Intelligent early warning method for stability of surrounding rock of deep coal seam roadway

Abstract

The invention provides a deep coal seam roadway surrounding rock stability intelligent early warning method, and relates to the technical field of coal mine dynamic disaster prevention and control. According to the intelligent early warning method for the stability of the surrounding rock of the deep coal seam roadway, a basic sample library is constructed through a dynamic and static combined loading rock test; synchronously collecting field surrounding rock static stress and micro-seismic data and carrying out time sequence polymerization; constructing an equivalent dynamic and static combined stress index based on a dynamic and static load superposition principle, and forming a seismic-force coupling feature vector; establishing an LSTM early warning model, adopting a transfer learning strategy, pre-training by using laboratory data, and then performing fine tuning by using field data; inputting the real-time feature vector into a model and outputting an instability probability; and verifying and correcting the probability based on actual deformation data of the surrounding rock, and applying a missing report sample to online incremental updating of the model. According to the method, the dynamic and static load coupling effect is quantified, the difficulty of small sample training is overcome, and accurate and self-adaptive early warning of the deep roadway surrounding rock instability risk is realized through a physical feedback closed-loop correction mechanism.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of coal mine dynamic disaster prevention, in particular to a deep coal seam roadway surrounding rock stability intelligent early warning method based on microseism-stress space-time evolution characteristics. BACKGROUND

[0002] With the continuous increase of coal resource mining depth, the geomechanical environment faced by deep coal seam roadway is increasingly complex, showing significant "high ground stress" and "strong mining disturbance" characteristics. In the deep high stress environment, the stability of the roadway surrounding rock is not only affected by static geological conditions, but also affected by dynamic load disturbance induced by roof fracture and other dynamic events. The nonlinear superposition of such dynamic and static loads is the main cause of dynamic disasters such as rock burst. At present, using deep learning and multi-physical field monitoring technology for early warning of rock burst and other dynamic disasters has become a research hotspot, but the existing technology still has the following shortcomings:

[0003] First, the existing method inputs microseismic and stress data as independent channels, ignoring the nonlinear disaster-causing mechanism of dynamic and static load superposition, and it is difficult to quantify the weakening effect of dynamic load on the static load bearing limit; second, the sample of rock burst and other catastrophic events is scarce, which makes the data-driven model difficult to train and has poor generalization ability; third, the existing early warning system lacks a closed-loop correction mechanism based on physical deformation feedback, which cannot real-time verify the prediction results and update the model online, resulting in a decrease in robustness over time.

[0004] For example, the Chinese patent document with publication number CN117312919A discloses a rock burst early warning method based on a BO-CNN-LSTM model, but it is a pure data-driven model lacking physical mechanism constraints, and does not consider the dynamic and static load coupling effect. The Chinese patent document with publication number CN111599137A discloses a multi-physical field monitoring and early warning system and method for underground engineering surrounding rock stability, which although integrates multi-physical field data, the feature fusion method is simple and lacks a special physical index construction and closed-loop correction mechanism. SUMMARY

[0005] The present application provides a deep coal seam roadway surrounding rock stability intelligent early warning method, which constructs a feature vector reflecting the superposition effect of dynamic and static loads, combines transfer learning and LSTM network, realizes accurate identification of rock burst risk of deep coal seam roadway surrounding rock, and introduces physical deformation feedback for closed-loop correction.

[0006] In order to achieve the above purpose, the technical solution adopted by the present application is as follows: A deep coal seam roadway surrounding rock stability intelligent early warning method, the method comprising the following steps: S1, constructing a basic sample library containing input feature vectors and stability labels through dynamic and static combined loading rock test; S2, synchronously collect the static stress of the field surrounding rock and the microseismic data, and map to a continuous time window for timing aggregation; S3, standardize the aggregated data, and construct a vibration-force coupling feature vector based on the dynamic and static load superposition principle; S4, establish a warning model based on a long short-term memory network; S5, adopt a transfer learning strategy, pre-train the model based on the S1 basic sample library, fine-tune using the S2 field collected data, and verify using the test set; S6, input the vibration-force coupling feature vector after real-time collection and processing into the model, and output the instability probability; S7, check and correct the instability probability based on the actual deformation data of the surrounding rock, and add the missed sample to the difficult sample library for online updating of the model.

[0007] Further, the S1 comprises: S11, drill hard rock and soft rock samples in the deep coal seam roadway surrounding rock; S12, process the drilled rock samples into rock test pieces of standard size; S13, use a rock triaxial pressure loading test machine and a pendulum impact test machine for combined loading, respectively applying static axial stress simulating the original rock stress and dynamic impact simulating vibration; S14, record the rock failure state under different combined working conditions, and construct a basic sample library : ; Wherein, is the input feature of the ith test, including the applied static axial stress value and dynamic impact energy value; is a stability label, which is marked as 0 if the rock test piece remains intact, and marked as 1 if the rock test piece fails; K is the total number of test samples, that is, the sum of the effective test times under different dynamic and static combined loading conditions.

[0008] Further, the S2 comprises: S21, interval arrangement of measuring stations in the key monitoring area of the deep roadway, each measuring station is equipped with a deep hole multi-point displacement meter, an anchor cable dynamometer and a microseismic pick-up array; S22, real-time collection of anchor / cable force sequence σ(t), and through the elastic wave signals captured by the microseismic pick-up array due to rock mass micro-fracture, the energy sequence E(t) of the microseismic event is solved; S23, map the discrete microseismic data to a continuous time window , define the first Polymer stress within a time window and aggregate energy for: ; ; in, For time window The number of stress sampling points within; For time window The total number of microseismic events that occurred within the area.

[0009] Further, S3 includes: S31. Clean the collected data, fill in missing data and remove outliers and noise. S32. Standardize each physical quantity using the Z-Score method. The standardized values ​​are... : ; in, This is the original monitoring data; The mean of the feature data; Standard deviation; S33. Construct an equivalent dynamic-static combined stress index with lithology self-adaptive capability. The calculation formula is as follows: ; in, The static stress aggregation value of the surrounding rock monitored at time t; E0 represents the aggregated energy released by the microseismic event at time t; E0 is the energy normalization baseline value. is the dynamic-static load dimension conversion coefficient, whose value is positively correlated with the impact tendency index of the surrounding rock of the roadway; ln is the natural logarithm function; S34. Calculate the first-order difference characteristic of the index. : ; in, The equivalent dynamic and static combined stress index aggregate value after standardization within the k-th time window; The equivalent dynamic and static combined stress index aggregate value after standardization within the (k-1)th time window; To represent the time interval between two consecutive time windows; S35. Constructing the model input feature vector : ; in, This represents data that has undergone standardization. The normalized static stress aggregate value of the surrounding rock at time k; The standardized aggregate energy value of the microseismic event at time k; The value of the equivalent dynamic-static combined stress index at time k is the standardized value.

[0010] Furthermore, the long short-term memory network early warning model established in S4, for time... input vector The state is updated using a forget gate, an input gate, and an output gate: Computational forget gate: ; in, The output vector of the forget gate has a value between (0,1), representing the proportion of cell state information retained from the previous time step. The closer the value is to 1, the more information is retained; the closer it is to 0, the more information is forgotten. Sigmoid is the activation function. Here is the weight matrix for the forget gate; This represents the hidden layer state from the previous time step. This is the input feature vector at the current time. This is the bias vector for the forget gate; Calculate the input gate: ; ; in, The output vector of the input gate determines what new information will be updated to the cell state at the current moment; The candidate cell state at the current moment is a new vector containing new information created through the tanh layer; tanh is the hyperbolic tangent activation function. and The weight matrices are calculated for the input gate and the candidate state, respectively. and These are the corresponding bias vectors; Perform a status update: = ⊙ + ⊙ ; in, The cell state updated at the current moment, which is calculated by forgetting the old information and adding the new candidate information; The cell state at the previous moment; represents the Hadamard product, that is, the operation of multiplying the corresponding position elements of matrices or vectors; Output the hidden layer features and hidden state at the current moment: ; = ⊙ ; Among them, The output vector of the output gate, which determines which information in the current cell state will be output. The closer the value is to 1, the more output, and the closer the value is to 0, the more shielding; The hidden state at the current moment, that is, the final feature representation filtered by the output gate, which will be passed to the LSTM unit at the next moment; The weight matrix of the output gate, used for weighted sum of input data and hidden state at the previous moment, determines the emphasis of output information; The bias term of the output gate, used to adjust the threshold value of the activation function and improve the fitting ability of the model; Calculate the output layer probability: ; Among them, The probability value of the model predicting that the surrounding rock will lose stability at the current moment t, which ranges between 0 and 1. The closer the value is to 1, the higher the risk of instability, and the closer the value is to 0, the more stable the surrounding rock state; The weight matrix of the output layer, used to map the hidden layer feature vector output by the LSTM to a one-dimensional decision space; The bias term of the output layer; e is the natural constant, the base of the natural logarithm; Further, the training process of the model uses the binary cross entropy loss function and adds the L2 regularization term , the loss function is: ; Among them, The number of batch samples; The true label; The probability predicted by the model; The L2 regularization term, The regularization coefficient; All weight parameters of the network.

[0011] Further, the S5 comprises: S51, pre-training the model by using rock test data loaded through dynamic and static combination, and similarity ratio conversion is required before training; S52, adopting a transfer learning strategy, keeping the weight parameters of the feature extraction layer in the pre-trained model unchanged, and only unfreezing and fine-tuning the parameters of the classification decision layer of the model; S53, calculating the F1 score of the model on the test set, if the F1 score is lower than the set threshold, then return to training after adjusting the hyperparameters.

[0012] Further, the S6 comprises: S61, generating a real-time seismic-force coupling feature vector from the real-time collected static stress of surrounding rock and microseismic data , inputting the early warning model; S62, the model outputs the instability probability value P of the surrounding rock at the current time.

[0013] Further, according to the instability probability value P, three levels of early warning levels are set: P<0.5 is the green safety zone, 0.5≤P<0.8 is the yellow warning zone, and P≥0.8 is the red alarm zone.

[0014] Further, the S7 comprises: S71, obtaining the actual deformation of the surrounding rock in real time through the field multi-point displacement meter , comparing with the set limit displacement threshold , calculating the normalized deformation ratio : ; S72, constructing the final warning probability after correction : ; wherein, is the confidence weight; is the original prediction probability of the model; is the correction linear unit; S73, online evolution of the model: when and , it is determined as "missed report", and the data of this period is added to the difficult example sample library for incremental training; wherein, is the rate threshold.

[0015] Compared with the prior art, the deep coal seam roadway surrounding rock stability intelligent early warning method of the present application has the following beneficial technical effects: 1. The present application realizes the scientific quantification of the dynamic and static load coupling effect by constructing the "equivalent dynamic and static combined stress" index. Traditional early warning methods often process microseismic and stress data independently, which cannot effectively reflect the physical nature of the disaster caused by the superposition of dynamic and static loads. The present application couples the static stress collected on site with the microseismic energy aggregation value to construct a feature vector representing the synergistic effect of dynamic and static loads, thereby transforming the isolated monitoring of the disaster-causing mechanism from multi-source information into direct quantitative characterization of the nonlinear superposition process of "static load bearing foundation" and "dynamic load disturbance weakening". This fundamentally overcomes the defects of the prior art that cannot depict the weakening effect of dynamic load on the bearing limit of surrounding rock, enables the early warning model to have clear physical mechanism constraints, and greatly improves the scientificity and accuracy of the early warning criterion.

[0016] 2. The present application innovatively adopts a hybrid modeling strategy of "laboratory test pre-training + field data transfer learning", which effectively solves the model training difficulties caused by the scarcity of disaster samples. In view of the problem of few samples of disaster events such as rock burst and poor generalization ability of pure data-driven models, the present application first constructs a basic sample library through dynamic and static combined loading test in the laboratory, so that the model can first learn the basic damage evolution law of rock under the force-seismic coupling effect; then, the knowledge of the pre-trained model in the laboratory is transferred to the engineering site by using the transfer learning technology, and only a small amount of field data is needed for fine tuning to adapt to the specific geological environment. This strategy significantly reduces the dependence on massive field disaster samples, ensures the rapid convergence and stable generalization ability of the model under small sample conditions, and solves the core bottleneck of the application of data-driven models in engineering practice.

[0017] 3. The present application establishes a closed-loop correction and online updating mechanism based on physical deformation feedback, which significantly improves the long-term robustness and self-adaptability of the system. Existing early warning systems are mostly open-loop, and the model is fixed once deployed, which cannot adapt to the dynamic changes of geological conditions. The present application introduces real-time surrounding rock deformation monitoring data as a physical verification benchmark to perform confidence weighted correction on the instability probability output by the model, constrains the prediction deviation of the model with physical measurement, and effectively reduces false positives and false negatives. More importantly, the system can automatically identify the data of the "false negative" period as difficult example samples, and perform online incremental learning on the model to realize the continuous self-optimization and evolution of the early warning model during operation. This enables the system to dynamically adapt to the influence of mining disturbance and changes in geological conditions, ensuring the long-term stability of the early warning accuracy.

[0018] In summary, the present application forms a complete early warning technology system from data perception, intelligent identification to dynamic optimization through the organic combination of three core technologies of mechanism-driven feature fusion, hybrid modeling of small sample adaptation and closed-loop evolution of physical feedback. This not only realizes more accurate and timely intelligent perception and early warning of deep roadway surrounding rock instability risk, but also fundamentally enhances the adaptability, reliability and practicality of the early warning system under complex geological conditions, providing strong technical support for safe and efficient mining of deep coal mines. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 A deep coal seam roadway surrounding rock core drilling position schematic diagram of an embodiment of the present application is shown in the figure; Figure 2 A hard rock specimen processed by an embodiment of the present application is shown in the figure; Figure 3 A soft rock specimen processed by an embodiment of the present application is shown in the figure; Figure 4 A schematic diagram of a rock triaxial pressure loading testing machine and a pendulum impact testing machine of an embodiment of the present application is shown in the figure; Figure 5 A layout plan of a measuring station of an embodiment of the present application is shown in the figure; Figure 6 An architecture diagram of an early warning model of an embodiment of the present application is shown in the figure; Figure 7 A method flowchart of an embodiment of the present application is shown in the figure; Figure 8 A model online feedback correction and closed-loop optimization logic schematic diagram of an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical scheme and beneficial effects of the present application clearer and more apparent, the present application will be further described in detail below in combination with specific embodiments and with reference to the drawings. Some but not all of the embodiments of the present application will be shown in the drawings. In fact, various embodiments of the present application can be implemented in many different forms, and should not be interpreted as being limited to the embodiments described herein; on the contrary, these embodiments are provided to meet the applicable legal requirements.

[0021] In the description of the present application, it should be noted that the terms "inner", "outer", "upper", "lower", "front", "rear" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and therefore cannot be understood as indicating or implying that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application. In addition, the terms "first" and "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0022] This embodiment is based on the actual situation of a coal mine and refers to... Figures 1 to 8 As shown, an intelligent early warning method for the stability of surrounding rock in deep coal seam roadways is provided. The method includes the following steps: S1. Construct a basic sample library containing input feature vectors and stability labels through dynamic and static combined loading rock tests.

[0023] S11, such as Figure 1 As shown, hard rock and soft rock samples were drilled in the surrounding rock of the deep coal seam roadway. Three core sampling points were selected on site, and three rock core samples were drilled from the two sides and the roof at each core sampling point. A total of nine rock core samples were drilled on site, and the length of each core sampling hole was 15m.

[0024] S12. Process the drilled rock samples into standard-sized rock specimens.

[0025] The rock core samples drilled on-site were wrapped in plastic wrap to prevent contact with air and water, and then transported to a processing plant for standardized processing to prepare standard-sized (50mm × 100mm) rock specimens. Figure 2 , Figure 3 As shown, a total of 15 hard rock specimens and 15 soft rock specimens were prepared.

[0026] S13, such as Figure 4 As shown, a rock triaxial compression loading tester and a pendulum impact testing machine were used for combined loading, applying static axial compression at different gradients. Simulating original rock stress and dynamic impact Simulated vibration.

[0027] S14. Record the rock failure state under different combinations of working conditions and construct a basic sample library. : ; in, The input features for the i-th test include the applied static axial stress value and the dynamic impact energy value; For stability labels, 0 is marked if the rock specimen remains intact, and 1 is marked if the rock specimen is damaged. K represents the total number of test samples, which is the sum of the effective number of tests under different dynamic and static combined loading conditions. In this test, K is set to 30.

[0028] S2. Synchronously acquire static stress and microseismic data of the surrounding rock at the site, and map them to a continuous time window for time-series aggregation.

[0029] S21. Measuring stations are deployed at intervals in key monitoring areas of deep roadways. Each station is equipped with a deep-hole multi-point displacement gauge, a bolt cable force gauge (sampling frequency set to 1-10Hz), and a micro-seismic pickup array. After installation, relevant software is used for debugging. Specifically, three sets of measuring stations are deployed at 100m intervals on-site.

[0030] S22. The anchor / cable stress sequence σ(t) is acquired in real time, and the elastic wave signal generated by the microfracture of the rock mass is captured by a microseismic sensor array to calculate the energy sequence E(t) of the microseismic event. A slow increase in σ(t) indicates that the surrounding rock is undergoing compression deformation and the static load is increasing. A sudden and violent fluctuation in σ(t) indicates that the internal structure of the rock mass has been adjusted. E(t) records the amount of energy released by the microseismic event occurring at time t.

[0031] S23. Map discrete microseismic data to a continuous time window. Above, define the first Polymer stress within a time window and aggregate energy for: ; ; in, For time window The number of stress sampling points within; For time window The total number of microseismic events that occurred within the area.

[0032] S3. Standardize the aggregated data and construct the seismic-force coupling feature vector based on the principle of dynamic and static load superposition.

[0033] S31. Clean the collected data, use Lagrange interpolation to complete missing data and remove outliers.

[0034] S32. Standardize each physical quantity using the Z-Score method. The standardized values ​​are... : ; in, This is the original monitoring data; The mean of the feature data; The standard deviation is denoted as .

[0035] S33. Construct an equivalent dynamic-static combined stress index with lithology self-adaptive capability. The calculation formula is as follows: ; wherein, is the aggregate value of the static stress of the surrounding rock monitored at time t; is the aggregate value of the energy released by the microseismic event at time t; E0 is the energy normalization reference value; is the dynamic-static load dimension conversion coefficient, which is positively correlated with the impact tendency index of the surrounding rock of the roadway; ln is the natural logarithm function.

[0036] S34, in order to capture the mutation trend before the catastrophe, the first order difference characteristics of the index is calculated : ; wherein, is the aggregate value of the equivalent dynamic-static combined stress index after standardization in the kth time window; is the aggregate value of the equivalent dynamic-static combined stress index after standardization in the k-1th time window (i.e. the last time); is the time interval (sampling step) between the two consecutive time windows.

[0037] S35, constructing a model input feature vector : ; wherein, denotes the data after standardization; is the aggregate value of the static stress of the surrounding rock after standardization at time k; is the aggregate value of the energy of the microseismic event after standardization at time k; is the value of the equivalent dynamic-static combined stress index after standardization at time k.

[0038] S4, establishing an early warning model based on a long short-term memory network, and defining a network structure to mine the time sequence evolution law of the feature vector.

[0039] For the input vector at time , the state is updated through the forget gate, the input gate and the output gate: Calculate the forget gate: ; wherein, is the output vector of the forget gate, whose value range is between (0, 1), which represents the retention proportion of the cell state information at the last time, the closer the value is to 1, the more retention, and the closer the value is to 0, the more forgetting; Sigmoid is the activation function; is the weight matrix of the forget gate; is the hidden state of the previous time step; is the input feature vector of the current time step; is the bias vector of the forget gate; Calculate the input gate: ; ; where, is the output vector of the input gate, which determines which new information at the current time step will be updated into the cell state; is the candidate cell state of the current time step, which is a new vector containing new information created by the tanh layer; tanh is the hyperbolic tangent activation function; and are the weight matrices calculated for the input gate and the candidate state, respectively; and are the corresponding bias vectors, respectively; Update the state: = ⊙ + ⊙ ; where, is the updated cell state at the current time step, which is calculated by forgetting old information and adding new candidate information; is the cell state of the previous time step; ⊙ represents Hadamard product, which is the operation of multiplying corresponding position elements of matrices or vectors.

[0040] Output the hidden layer features and hidden state at the current time step: ; = ⊙ ; where, is the output vector of the output gate, which determines which information in the current cell state will be output, and the closer the value is to 1, the more output, and the closer the value is to 0, the more shielding; is the hidden state at the current time step, which is the final feature representation after filtering by the output gate, and it will be passed to the LSTM unit of the next time step; is the weight matrix of the output gate, which is used to weight sum the input data and the hidden state of the previous time step, and determines the emphasis of the output information; is the bias term of the output gate, which is used to adjust the threshold value of the activation function and improve the fitting ability of the model; Calculate the output layer probability: ; in, The value is the probability value of the surrounding rock becoming unstable at the current time t, which is predicted by the model. Its value ranges between [0,1]. The closer the value is to 1, the higher the risk of instability. The closer the value is to 0, the more stable the surrounding rock is. The weight matrix of the output layer is used to map the hidden feature vectors output by the LSTM to a one-dimensional decision space. Here, represents the bias term for the output layer; e is the natural constant, the base of the natural logarithm; the model training process uses a binary cross-entropy loss function with added L2 regularization. The loss function is: ; in, This represents the batch sample size. This is a real label; Predict probabilities for the model; For L2 regularization terms, The regularization coefficient is used. These are all network weight parameters.

[0041] S5. Using a transfer learning strategy, the model is first pre-trained based on the S1 basic sample library to learn the basic temporal characteristics of rocks under force-seismic coupling. Then, the model is fine-tuned using the S2 field-collected data and validated using the test set.

[0042] S51. The model is pre-trained using rock test data with dynamic and static loading. Since the size of the rock specimen (50mm×100mm) is very different from the scale of the roadway on site, and the loading energy level in the laboratory is low, a similarity ratio conversion is required before training so that the model can learn the basic "stress-vibration-damage" evolution law.

[0043] S52. A transfer learning strategy is adopted to keep the weight parameters of the feature extraction layer in the pre-trained model unchanged. This layer has learned the basic temporal characteristics of rocks under force-seismic coupling in the dynamic and static combined loading rock test data. Only the parameters of the classification decision layer of the model are unfrozen and fine-tuned. The parameters of this layer are updated using a small amount of field-collected data to adapt to the specific geological environment and noise distribution on site.

[0044] S53. Calculate the F1 score of the model on the test set. If the F1 score is lower than the set threshold (0.8), adjust the hyperparameters. They then returned to training.

[0045] ; in, TP (True Positive) represents that the model "successfully captures the danger of surrounding rock instability"; FP (False Positive) represents that the model issues a "false alarm" in the stable state of surrounding rock; FN (False Negative) represents that the model does not issue any warning for surrounding rock instability.

[0046] S6, input the real-time collected and processed seismic-force coupling feature vector into the model, and output the instability probability and warning level.

[0047] S61, generate real-time seismic-force coupling feature vector from real-time collected surrounding rock static stress and microseismic data , and input it into the warning model.

[0048] S62, the model outputs the instability probability value P of the surrounding rock at the current time.

[0049] According to the instability probability value P, three warning levels are set: P<0.5 is the green safety zone, the surrounding rock is in a stable state, and normal monitoring is maintained; 0.5≤P<0.8 is the yellow warning zone, the monitoring index shows abnormal fluctuations, suggesting that support should be strengthened or the excavation speed should be reduced; P≥0.8 is the red alarm zone, the surrounding rock is in a critical instability state, the system automatically triggers an audible and visual alarm, and the terminal notifies the on-site personnel to evacuate immediately.

[0050] S7, check and correct the instability probability based on the actual deformation data of the surrounding rock, and add the missed samples to the difficult sample library for online updating of the model.

[0051] Physical verification: real-time acquisition of the actual deformation of the surrounding rock, and calculation of the physical deformation deviation index by comparing the actual deformation with the limit displacement threshold.

[0052] Probability correction: using the physical deformation deviation index to weight and fuse the initial prediction probability to generate the final warning level, and using the physical measured data to constrain the prediction deviation of the model.

[0053] Online updating: real-time monitoring of the deformation rate of the surrounding rock, and determining the missed difficult sample when the model predicts safety but the actual deformation rate exceeds the preset threshold, and adding it to the training library for online incremental updating of the model, to realize the adaptive evolution of the warning model.

[0054] S71, real-time acquisition of the actual deformation of the surrounding rock by on-site multi-point displacement meter , comparison with the set limit displacement threshold , calculation of the normalized deformation ratio : .

[0055] S72, construction of the final warning probability after correction : ; wherein, is a confidence weight; is a model raw prediction probability; is a rectified linear unit.

[0056] S73, model online evolution: define when and , it is determined as "false alarm", the period data is marked as y=1, the period data is added to the difficult example sample library, and incremental training is performed; wherein, is a rate threshold.

[0057] So far, the present embodiment has been described in detail in conjunction with the drawings. According to the above description, those skilled in the art should have a clear understanding of the intelligent early warning method for stability of surrounding rock of deep coal seam roadway of the present application. The above-described specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only for specific embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for intelligent early warning of surrounding rock stability in deep coal seam roadways, characterized in that, The method includes the following steps: S1. Construct a basic sample library containing input feature vectors and stability labels through dynamic and static combined loading rock tests; S2. Simultaneously acquire static stress and microseismic data of the surrounding rock at the site, and map them to a continuous time window for time-series aggregation; S3. Standardize the aggregated data and construct a seismic-force coupling feature vector based on the principle of dynamic and static load superposition. S4. Establish an early warning model based on long short-term memory networks; S5. Adopting a transfer learning strategy, first pre-train the model based on the S1 basic sample library, then fine-tune it using the on-site collected data in S2, and validate it using the test set. S6. Input the real-time collected and processed seismic-force coupling feature vector into the model and output the instability probability. S7. The instability probability is verified and corrected based on the actual deformation data of the surrounding rock, and the missed samples are added to the difficult case sample library to update the model online.

2. The intelligent early warning method for the stability of surrounding rock in deep coal seam roadways according to claim 1, characterized in that, S1 includes: S11. Drill hard rock and soft rock samples in the surrounding rock of deep coal seam roadways; S12. Process the drilled rock samples into rock specimens of standard size; S13. A rock triaxial compression loading tester and a pendulum impact tester were used for combined loading, applying static axial compression simulating the original rock stress and dynamic impact simulating vibration, respectively. S14. Record the rock failure state under different combinations of working conditions and construct a basic sample library. : ； in, The input features for the i-th test include the applied static axial stress value and the dynamic impact energy value; For stability labels, 0 is marked if the rock specimen remains intact, and 1 is marked if the rock specimen is damaged. K represents the total number of test samples, which is the sum of the effective number of tests under different dynamic and static combined loading conditions.

3. The intelligent early warning method for the stability of surrounding rock in deep coal seam roadways according to claim 1, characterized in that, S2 includes: S21. Set up monitoring stations at intervals in key monitoring areas of deep roadways. Each monitoring station is equipped with a deep-hole multi-point displacement gauge, an anchor cable force gauge and a micro-vibration pickup array. S22. Real-time acquisition of anchor bolt / cable force sequence σ(t), and by capturing elastic wave signals generated by micro fractures in the rock mass through a microseismic sensor array, the energy sequence E(t) of the microseismic event is calculated. S23. Map discrete microseismic data to a continuous time window. Above, define the first Polymer stress within a time window and aggregate energy for: ； ； in, For time window The number of stress sampling points within; For time window The total number of microseismic events that occurred within the area.

4. The intelligent early warning method for the stability of surrounding rock in deep coal seam roadways according to claim 1 or 3, characterized in that, S3 includes: S31. Clean the collected data, fill in missing data and remove outliers and noise. S32. Standardize each physical quantity using the Z-Score method. The standardized values ​​are... : ； in, This is the original monitoring data; The mean of the feature data; Standard deviation; S33. Construct an equivalent dynamic-static combined stress index with lithology self-adaptive capability. The calculation formula is as follows: ； in, The static stress aggregation value of the surrounding rock monitored at time t; E0 represents the aggregated energy released by the microseismic event at time t; E0 is the energy normalization baseline value. is the dynamic-static load dimension conversion coefficient, whose value is positively correlated with the impact tendency index of the surrounding rock of the roadway; ln is the natural logarithm function; S34. First-order difference characteristics of the calculation index : ； in, The equivalent dynamic and static combined stress index aggregate value after standardization within the k-th time window; The equivalent dynamic and static combined stress index aggregate value after standardization within the (k-1)th time window; The time interval between two consecutive time windows; S35. Constructing the model input feature vector : ； in, This represents data that has undergone standardization. The normalized static stress aggregate value of the surrounding rock at time k; The standardized aggregate energy value of the microseismic event at time k; The value of the equivalent dynamic-static combined stress index at time k is the standardized value.

5. The intelligent early warning method for the stability of surrounding rock in deep coal seam roadways according to claim 1, characterized in that, The long short-term memory network early warning model established in S4, for time... input vector The state is updated using a forget gate, an input gate, and an output gate: Computational forget gate: ； in, The output vector of the forget gate has a value between (0,1), representing the proportion of cell state information retained from the previous time step. The closer the value is to 1, the more information is retained; the closer it is to 0, the more information is forgotten. Sigmoid is the sigmoid activation function. Here is the weight matrix for the forget gate; This represents the hidden layer state from the previous time step. This is the input feature vector at the current time. This is the bias vector for the forget gate; Calculate the input gate: ； ； in, The output vector of the input gate determines what new information will be updated to the cell state at the current moment; The candidate cell state at the current moment is a new vector containing new information created through the tanh layer; tanh is the hyperbolic tangent activation function. and The weight matrices are calculated for the input gate and the candidate state, respectively. and These are the corresponding bias vectors; Perform a status update: = ⊙ + ⊙ ； in, The cell state at the current moment is calculated by forgetting old information and adding new candidate information; The cell state at the previous time step; ⊙ represents the Hadamard product, which is the operation of multiplying corresponding elements of a matrix or vector; Output the hidden layer features and hidden states at the current time step: ； = ⊙ ； in, The output vector of the output gate determines which information in the current cell state will be output. The closer its value is to 1, the more information will be output, and the closer it is to 0, the more information will be masked. This is the hidden state at the current time step, which is the final feature representation after filtering by the output gate. It will be passed to the LSTM unit at the next time step. The weight matrix of the output gate is used to perform a weighted summation of the input data and the hidden state at the previous time step, determining the emphasis of the output information; This is the bias term for the output gate, used to adjust the threshold of the activation function and improve the model's fitting ability. Calculate the output layer probability: ； in, The value is the probability value of the surrounding rock becoming unstable at the current time t, which is predicted by the model. Its value ranges between [0,1]. The closer the value is to 1, the higher the risk of instability. The closer the value is to 0, the more stable the surrounding rock is. The weight matrix of the output layer is used to map the hidden feature vectors output by the LSTM to a one-dimensional decision space. is the bias term for the output layer; e is the natural constant, the base of the natural logarithm.

6. The intelligent early warning method for the stability of surrounding rock in deep coal seam roadways according to claim 5, characterized in that, The model training process uses a binary cross-entropy loss function with an added L2 regularization term. The loss function is: ； in, This represents the batch sample size. This is a real label; Predict probabilities for the model; For L2 regularization terms, The regularization coefficient is used. These are all network weight parameters.

7. The intelligent early warning method for the stability of surrounding rock in deep coal seam roadways according to claim 1, characterized in that, S5 includes: S51. Use rock test data with dynamic and static loading to pre-train the model. Similarity ratio conversion is required before training. S52. Adopt a transfer learning strategy, keep the weight parameters of the feature extraction layer in the pre-trained model unchanged, and only unfreeze and fine-tune the parameters of the classification decision layer of the model. S53. Calculate the F1 score of the model on the test set. If the F1 score is lower than the set threshold, adjust the hyperparameters and return to training.

8. The intelligent early warning method for the stability of surrounding rock in deep coal seam roadways according to claim 1, characterized in that, S6 includes: S61. Generate a real-time seismic-force coupling feature vector by combining the real-time acquired static stress of the surrounding rock with the microseismic data. Input the early warning model; S62. The model outputs the instability probability value P of the surrounding rock at the current moment.

9. The intelligent early warning method for the stability of surrounding rock in deep coal seam roadways according to claim 8, characterized in that, Based on the instability probability value P, three warning levels are set: green safety zone when P < 0.5, yellow warning zone when 0.5 ≤ P < 0.8, and red alarm zone when P ≥ 0.

8.

10. The intelligent early warning method for the stability of surrounding rock in deep coal seam roadways according to claim 1, characterized in that, S7 includes: S71. The actual deformation of the surrounding rock is obtained in real time through multi-point displacement gauges on site. Compare with the set limit displacement threshold Calculate the normalized deformation ratio : ； S72. Constructing the revised final early warning probability : ； in, Confidence weights; The original predicted probabilities of the model; To correct the linear unit; S73, Online Model Evolution: Define when and If a data point is missed, it is considered a "missed report," and the data for that period is added to the difficult case sample library for incremental training. in, This is the rate threshold.

Images