures at construction sites, have significant practical implications.

[0004] However, in the field of rockburst early warning technology, most evaluation methods for the early warning effectiveness of various rockburst early warning systems remain at the level of calculating the early warning accuracy rate, i.e., the ratio of the total number of accurate rockburst level warnings to the total number of warnings. While this evaluation method is convenient to use, a single evaluation index cannot assess the early warning effectiveness of a rockburst early warning system at a deeper level and is insufficient to reveal the true effectiveness of the system. To further deepen the evaluation of rockburst early warning effectiveness, it is necessary to explore more comprehensive evaluation methods. Summary of the Invention

[0005] This invention addresses the technical problems existing in the prior art by providing a method for evaluating the cumulative effect of rockburst throughout its entire cycle and for visualizing the effect of a single early warning.

[0006] According to a first aspect of the present invention, a method for evaluating the cumulative early warning effect of rockburst throughout its entire lifecycle is provided, comprising:

[0007] Acquire rockburst level data during the warning period, the rockburst level data including the warning rockburst level data during the warning period and the corresponding actual rockburst level data at the scene;

[0008] Based on the rockburst level data during the warning period, a rockburst full-cycle cumulative warning effect evaluation matrix is ​​generated. The rockburst full-cycle cumulative warning effect evaluation matrix is ​​an evaluation matrix that represents the superposition of daily warning rockburst level data and actual rockburst level data during the warning period. The actual rockburst level recorded on site is listed as the warning rockburst level of the system.

[0009] Based on the cumulative early warning effect evaluation matrix for the entire rockburst cycle, calculate the early warning F1 score for each rockburst level;

[0010] Machine learning methods were used to calculate the weight of the quantitative index of rockburst damage degree for each rockburst level;

[0011] The early warning effect score weight for each rockburst level is calculated based on the weight of the quantitative index of rockburst damage degree for each rockburst level.

[0012] Based on the F1 score of each rockburst level and the weight of the warning effect score for each rockburst level, the cumulative warning effect score for the entire rockburst cycle is calculated, and the cumulative warning effect for the entire rockburst cycle is evaluated based on the cumulative warning effect score for the entire rockburst cycle.

[0013] According to a second aspect of the present invention, a method for evaluating the effect of a single early warning of rockburst throughout its entire lifecycle is provided, comprising:

[0014] A grid array for evaluating the effectiveness of a single warning is generated based on the daily warning rockburst level data and the daily actual rockburst level data. The grid array for evaluating the effectiveness of a single warning represents the relationship between the daily warning rockburst level and the daily actual rockburst level.

[0015] Based on the grid array used for evaluating the single-time early warning effect, a single-time early warning effect score for the entire rockburst cycle is generated. This score includes the actual rockburst level for the day and the daily early warning deviation, expressed as:

[0016] Rockburst full-cycle single warning effectiveness score = [actual rockburst level on the day, single-day warning deviation];

[0017] The actual rockburst level on that day is obtained from feedback from the construction site. The daily warning deviation is the distance from the grid data of that day to the diagonal grid data of the same row. The diagonal grid data of the same row is the grid data when the warning rockburst level and the actual rockburst level are equal.

[0018] According to a third aspect of the present invention, a visualization method for evaluating the effectiveness of rockburst early warning is provided, comprising:

[0019] The evaluation matrix for the cumulative early warning effect of rockburst throughout its entire cycle is generated based on the evaluation method for the cumulative early warning effect of rockburst throughout its entire cycle, and the grid array for the evaluation of the single early warning effect is generated based on the evaluation method for the single early warning effect of rockburst throughout its entire cycle.

[0020] Based on the cumulative early warning effect evaluation matrix of the entire rockburst cycle and the grid array of the single early warning effect evaluation, a multidimensional rockburst early warning heat map is generated, and the multidimensional rockburst early warning heat map is visualized.

[0021] This invention provides a method and visualization method for evaluating the cumulative effect of rockburst throughout its entire lifecycle and for single-event early warning. By calculating F1 scores for different rockburst levels and assigning weights based on the severity of each rockburst level, the evaluation method is given different sensitivities to changes in each rockburst level. This method scientifically assesses the overall performance of the early warning system. Especially when rockburst data is unbalanced, this independent calculation and weighting design effectively avoids the bias of traditional accuracy statistics, preventing high-frequency early warning results for low-level rockbursts from masking the early warning system's effectiveness for high-level rockbursts. Attached Figure Description

[0022] Figure 1 A flowchart of a method for evaluating the cumulative early warning effect of rockburst throughout its entire life cycle, provided by this invention;

[0023] Figure 2 This is a schematic diagram of rockburst severity data;

[0024] Figure 3 A schematic diagram of a flowchart for evaluating the effect of a single early warning of rockburst throughout its entire cycle, provided by this invention;

[0025] Figure 4 A schematic diagram of a grid array for evaluating the effectiveness of a single early warning;

[0026] Figure 5 A schematic diagram of the grid array and early warning deviation for evaluating the effectiveness of a single early warning;

[0027] Figure 6 A flowchart illustrating a visualization method for evaluating the effectiveness of rockburst early warning provided by this invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. In addition, the technical features of the various embodiments or individual embodiments provided by the present invention can be arbitrarily combined with each other to form feasible technical solutions. Such combinations are not constrained by the order of steps and / or structural composition patterns, but must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0029] For rockburst early warning, using traditional accuracy statistics has the following shortcomings:

[0030] 1. Inability to distinguish between different rockburst levels: Different levels of rockbursts cause accidents with varying degrees of harm to the site. Accuracy rate alone cannot reflect the difference in early warning effectiveness between high-level and low-level rockbursts. For example, the consequences of a false warning for a low-level rockburst are different from those for a high-level rockburst, but accuracy rate calculations alone do not distinguish between these differences.

[0031] 2. Ignoring the impact of false alarms and missed alarms: Accuracy only focuses on the number of correct warnings, ignoring false alarms (i.e., the system's warning of a rockburst level is higher than the actual rockburst level) and missed alarms (i.e., the system's warning of a rockburst level is lower than the actual rockburst level). A high false alarm rate leads to wasted resources and uncertainty in on-site decision-making, while missed alarms pose safety risks. Accuracy alone cannot quantify these impacts.

[0032] 3. Ignoring the degree of deviation in early warning: Accuracy rates only record whether the prediction is correct, without considering the degree of deviation between the predicted and actual results. For example, if the system predicts a minor rockburst but the actual rockburst is a moderate one, the impact of this deviation is not reflected in the accuracy rate. This neglect may lead to a lack of precision in the system's early warning levels.

[0033] 4. Poor performance in evaluating unbalanced data: If there are many low-level events and few high-level events, the accurate prediction of a large number of low-level rockbursts will overshadow the inaccurate prediction of high-level rockbursts. Relying solely on accuracy can easily lead to misleading evaluations due to the correct prediction of most events.

[0034] 5. Difficulty in reflecting overall model performance: The accuracy rate of a single calculation-based early warning system cannot provide comprehensive information about the model. More comprehensive indicators such as precision, recall, and F1 score can quantify the system's performance in different aspects, while accuracy alone is insufficient to identify the system's actual performance.

[0035] Based on this, see Figure 1 This invention provides a method for evaluating the cumulative early warning effect of rockburst throughout its entire lifecycle, which mainly includes the following steps:

[0036] Step 1: Obtain rockburst level data during the warning period. The rockburst level data includes the rockburst level data during the warning period and the corresponding actual rockburst level data at the site.

[0037] Understandably, for the warning period requiring rockburst alerts, rockburst level data for that period is obtained. This data includes daily predicted rockburst level data and daily actual rockburst level data. Predicted rockburst level data refers to the rockburst level data predicted by the rockburst warning system, while actual rockburst level data refers to the rockburst level data determined on-site based on the damage caused by rockbursts. The rockburst level is categorized into five levels: no rockburst, minor rockburst, moderate rockburst, severe rockburst, and extremely severe rockburst. Rockburst level data for a given period is shown in Table 1; the accuracy and reliability of this data need to be verified.

[0038] Table 1 Rockburst Level Data

[0039]

[0040] Step 2: Based on the rockburst level data during the warning period, generate a rockburst full-cycle cumulative warning effect evaluation matrix. The rockburst full-cycle cumulative warning effect evaluation matrix is ​​an evaluation matrix that represents the superposition of daily warning rockburst level data and actual rockburst level data during the warning period. The actual rockburst level recorded on-site is listed as the warning rockburst level of the system.

[0041] Understandably, based on the rockburst level data within the warning period in Table 1 above, a cumulative rockburst warning effect evaluation matrix is ​​generated. The rows of the cumulative rockburst warning effect evaluation matrix represent the actual rockburst level on site, and the columns represent the warning rockburst level. Since rockburst levels are divided into five levels—no rockburst, minor rockburst, moderate rockburst, severe rockburst, and extremely severe rockburst—the five rows i=1, 2, 3, 4, 5 and the five columns j=1, 2, 3, 4, 5 of the cumulative rockburst warning effect evaluation matrix represent the five rockburst levels: no rockburst, minor rockburst, moderate rockburst, severe rockburst, and extremely severe rockburst, respectively. Elements of the cumulative rockburst warning effect evaluation matrix. As a superposition of the actual daily rockburst level and the daily rockburst warning level, i represents the actual daily rockburst level, j represents the daily rockburst warning level, and the elements in the evaluation matrix are... This represents the total number of times a rockburst of level i is actually recorded on a single day, and the warning level is j. For example, if the rockburst warning level is minor and the actual rockburst level is moderate, then the count in the 2nd row and 3rd column of the matrix is ​​1 (1 represents the number of times). Based on this, a 5×5 rockburst warning effect evaluation matrix is ​​generated by overlaying rockburst data over a period of time. The generated cumulative rockburst warning effect evaluation matrix for the entire cycle can be found in [reference needed]. Figure 2 .

[0042] Step 3: Calculate the early warning F1 score for each rockburst level based on the cumulative early warning effect evaluation matrix for the entire rockburst cycle.

[0043] The calculation of the F1 score for each rockburst level includes: calculating the warning precision rate and warning recall rate for each rockburst level, and then harmonizing the warning precision rate and warning recall rate for each rockburst level to obtain the warning F1 score for each rockburst level. The warning precision rate refers to the proportion of rockburst levels that are actually predicted to occur, and the warning recall rate refers to the proportion of actual rockburst levels that are predicted to occur.

[0044] Specifically, the F1 score for each rockburst level is the harmonic mean of the precision and recall rates for that level. Prediction precision is the proportion of rockbursts of a given level that were predicted to actually occur, while prediction recall is the proportion of rockbursts of a given level that were predicted to actually occur.

[0045] Elements of the 5×5 Rockburst Full-Cycle Cumulative Early Warning Effect Evaluation Matrix This indicates a superposition of actual rockburst level i and warning rockburst level j (i, j=1, 2, 3, 4, 5 correspond to no rockburst, slight rockburst, moderate rockburst, strong rockburst, and extremely strong rockburst, respectively).

[0046] The F1 score for rockburst level i is: ;

[0047] The accuracy rate of the early warning for rockburst level i is: ;

[0048] The recall rate for rockburst level i warnings is:

[0049] ;

[0050] in, The warning F1 score for rockburst level i, For rockburst level i, the early warning accuracy is... This represents the early warning summoning rate for Rockburst level i.

[0051] The warning F1 score for each rockburst level is calculated using the above calculation method, as shown in Table 2 below.

[0052] Table 2. Warning F1 Score for Each Rockburst Level

[0053]

[0054] Step 4: Use machine learning methods to calculate the weight of the quantitative index of rockburst damage degree for each rockburst level.

[0055] The quantitative indicators for the degree of rockburst damage for each rockburst level include post-rockburst cleanup time, rockburst crater depth, rockburst crater axis length, and initial velocity of rockburst fragments. The weights of these quantitative indicators for each rockburst level are calculated using machine learning methods, including:

[0056] Determine the numerical range of the quantitative indicators of rockburst damage degree for each rockburst level, and generate the median of four quantitative indicators of rockburst damage degree for each early warning rockburst level: post-rockburst cleanup time, rockburst crater depth, rockburst crater axis length, and initial velocity of rockburst fragments.

[0057] A random forest nonlinear prediction model was established, with the median of four quantitative indicators of rockburst damage as input and the early warning rockburst level as output.

[0058] The weights of each rockburst damage index corresponding to each rockburst level are calculated based on the established random forest nonlinear prediction model.

[0059] Specifically, the quantitative indicators for rockburst damage are selected based on the actual destructive effects of rockburst disasters to characterize the impact of rockbursts on the engineering construction environment. The selected indicators also need to be easily quantifiable. In this embodiment of the invention, the following parameters are selected as the quantitative indicators for rockburst damage: post-rockburst cleanup time (in hours), rockburst crater depth (in meters), rockburst crater axis length (in meters), and initial velocity of rockburst fragments (in meters per second). The numerical range of the quantitative indicators for each rockburst level is also determined, as shown in Table 3 below.

[0060] Table 3 Quantitative Indicators of Rockburst Damage

[0061]

[0062] The weights of rockburst damage quantification indicators are obtained using a random forest algorithm-based rockburst severity prediction model. Predicting rockburst severity using these indicators involves handling multi-feature nonlinear relationships, and random forests are well-suited for such tasks; therefore, a random forest model is used to build the prediction model. First, the medians of four rockburst damage quantification indicators—post-rockburst cleanup time (h), rockburst crater depth (m), rockburst crater axis length (m), and initial rockburst fragment velocity (m / s)—are calculated. Then, a random forest nonlinear prediction model is built, using the medians of the four rockburst damage quantification indicators as input and the predicted rockburst severity as output. The feature importance of the four rockburst damage quantification indicators in the random forest prediction model is calculated as the weights of the rockburst damage quantification indicators. Random forests indirectly calculate feature importance by statistically analyzing the effect of split nodes. During the construction of the decision tree, each feature is used as a split node (i.e., a feature and a threshold are selected to split the data). Each time a feature is used for splitting, the impurity of the target variable (e.g., rockburst severity) is reduced. Feature importance is calculated based on the reduction in impurity. Because a random forest contains many trees, each tree contributes differently to feature importance. The overall importance of a feature is obtained by weighted averaging of the impurity reduction for each feature across all trees. The final feature importance is a relative value, representing the contribution of a feature to the model's prediction relative to the contributions of other features. The weights of each damage index are obtained by summing and normalizing the importance of all features. To ensure the scientific validity of the weight calculation, multiple sets of randomly generated data within the range of the rockburst damage index can be used, divided into training and test sets, for result verification. The final weights need to be normalized; the normalized weights of the rockburst damage index are shown in Table 4.

[0063] Table 4 Weights of Quantitative Indicators for Rockburst Damage Degree

[0064]

[0065] Step 5: Calculate the early warning effect score weight for each rockburst level based on the quantitative index weight of the rockburst damage degree for each rockburst level.

[0066] Specifically, after obtaining the weights of the rockburst damage quantification index in rockburst level prediction based on step 4, these weights are multiplied by the median of the damage quantification indexes for minor, moderate, severe, and extremely severe rockbursts to calculate the corresponding rockburst level score weights. Since no rockburst will cause damage on-site, there is no rockburst damage quantification index; therefore, the initial score weight for no rockburst needs to be set with reference to the score weight for minor rockburst. Finally, the score weights for the five rockburst levels are normalized to obtain the final score weights.

[0067] The weighting of the early warning effect score for rockburst level i is as follows: ;

[0068] Median post-rockburst cleanup time for rockburst level i: ;

[0069] Median crater depth for rockburst class i: ;

[0070] Median crater length for rockburst grade i: ;

[0071] Median initial velocity of rockburst fragments for rockburst grade i: ;

[0072] Where i=1, 2, 3, 4, 5 correspond to no rockburst, slight rockburst, moderate rockburst, strong rockburst, and extremely strong rockburst; The weighting for the early warning effect score of rockburst level i; The median weight of post-rockburst cleanup time is used as the weighting factor in the quantitative index of rockburst damage severity. The weight of rockburst crater depth is used as the weight in the quantitative index of rockburst damage degree. The weight of the rockburst crater axis length in the weighting of the quantitative index of rockburst damage degree; The initial velocity weight of rockburst fragments is used as the weight in the quantitative index of rockburst damage degree. The median cleanup time after a rockburst for rockburst class i; The median depth of the rockburst crater is i; is the median length of the rockburst crater axis for rockburst grade i; Let i be the median initial velocity of rockburst fragments for rockburst grade i. and These are the minimum and maximum cleanup times after a rockburst, respectively, for rockburst level i. and These are the minimum and maximum values ​​of the rockburst crater depth for rockburst level i, respectively. and These are the minimum and maximum values ​​of the axial length of the rockburst crater for rockburst grade i, respectively. and These are the minimum and maximum initial velocities of rockburst fragments for rockburst grade i, respectively.

[0073] The initial weights for the early warning effect score of rockburst level can be found in Table 5, and the final weights after normalization can be found in Table 6.

[0074] Table 5 Initial Weights for Rockburst Early Warning Effectiveness Score

[0075]

[0076] Table 6

[0077]

[0078] Step 6: Calculate the cumulative early warning effect score for the entire rockburst cycle based on the early warning F1 score for each rockburst level and the early warning effect score weight for each rockburst level. Evaluate the cumulative early warning effect for the entire rockburst cycle based on the cumulative early warning effect score for the entire rockburst cycle.

[0079] Understandably, the early warning F1 score for each rockburst level and its corresponding early warning effect score weight are weighted and summed to generate the cumulative early warning effect score for the entire rockburst cycle. The cumulative early warning effect score for the entire rockburst cycle is the sum of the weighted early warning F1 scores for each rockburst level. It should be noted that if a certain level of rockburst in the early warning data has never actually occurred, the calculation portion of that level of rockburst should be removed, and the rockburst levels that actually occurred should be normalized again according to the proportional relationship of their respective early warning effect score weights before calculation. For example, in Example 1, only no rockburst and minor rockburst actually occurred, so the proportional relationship is calculated according to their corresponding early warning effect score weights of 0.01 and 0.03, and then normalized again, resulting in early warning effect score weights of 0.25 and 0.75 for no rockburst and minor rockburst, respectively.

[0080] The evaluation index for the cumulative early warning effect of rockburst early warning throughout the entire cycle is as follows:

[0081] The cumulative early warning effectiveness score is calculated for the entire lifecycle of rockburst early warning. Weighting of the early warning effect score for rockburst level i; The warning F1 score is the score for rockburst level i.

[0082] The calculation results of the cumulative early warning effect score for rockburst early warning throughout the entire cycle of one embodiment:

[0083] .

[0084] The above-mentioned evaluation method for the full-cycle early warning effect of rockburst is described in [reference needed]. Figure 3 The method for evaluating the effect of a single early warning of rockburst throughout its entire cycle, as described in this invention, mainly includes the following steps:

[0085] Step 1': Generate a grid array for evaluating the effect of a single warning based on the daily warning rockburst level data and the daily actual rockburst level data. The grid array for evaluating the effect of a single warning represents the relationship between the daily warning rockburst level and the daily actual rockburst level.

[0086] Understandably, it is necessary to collect daily rockburst warning data and daily actual rockburst data. For example, if the rockburst warning level is severe rockburst on a certain day, but the actual rockburst level is minor rockburst, the daily rockburst level data can be found in Table 7.

[0087] Table 7 Daily Rockburst Level Data

[0088]

[0089] A grid array for evaluating the effectiveness of a single early warning is generated based on daily early warning rockburst level data and actual rockburst level data. Table 7 shows an early warning rockburst level of "strong rockburst" and an actual rockburst level of "minor rockburst," resulting in the generated grid array for evaluating the effectiveness of a single early warning. This grid array consists of 25 grid points in 5 rows and 5 columns. The 5 rows i=1, 2, 3, 4, 5 represent the five levels of actual rockburst, and the 5 columns j=1, 2, 3, 4, 5 represent the five levels of early warning rockburst.

[0090] According to the composition rules of the grid array, the diagonal grid points of the grid array used to evaluate the effectiveness of a single early warning are those where the warning rockburst level is equal to the actual rockburst level, represented by the first graphic symbol; the lower left triangular grid points are those where the warning rockburst level is lower than the actual rockburst level, represented by the second graphic symbol; and the upper right triangular grid points are those where the warning rockburst level is higher than the actual rockburst level, represented by the third graphic symbol. The grid data for the day is a combination of the actual rockburst level and the rockburst warning level for the day. That is, i represents the actual rockburst level for the day and j represents the rockburst warning level for the day, which is represented by the fourth graphic symbol.

[0091] Specifically, in the single-warning effect evaluation grid array, the diagonal grid points represent warning rockburst levels equal to the actual rockburst levels, denoted by ○. The lower left triangular grid points represent warning rockburst levels lower than the actual rockburst levels, denoted by ▽. The upper right triangular grid points represent warning rockburst levels higher than the actual rockburst levels, denoted by △. Grid points The rockburst level is a combination of the actual daily rockburst level and the daily rockburst warning level, where i represents the actual daily rockburst level and j represents the daily rockburst warning level, denoted by ●. For a grid array used to evaluate the effectiveness of a single warning, please refer to [reference needed]. Figure 3 .

[0092] Step 2': Generate a single rockburst warning effect score for the entire cycle based on the single warning effect evaluation grid array. The single rockburst warning effect score for the entire cycle includes the actual rockburst level on the day and the single warning deviation.

[0093] Understandably, the score for the effectiveness of a single early warning throughout the rockburst cycle is composed of the actual rockburst level on that day and the daily early warning deviation, i.e.: the score for the effectiveness of a single early warning throughout the rockburst cycle = [the actual rockburst level on that day, and the daily early warning deviation].

[0094] The actual rockburst level for the day is obtained from feedback at the construction site. The daily warning deviation is the distance from the grid point data of that day to the diagonal grid point of the same row (the distance between adjacent grid points is 1), that is, the distance from grid point ● to grid point ○ in the same row. If grid point ● is to the right of grid point ○, the daily warning deviation is positive, indicating that the warned rockburst level for that day is higher than the actual rockburst level. If grid point ● is to the left of grid point ○, the daily warning deviation is negative, indicating that the warned rockburst level for that day is lower than the actual rockburst level. A schematic diagram of the grid array and warning deviation for evaluating the effect of a single warning can be found in [reference needed]. Figure 4 , Figure 4 In the medium range, the single-time early warning effect score for rockburst throughout the entire cycle is: [Minor rockburst, +2].

[0095] See Figure 5 This invention provides a visualization method for evaluating the effectiveness of rockburst early warning, comprising:

[0096] Step 1” generates a rockburst full-cycle cumulative early warning effect evaluation matrix based on the rockburst full-cycle cumulative early warning effect evaluation method, and generates a single early warning effect evaluation grid array based on the rockburst early warning full-cycle single early warning effect evaluation method.

[0097] Step 2”: Generate a multi-dimensional rockburst early warning heat map based on the cumulative early warning effect evaluation matrix of the entire rockburst cycle and the grid array of the single early warning effect evaluation, and visualize the multi-dimensional rockburst early warning heat map.

[0098] Understandably, based on the aforementioned method for evaluating the full-cycle early warning effect of rockburst, a cumulative early warning effect evaluation matrix for the full-cycle rockburst can be obtained, as well as a grid array for evaluating the single early warning effect of rockburst based on the aforementioned method for evaluating the single early warning effect of rockburst. Furthermore, multiple rockburst early warning heatmaps can be generated based on the cumulative evaluation matrix for the full-cycle rockburst and the grid array for evaluating the single early warning effect, which can more intuitively evaluate the rockburst early warning effect.

[0099] This invention provides a method and visualization method for evaluating the cumulative effect of rockburst throughout its entire cycle and for single-event early warning, which has the following beneficial effects:

[0100] (1) Considering the differences in the hazards of different levels of rockburst

[0101] Traditional accuracy statistics primarily focus on overall prediction accuracy, failing to account for the varying degrees of harm caused by false alarms and missed alarms at different rockburst levels. However, by calculating F1 scores separately for different accident levels and assigning weights based on the severity of each rockburst level, the evaluation method gains a sensitivity to changes in each rockburst level. This method scientifically assesses the overall model performance, especially in cases of imbalanced accident data. This independent calculation and weighting design effectively avoids the biases of traditional accuracy statistics, preventing the high-frequency warning effect of low-level rockbursts from masking the system's warning effect for high-level rockbursts.

[0102] (2) Dynamic weight design based on rockburst damage degree

[0103] The evaluation method assigns corresponding weights to different levels of rockbursts based on quantitative indicators of rockburst damage severity (post-accident cleanup time, crater depth, crater axis length, and initial fragment velocity), ensuring that the evaluation method pays attention to the corresponding levels of rockbursts. Compared to simply calculating the accuracy of early warnings, traditional accuracy methods do not consider the impact of the severity of different levels of rockbursts on engineering projects. The weighted design allows the evaluation method to place greater emphasis on the accurate early warning effect of severe accidents, which is more important in actual engineering risk management.

[0104] (3) Data-driven weighted mechanism

[0105] By employing machine learning algorithms (random forests) to automatically calculate the weights of each rockburst destructive index, the weight design becomes more objective. Each random forest algorithm calculates the importance of each destructive index in the prediction process using the four rockburst destructive indices, thus representing the weight of each destructive index in each rockburst level. Using a data-driven approach also allows the evaluation method to adjust the calculated weights in real time according to actual engineering conditions, ensuring the applicability of the evaluation method across different engineering projects and significantly improving the scientific rigor and stability of the model evaluation.

[0106] (4) Data visualization

[0107] A heatmap using a multidimensional confusion matrix can show the comparison between actual accident levels and predicted levels. The intensity of the color directly reflects the distribution of correct and incorrect warnings, helping to analyze the distribution of warnings at different accident levels, displaying the prediction situation for each level, and ensuring clear information delivery.

[0108] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0109] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0110] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0111] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0112] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0113] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0114] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for evaluating the cumulative early warning effect of rockburst throughout its entire lifecycle, characterized in that, include: Acquire rockburst level data during the warning period, the rockburst level data including the warning rockburst level data during the warning period and the corresponding actual rockburst level data at the scene; Based on the rockburst level data during the warning period, a rockburst full-cycle cumulative warning effect evaluation matrix is ​​generated. The rockburst full-cycle cumulative warning effect evaluation matrix is ​​an evaluation matrix that represents the superposition of daily warning rockburst level data and actual rockburst level data during the warning period. The actual rockburst level recorded on site is listed as the warning rockburst level of the system. Based on the cumulative early warning effect evaluation matrix for the entire rockburst cycle, calculate the early warning F1 score for each rockburst level; Machine learning methods were used to calculate the weight of the quantitative index of rockburst damage degree for each rockburst level; The early warning effect score weight for each rockburst level is calculated based on the weight of the quantitative index of rockburst damage degree for each rockburst level. Based on the F1 score of each rockburst level and the weight of the warning effect score for each rockburst level, the cumulative warning effect score for the entire rockburst cycle is calculated, and the cumulative warning effect for the entire rockburst cycle is evaluated based on the cumulative warning effect score for the entire rockburst cycle. The calculation of the early warning effect score weight for each rockburst level based on the quantitative index weight of rockburst damage degree for each rockburst level includes: The weight of each rockburst damage degree quantitative index is multiplied by the median of the damage degree quantitative index corresponding to each rockburst level to calculate the early warning effect score weight of each rockburst level. The weighting of the early warning effect score for rockburst level i is as follows: ; Median post-rockburst cleanup time for rockburst level i: ; Median crater depth for rockburst class i: ; Median crater length for rockburst grade i: ; Median initial velocity of rockburst fragments for rockburst class i: ; Where i=1, 2, 3, 4, 5 correspond to no rockburst, slight rockburst, moderate rockburst, strong rockburst, and extremely strong rockburst; The weighting of the early warning effect score for rockburst level i; The median weight of post-rockburst cleanup time is used as the weighting factor in the quantitative index of rockburst damage severity. The weight of rockburst crater depth is used as the weight in the quantitative index of rockburst damage degree. The weight of the rockburst crater axis length in the weighting of the quantitative index of rockburst damage degree; The initial velocity weight of rockburst fragments is used as the weight in the quantitative index of rockburst damage degree. The median cleanup time after a rockburst for rockburst class i; The median depth of the rockburst crater is i; is the median length of the rockburst crater axis for rockburst grade i; Let i be the median initial velocity of rockburst fragments for rockburst grade i. and These are the minimum and maximum cleanup times after a rockburst, respectively, for rockburst level i. and These are the minimum and maximum values ​​of the rockburst crater depth for rockburst level i, respectively. and These are the minimum and maximum values ​​of the axial length of the rockburst crater for rockburst grade i, respectively. and These are the minimum and maximum initial velocities of rockburst fragments for rockburst grade i, respectively.

2. The method for evaluating the cumulative early warning effect of rockburst throughout its entire lifecycle according to claim 1, characterized in that, The rockburst severity level is divided into five levels: no rockburst, minor rockburst, moderate rockburst, severe rockburst, and extremely severe rockburst. The elements of the rockburst full-cycle cumulative early warning effect evaluation matrix... This represents the total number of times that the actual rockburst level is i and the warning rockburst level is j on a single day.

3. The method for evaluating the cumulative early warning effect of rockburst throughout its entire lifecycle according to claim 2, characterized in that, Based on the aforementioned cumulative early warning effect evaluation matrix for the entire rockburst cycle, the early warning F1 score for each rockburst level is calculated, including: Calculate the warning precision rate and warning recall rate for each rockburst level separately. Then, harmonic average the warning precision rate and warning recall rate for each rockburst level to obtain the warning F1 score for each rockburst level. The warning precision rate refers to the proportion of rockburst levels that are actually predicted to occur, and the warning recall rate refers to the proportion of actual rockburst levels that are predicted to occur.

4. The method for evaluating the cumulative early warning effect of rockburst throughout its entire lifecycle according to claim 3, characterized in that, The method of harmonic averaging the warning precision and warning recall rates for each rockburst level to obtain the warning F1 score for each rockburst level includes: ； The accuracy rate of the early warning for rockburst level i is: ; The recall rate for rockburst level i warnings is: ； in, This represents the F1 score for the warning level i of rockburst. For rockburst level i, the early warning accuracy is... denoted as the early warning recall rate for rockburst level i.

5. The method for evaluating the cumulative early warning effect of rockburst throughout its entire lifecycle according to claim 1, characterized in that, The quantitative indicators of rockburst damage severity for each rockburst level include post-rockburst cleanup time, rockburst crater depth, rockburst crater axis length, and initial velocity of rockburst fragments. The weights of these quantitative indicators for each rockburst level are calculated using machine learning methods, including: Determine the numerical range of the quantitative indicators of rockburst damage degree for each rockburst level, and generate the median of four quantitative indicators of rockburst damage degree for each early warning rockburst level: post-rockburst cleanup time, rockburst crater depth, rockburst crater axis length, and initial velocity of rockburst fragments. A random forest nonlinear prediction model was established, with the median of four quantitative indicators of rockburst damage as input and the early warning rockburst level as output. The weights of each rockburst damage index corresponding to each rockburst level are calculated based on the established random forest nonlinear prediction model.

6. The method for evaluating the cumulative early warning effect of rockburst throughout its entire lifecycle according to claim 1, characterized in that, The cumulative early warning effect score for the entire rockburst cycle is calculated based on the early warning F1 score for each rockburst level and the early warning effect score weight for each rockburst level, including: The cumulative early warning effect score for the entire rockburst cycle is obtained by weighting and summing the early warning F1 score for each rockburst level and the early warning effect score for each rockburst level.

7. A visualization method for evaluating the effectiveness of rockburst early warning, characterized in that, include: The method for evaluating the cumulative early warning effect of rockburst throughout its entire cycle, as described in claim 1, generates an evaluation matrix for the cumulative early warning effect of rockburst throughout its entire cycle, and generates a grid array for evaluating the single early warning effect based on the method for evaluating the single early warning effect of rockburst throughout its entire cycle. Based on the cumulative early warning effect evaluation matrix of the entire rockburst cycle and the grid array of the single early warning effect evaluation, a multidimensional rockburst early warning heat map is generated, and the multidimensional rockburst early warning heat map is visualized. The method for evaluating the effectiveness of a single rockburst early warning throughout its entire lifecycle includes: A grid array for evaluating the effectiveness of a single warning is generated based on the daily warning rockburst level data and the daily actual rockburst level data. The grid array for evaluating the effectiveness of a single warning represents the relationship between the daily warning rockburst level and the daily actual rockburst level. Based on the grid array used for evaluating the single-time early warning effect, a single-time early warning effect score for the entire rockburst cycle is generated. This score includes the actual rockburst level for the day and the daily early warning deviation, expressed as: Rockburst full-cycle single warning effectiveness score = [actual rockburst level on the day, single-day warning deviation]; The actual rockburst level on that day is obtained from feedback from the construction site. The daily warning deviation is the distance from the grid data of that day to the diagonal grid data of the same row. The diagonal grid data of the same row is the grid data when the warning rockburst level and the actual rockburst level are equal.

Images