Tunnel disaster early warning method and device, equipment and storage medium

Abstract

The invention provides a tunnel disaster early warning method and device, equipment and a storage medium. Comprising the following steps: determining lithologic conditions and geologic structures along a tunnel, and determining a plurality of types of unfavorable geologic bodies and resistivity ranges thereof; establishing a tunnel finite element simulation model according to the lithologic conditions; unfavorable geologic bodies are respectively arranged at different distances in front of a tunnel face of the tunnel finite element simulation model, simulation is carried out, a plurality of samples are obtained, and a data set is constructed; the sample comprises the distance, the sample resistivity and the geologic body type, and the sample resistivity is the resistivity of the unfavorable geologic body measured by the tunnel face; constructing a plurality of machine learning models; training a plurality of machine learning models by using the data set, evaluating the performance of the models, and taking the optimal machine learning model as a tunnel disaster early warning model; acquiring the resistivity of a geologic body in front of the tunnel face; and according to the resistivity of the geologic body in front of the tunnel face and the tunnel disaster early warning model, predicting the type of a disaster occurring in front of the tunnel face.

Description

Technical Field

[0001] The present invention relates to the technical field of disaster warning, and in particular to a tunnel disaster warning method and device, equipment and storage medium. Background Art

[0002] With the continuous improvement of the national water network and the development of urban underground space, higher requirements will be placed on the safety and speed of tunnel construction. At present, tunnel construction often passes through strata with complex engineering geological characteristics, resulting in frequent geological disasters such as sudden water inrush, face collapse, and soft-hard composite interface during tunnel construction, which in turn causes low tunnel construction efficiency or causes serious machine jams or flooding of full-section hard rock tunnel boring machines (TBMs). Early warning of geological disasters ahead of the tunnel face and guidance of advanced treatment construction will greatly improve construction efficiency and reduce construction delays and property losses caused by engineering disasters.

[0003] Existing advance detection equipment is relatively expensive and complicated to operate, focusing on identifying geological structures. Resistivity measurement is relatively fast and low-cost, providing a convenient means for geological risk prediction, but it still requires professionals to interpret and judge the possible disaster risks ahead, and the prediction results will lag behind the construction progress to a certain extent. The patent of this invention is based on detailed preliminary geological survey data, and proposes a method for establishing a machine learning early warning model for tunnel advance disasters based on numerical simulation of resistivity changes. It has the characteristics of faster early warning of geological disasters ahead and does not require professional interpretation. Summary of the invention

[0004] The present invention provides a tunnel disaster warning method and device, equipment and storage medium, which can construct a tunnel disaster warning model with high prediction accuracy, so as to predict the tunnel disasters that are about to occur in front of the tunnel face during the tunnel construction process, thereby ensuring the safety of the construction process.

[0005] On the one hand, a tunnel disaster early warning method is provided, comprising: Obtain the lithology conditions and geological structures along the tunnel; Determine several types of adverse geological bodies and their resistivity ranges based on lithological conditions and geological structures; Establish a tunnel finite element simulation model based on the adverse geological body and its resistivity range; Unfavorable geological bodies are set at different distances in front of the tunnel face of the tunnel finite element simulation model, and simulation is performed. The type of each unfavorable geological body and the distance between it and the tunnel face are recorded to obtain several samples and construct a data set. The samples include distance, sample resistivity and geological body type. The sample resistivity is the resistivity of the unfavorable geological body measured at the tunnel face. Build multiple machine learning models; The dataset is used to train multiple machine learning models, and the model performance is evaluated, with the optimal machine learning model used as the tunnel disaster warning model.

[0006] Optionally, the poor geological body is a possible poor geological body determined based on lithological conditions and geological structure, and the poor geological body includes water-rich faults, complex mixed strata and cavities; the resistivity range of the poor geological body is determined based on the resistivity of the geological structure obtained based on the geological structure and lithological conditions.

[0007] Optionally, during simulation, at least two or more spacings are set for each unfavorable geological body; at each spacing, each unfavorable geological body is divided into two types according to the resistivity range of the unfavorable geological body, with a resistivity increment of Increase the resistivity of the adverse geological body, perform multiple simulations, and obtain multiple groups of samples.

[0008] Optionally, when training multiple machine learning models, the data set is divided into a training set and a test set, the training set is used to train the multiple machine learning models, and the test set is used to test the trained multiple machine learning models to obtain the best machine learning model; in the process of training the multiple machine learning models with the training set, the hyperparameters of each machine learning model are optimized by the improved sparrow optimization algorithm to obtain the respective optimal hyperparameter combinations, and the machine learning model with the optimal hyperparameter combination is used for testing.

[0009] Optionally, the method for optimizing the hyperparameters of each machine learning model by using the improved sparrow optimization algorithm is as follows: The hyperparameter group of the machine learning model to be optimized is used to construct coordinates to represent the position of the sparrow in the sparrow optimization algorithm, and the performance index of the machine learning model to be optimized obtained by training according to the sparrow's coordinates is used as the fitness value of the sparrow; Classify sparrows in a sparrow population into finders, followers, and scouts; Initialize the sparrow population and calculate the fitness values ​​of all sparrows; Position update: Select the butterfly search strategy or the triangle walk strategy according to the size of the random number to update the discoverer's position; update the positions of the followers and scouts; Recalculate the fitness values ​​of all sparrows; Repeat the position update and fitness value calculation until the set number of iterations is reached; The sparrow position with the best fitness value is output as the optimal hyperparameter group of the machine learning model to be optimized.

[0010] Optionally, a cubic chaos mapping strategy is used to initialize the sparrow population. The initialization formula is as follows:

[0011]

[0012] Indicates the first The sparrow's The coordinates of the dimension, i.e. The first of the hyperparameter groups The value of the hyperparameter; Represents the hyperparameter group The upper limit of the hyperparameters; Represents the hyperparameter group The lower bound of the hyperparameters; For the A random number of sparrows.

[0013] Optionally, the method for updating the finder position according to the butterfly search strategy and the triangle walk strategy is as follows: Set a safety threshold. When the random number is not greater than the safety threshold, use the butterfly search strategy to update the discoverer's position; otherwise, use the triangle walk strategy to update the discoverer's position. The update formula of the butterfly search strategy is as follows:

[0014]

[0015] Indicates A sparrow in the The position is updated before the iteration Dimensional coordinates; Indicates A sparrow in the After the first discoverer position update in the iteration Dimensional coordinates; Indicates A sparrow in the After the second position update in the iteration process Dimensional coordinates; for The sparrow with the best fitness value in the iteration Dimensional coordinates; is a random number; is the maximum number of iterations; Indicates the number of iterations; Indicates the amount of aroma perception; For the The amount of scent perceived by a sparrow; is the stimulus intensity, The optimal fitness value of the current sparrow population equal; It's the fragrance. It is a form of perception. , is a constant; is a constant; The update formula of the triangle walk strategy is as follows:

[0016]

[0017] , ,

[0018] Indicates A sparrow in the The position is updated before the iteration The vector formed by the coordinates of the dimension; Indicates A sparrow in the After the first discoverer position update in the iteration The vector formed by the coordinates of the dimension; Indicates A sparrow in the After the second position update in the iteration process Dimensional coordinates; for The individual with the best fitness value in the iteration The vector formed by the coordinates of the dimension; is a random number that follows a normal distribution; To include elements, All the elements in are 1, The size of is equal to the number of dimensions of the sparrow coordinates; is a unit vector with a random direction.

[0019] On the other hand, a tunnel disaster warning device is provided, comprising: The acquisition module is used to obtain the lithological conditions and geological structures along the tunnel; determine several types of adverse geological bodies and their resistivity ranges based on the lithological conditions and geological structures; The finite element simulation module is used to establish a tunnel finite element simulation model based on the unfavorable geological body and its resistivity range; unfavorable geological bodies are respectively set at different distances in front of the tunnel face of the tunnel finite element simulation model, and simulation is performed, and the type of each unfavorable geological body and the distance between it and the tunnel face are recorded to obtain a number of samples and construct a data set; the samples include distance, sample resistivity and geological body type, and the sample resistivity is the resistivity of the unfavorable geological body measured at the tunnel face; The model training module is used to build a variety of machine learning models; use the data set to train a variety of machine learning models, and evaluate the model performance, and use the optimal machine learning model as the tunnel disaster warning model.

[0020] On the other hand, an electronic device is provided, which includes the tunnel disaster warning device as described above.

[0021] On the other hand, a computer-readable storage medium is provided, wherein at least one program code is stored in the computer-readable storage medium, and the program code is executed by a processor to implement the tunnel disaster warning method as described in any one of the above items.

[0022] The technical solution provided by the embodiments of the present disclosure brings the following beneficial effects: In the disclosed embodiment, by obtaining the lithological conditions and geological structures along the tunnel and determining the type and resistivity range of the unfavorable geological body, a tunnel finite element simulation model is established according to the type and resistivity range of the unfavorable geological body, and simulation is performed to form sample data that matches the actual tunnel construction situation. Subsequently, a machine learning model is trained based on the sample data, and the machine learning model with the best model performance obtained through training is used as a tunnel disaster warning model, which can achieve advance prediction of disasters during the construction process. By obtaining the resistivity of the geological body in front of the tunnel face and the distance between the geological body and the tunnel face during the tunnel construction process, the tunnel disaster warning model can be used to predict the geological disasters that are about to occur in front of the tunnel face. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 A method flow chart of a tunnel disaster warning method provided by an embodiment of the present disclosure; Figure 2 A schematic diagram of a tunnel finite element model provided in an embodiment of the present disclosure; Figure 3 A training flow chart of a tunnel disaster warning model provided in an embodiment of the present disclosure; Figure 4 A training flow chart of another tunnel disaster warning model provided by an embodiment of the present disclosure; Figure 5 A structural block diagram of a tunnel disaster warning device provided by an embodiment of the present disclosure; Figure 6 A structural block diagram of an electronic device provided in an embodiment of the present disclosure.

[0025] The reference numerals are as follows: 11: Tunnel face; 12: Unfavorable geological body; 121: Water-rich fault; 122: Complex mixed strata; 123: Cavity; 13: Resistivity measurement mechanism; 131: Power supply; 132: Voltmeter; 133: Measurement column; 21: acquisition module; 22: finite element simulation module; 23: model training module; 31: processor; 32: memory. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] Figure 1 A method flow chart of a tunnel disaster warning method provided by an embodiment of the present disclosure. Figure 1 ,include: S1. Obtain the lithological conditions and geological structure along the tunnel.

[0028] In an achievable embodiment provided by the present disclosure, the specific method of step S1 is as follows: By setting up multiple geological exploration points at intervals along the tunnel, the underground rock conditions and geological structure are determined through the geological exploration points.

[0029] For example, assuming that the total length of the tunnel to be excavated is 10 km, 1,000 geological exploration points can be set along the tunnel to determine the lithological conditions and geological structure along the tunnel.

[0030] Of course, the above setting of geological exploration points is only an example and is not intended to limit the present disclosure. In actual implementation, the number of geological exploration points can be selectively set. The more geological exploration points are set, the clearer the lithological conditions and geological structures along the tunnel will be. However, the more geological exploration points are set, the more difficult the geological exploration work will be.

[0031] In step S1, the lithology condition refers to the main rock mass structure in the geological structure, for example, the rock mass structure may be limestone, sandstone, etc. The geological structure may be composed of rock mass structure, soil, water, etc. The composition of the geological structure along the tunnel may be different, so it is necessary to determine the specific geological structure through exploration.

[0032] In another achievable embodiment provided by the present disclosure, the lithological conditions and geological structures along the tunnel may be pre-explored by other companies or individuals, and the data provided by other companies or individuals may be directly obtained.

[0033] S2. Determine several types of adverse geological bodies and their resistivity ranges based on lithological conditions and geological structures.

[0034] In step S1, the bad geological body is a possible bad geological body determined according to lithological conditions and geological structure, and the bad geological body includes water-rich faults, complex mixed strata and cavities; the resistivity range of the bad geological body is determined based on the resistivity of the geological structure obtained according to the geological structure and lithological conditions.

[0035] In the disclosed embodiment, during the tunnel excavation process, if the tunnel boring machine excavates a bad geological body, disasters such as tunnel collapse and tunnel boring machine damage may occur. Whether a bad geological body will be formed mainly depends on the underground lithological conditions and geological structure. All possible bad geological bodies are determined by the lithological conditions and geological structure, so as to facilitate the subsequent construction of a data set based on the possible bad geological bodies. In addition, the different geological structures at the geological exploration points will cause the resistivity obtained at the geological exploration points to be different. For example, if the lithological conditions in the geological structure at multiple geological exploration points are limestone, if the proportions of limestone, soil, water, etc. in the geological structure are different, the resistivity obtained at the geological exploration points will be different. Therefore, it is necessary to determine the resistivity of the geological structure based on the actual lithological conditions and geological structure explored, so as to ensure that the data set finally used for model training matches the actual data along the tunnel, thereby ensuring the prediction accuracy of the trained model.

[0036] The above poor geological bodies are only some examples provided by the present disclosure. In actual situations, poor geological bodies include but are not limited to the types of the above poor geological bodies.

[0037] S3. Establish a tunnel finite element simulation model based on the unfavorable geological body and its resistivity range.

[0038] In an example provided in the present disclosure, the tunnel finite element model includes at least a tunnel face and a poor geological body. In addition, at least one resistivity measurement mechanism is provided at the tunnel face. Of course, the tunnel finite element model may further include a tunnel, which is not limited in the present disclosure.

[0039] In one example, the resistivity measurement mechanism measures the resistivity of a sample of a poor geological body using a Wenner quadrupole resistivity measurement method.

[0040] Figure 2 A schematic diagram of a tunnel finite element model provided by an embodiment of the present disclosure. Figure 2 The constructed tunnel finite element model includes a tunnel face 11, a poor geological body 12 and a resistivity measurement mechanism 13. The poor geological body 12 includes a water-rich fault 121, a complex mixed bottom layer 122 and a cavity 123.

[0041] In the embodiment of the present disclosure, according to Figure 2 A brief explanation of the setting principle of the resistivity measurement mechanism: Figure 2 The resistivity measuring mechanism 13 shown in FIG. 1 is actually a circuit diagram of the resistivity measuring mechanism 13. In actual situations, when the resistivity measuring mechanism 13 is set at the tunnel face 11, a power supply 131, a voltmeter 132 and a plurality of measuring columns 133 may be set. The power supply 131 is used for power supply, the voltmeter 132 is used for measuring voltage, and the setting positions of the plurality of measuring columns 133 may refer to Figure 2 At the four positions A, B, C, and D in the figure, the spacing between each measuring column 133 can be The length of the measuring column 133 can be set as needed. The longer the length of the measuring column 133 is, the resistivity of the poor geological body farther away from the tunnel face can be measured by the measuring column 133, but it will also make the measuring column easy to be damaged. Therefore, the length of the measuring column needs to be determined according to actual conditions.

[0042] The resistivity calculation formula of the resistivity measuring mechanism 13 is as follows:

[0043] is the measured resistivity of the adverse geological body, i.e., the sample resistivity; To measure the distance between the columns; is the voltage between the measuring column (C) and the measuring column (D); The current supplied by the power supply.

[0044] S4. Unfavorable geological bodies are set at different distances in front of the tunnel face of the tunnel finite element simulation model, and simulation is performed. The type of each unfavorable geological body and the distance between it and the tunnel face are recorded to obtain several samples and construct a data set. The samples include distance, sample resistivity and geological body type. The sample resistivity is the resistivity of the unfavorable geological body measured at the tunnel face.

[0045] In an example provided in the present disclosure, step S4 includes the following steps: During simulation, at least two spacings are set for each unfavorable geological body; at each spacing, each unfavorable geological body is divided into two groups according to the resistivity range of the unfavorable geological body and the resistivity increment. Increase the resistivity of the adverse geological body, perform multiple simulations, and obtain multiple groups of samples.

[0046] In the disclosed embodiment, a finite element model is established, and a finite element simulation analysis is performed according to the range of resistivity of one or more possible disasters, thereby obtaining multiple sets of data on the distance between the tunnel face and the bad geological body, the sample resistivity, and the type of the bad geological body. The data obtained by the finite element simulation analysis can reflect the actual values ​​of the distance between the tunnel face and the area to be constructed and the resistivity of the area to be constructed when a disaster occurs under actual construction conditions. The data obtained by the finite element simulation analysis is used for model training, so that during the tunnel construction process, the tunnel disaster warning model can determine whether a bad geological body appears in the area in front of the tunnel face based on the measured resistivity value of the area in front of the tunnel face and the distance from the tunnel face, thereby predicting whether a disaster is about to occur.

[0047] The specific implementation process of step S4 is briefly introduced below based on a specific example: Assuming that the tunnel diameter is D, models of unfavorable geological bodies such as water-rich faults, complex mixed strata and cavities are established at a certain distance (e.g., 0, 0.05D, 0.1D, 0.15D, 0.2D, 0.25D, 0.3D, 0.35D, 0.4D, 0.45D, and 0.5D) in front of the tunnel face in the finite element simulation model.

[0048] Assume that at a certain geological exploration point, the underground lithology is mainly limestone, and the resistivity of the determined geological structure is 400 According to the lithology and geological structure, the unfavorable geological bodies include water-rich faults, complex mixed strata and cavities. It can be determined based on the resistivity of the geological structure. The resistivity range of water-rich faults is 200~300. The resistivity of complex mixed formations ranges from 300 to 400 , the resistivity of the cavity ranges from 100 to 200 .

[0049] Under the condition of fixed distance (the distance between the unfavorable geological body and the tunnel face) (i.e., at a certain spacing, for example, 0.05D), the (Resistivity increment ) intervals are assigned to each unfavorable geological body, and the sample resistivity value of the unfavorable geological body is measured by the resistivity measuring device at the tunnel face.

[0050] Then, adjust the distance between the unfavorable geological body and the tunnel face (for example, adjust the value to 0.1D), and then adjust the distance to 0.5D. (Resistivity increment ) intervals are assigned to each unfavorable geological body, and the sample resistivity value of the unfavorable geological body is measured again by the resistivity measuring device at the tunnel face.

[0051] Ultimately, 6,600 sets of samples including the distance between the tunnel face and the adverse geological body, sample resistivity, and the type of adverse geological body will be obtained to form a data set.

[0052] S5. Build multiple machine learning models.

[0053] In some examples provided in the present disclosure, machine learning models include a distributed gradient boosting library (XGBoost) model, a lightweight gradient boosting machine (LightGBM) model, and a classification feature boosting (CatBoost) model.

[0054] Of course, the actual implementation process may include but is not limited to the above-mentioned machine learning models.

[0055] S6. Use the dataset to train multiple machine learning models and evaluate the model performance, with the optimal machine learning model used as the tunnel disaster warning model.

[0056] In one example, step S6 includes the following steps: When training multiple machine learning models, the data set is divided into a training set and a test set. The training set is used to train the multiple machine learning models, and the test set is used to test the trained multiple machine learning models to obtain the best machine learning model. In the process of training multiple machine learning models with the training set, the hyperparameters of each machine learning model are optimized using the improved sparrow optimization algorithm to obtain the optimal hyperparameter combination of each model, and the machine learning model with the optimal hyperparameter combination is used for testing.

[0057] In the disclosed embodiments, during the training process of the machine learning model, the model performance of the machine learning model finally trained using different hyperparameters may be different. In order to ensure the model performance of the final machine learning model, during the training process of the machine learning model, the hyperparameters of the machine learning model are optimized to obtain the optimal hyperparameter combination for each machine learning model. The machine learning model is trained with the optimal hyperparameter combination, so that the model performance of each type of machine learning model obtained through training can be ensured. Finally, by evaluating the model performance of multiple types of machine learning models trained using the optimal hyperparameter combination, the machine learning model with the best model performance is determined as the tunnel disaster prediction model.

[0058] In the disclosed embodiment, the training strategy adopted by the machine learning model may be a ten-fold cross-validation training strategy.

[0059] In some examples provided in the present disclosure, the hyperparameters of each machine learning model are optimized by using the improved sparrow optimization algorithm to obtain the respective optimal hyperparameter combinations, including the following steps: Step 1: Construct coordinates with the hyperparameter group of the machine learning model to be optimized to represent the position of the sparrow in the sparrow optimization algorithm, and use the performance index of the machine learning model to be optimized obtained by training according to the sparrow's coordinates as the fitness value of the sparrow.

[0060] In the embodiment of the present disclosure, the performance indicators of the machine learning model to be optimized include the harmonic mean of precision and recall (ie, F1 value), accuracy, etc.

[0061] In the embodiment of the present disclosure, the F1 value is used as the fitness value of the sparrow, and the calculation process of the F1 value is as follows: , , ; in, is the number of samples that the model correctly predicts as positive class samples. The number of samples that the model incorrectly predicts as positive samples from the negative category; The number of samples that the model incorrectly predicts as negative class samples from the positive class; for accuracy; is the recall rate.

[0062] Step 2: Divide the sparrows in the sparrow population into discoverers, followers, and scouts.

[0063] In one example, the proportions of finders, followers, and scouts in a sparrow population are 20%, 70%, and 10%, respectively.

[0064] Of course, the above ratio is only an example provided by the present disclosure, and can be set as needed during actual implementation.

[0065] Step 3: Initialize the sparrow population and calculate the fitness values ​​of all sparrows.

[0066] In one example, step 3 includes: The cubic chaos mapping strategy is used to initialize the sparrow population. The initialization formula is as follows:

[0067]

[0068] Indicates the first The sparrow's The coordinates of the dimension, i.e. The first of the hyperparameter groups The value of the hyperparameter; Represents the hyperparameter group The upper limit of the hyperparameters; Represents the hyperparameter group The lower bound of the hyperparameters; For the A random number of sparrows.

[0069] In the embodiment of the present disclosure, according to the above formula, a random number As an independent variable, As the dependent variable, the coordinates of the sparrow can be initialized. After the sparrow coordinates are initialized, The value of The value of To initialize the By applying the cubic chaos mapping strategy to the initialization particle state of the sparrow optimization algorithm, the uniformity of particle initialization can be improved. In the embodiment of the present disclosure, taking the hyperparameter combination including two hyperparameters as an example (n_estimators and learning_rate), the value range of n_estimators can be , the value range of learning_rate can be . The value of n_estimators is used as the coordinate value of the first dimension of the sparrow, and learning_rate is used as the coordinate value of the second dimension of the sparrow. Within the above range, the position of the sparrow is determined according to the cubic chaos mapping strategy to initialize the sparrow population.

[0070] Step 4, position update: select the butterfly search strategy or the triangle walk strategy according to the size of the random number to update the discoverer's position, and update the positions of the followers and scouts.

[0071] In one example, a method for updating the finder position according to the butterfly search strategy and the triangle walk strategy is as follows: Set a safety threshold. When the random number is not greater than the safety threshold, use the butterfly search strategy to update the finder's position; otherwise, use the triangle walk strategy to update the finder's position. For example, for The safety threshold can be , is a constant, and The value range of .

[0072] The update formula of the butterfly search strategy is as follows:

[0073]

[0074] Indicates A sparrow in the The position is updated before the iteration Dimensional coordinates; Indicates A sparrow in the After the first discoverer position update in the iteration Dimensional coordinates; Indicates A sparrow in the After the second position update in the iteration process Dimensional coordinates; for The sparrow with the best fitness value in the iteration Dimensional coordinates; for Random numbers within is the maximum number of iterations; Indicates the number of iterations; Indicates the amount of aroma perception; For the The amount of scent perceived by a sparrow; is the stimulus intensity, The optimal fitness value of the current sparrow population equal; It's the fragrance. It is a form of perception. , is a constant; is a constant; , , The value range of .

[0075] The update formula of the triangle walk strategy is as follows:

[0076]

[0077] , ,

[0078] Indicates A sparrow in the The position is updated before the iteration The vector formed by the coordinates of the dimension; Indicates A sparrow in the After the first discoverer position update in the iteration The vector formed by the coordinates of the dimension; Indicates A sparrow in the After the second position update in the iteration process Dimensional coordinates; for The individual with the best fitness value in the iteration The vector formed by the coordinates of the dimension; is a random number that follows a normal distribution; To include elements, All the elements in are 1, The size of is equal to the number of dimensions of the sparrow coordinates; is a unit vector with a random direction.

[0079] In the disclosed embodiment, the finder leads the population to search for food globally and updates its own position. It can be seen from the improved sparrow optimization algorithm that when the finder's current environment is not dangerous (i.e. ), the discoverer's own position is updated twice through the butterfly search strategy. During the second position update, the butterfly search strategy uses the smell between individuals to enhance the information exchange between individuals during the search process, improving the defect of the original sparrow optimization algorithm that lacks information exchange between individuals and improving the global search capability. When the current environment of the discoverer is dangerous When the sparrows are moving, they do not need to approach the food directly. Instead, they move around the food using a triangular wandering strategy, increasing the randomness of individual sparrows.

[0080] For example, If the value of is 0.5, If it is 2, the number of dimensions of the sparrow's coordinates is 3, then It can be expressed as .

[0081] In an example provided in the present disclosure, the follower uses the following formula to update the location:

[0082] For the A sparrow in the The first iteration process performs the location update before The vector formed by the coordinates of the dimension; No. A sparrow in the After the location update in the first iteration The vector formed by the coordinates of the dimension; For the The sparrow with the worst fitness value in the iteration The vector formed by the coordinates of the dimension; For the The sparrow with the best fitness value in the iteration The vector formed by the coordinates of the dimension; is a random number that follows a normal distribution; Indicates including elements, All the elements in are 1; for A random number between Represents the number of sparrows in a sparrow population.

[0083] In the improved sparrow optimization algorithm, followers move toward the finder (also known as the leader) in a better position. Followers try to improve their position by reducing the difference between their position and the leader's position. This reflects a strategy of exploring for a better position, that is, moving closer to the leader's position.

[0084] In an example provided by the present disclosure, the scout uses the following formula to update the location:

[0085] For the The position of the sparrow before the iteration process is updated Dimensional coordinates; For the The position of the sparrow after the iteration is updated Dimensional coordinates; For the front The sparrow with the best fitness value in the iteration Dimensional coordinates; is the step size control parameter, and is a random value that follows a normal distribution; for Random value within ; is a constant; Indicates the current sparrow population The fitness value of a sparrow; Indicates the optimal fitness value in the current sparrow population.

[0086] In the improved sparrow optimization algorithm, the scout is in a dangerous situation (such as encountering a predator or being at the edge of the group, that is, ), react quickly and move away from danger or toward a better-positioned individual, in a safe situation (i.e. ), randomly move towards other individuals in the search space.

[0087] In the disclosed embodiment, Used to ensure that the denominator is not 0, It can be a smaller number, such as 0.0001.

[0088] Step 5: Recalculate the fitness values ​​of all sparrows.

[0089] The steps for calculating the fitness value refer to step 1. Step 6: Repeat the position update and fitness value calculation until the set number of iterations is reached.

[0090] Exemplarily, the number of iterations may be 200. Of course, the present disclosure does not limit the number of iterations, and it may be selectively set.

[0091] Step 7: Output the sparrow position with the best fitness value as the optimal hyperparameter group for the machine learning model to be optimized.

[0092] In the disclosed embodiment, the sparrow optimization algorithm is a new type of group intelligence optimization algorithm inspired by the foraging behavior and anti-predation behavior of sparrows. It can be abstracted as a finder-follower model, and a reconnaissance and early warning mechanism is added. In the process of sparrow foraging, the entire population can be divided into finders and followers. The finder is responsible for finding food in the population and providing the foraging area and direction for the entire sparrow population, while the follower relies on the finder to obtain food. In order to obtain food, sparrows can adopt two types of behavioral strategies, finders and followers, to forage. At the same time, a certain proportion of individuals in the population will be selected for reconnaissance and early warning. If danger is found, the food will be abandoned to ensure safety. The position of the finder is updated by randomly selecting the butterfly search strategy and the triangle wandering strategy. The butterfly search strategy can improve the defect of the original sparrow optimization algorithm that lacks information exchange between individuals and enhance the global search capability. Wandering around the food through the triangle wandering strategy can increase the randomness of the sparrow individuals.

[0093] In the disclosed embodiment, step 7 will obtain the optimal hyperparameter combination of multiple types of machine learning models, such as the optimal hyperparameter combination of the distributed gradient boosting library (XGBoost) model, the optimal hyperparameter combination of the lightweight gradient boosting machine (LightGBM) model, and the optimal hyperparameter combination of the classification feature boosting (CatBoost) model, thereby training the distributed gradient boosting library (XGBoost) model, the lightweight gradient boosting machine (LightGBM) model, and the classification feature boosting (CatBoost) model with the best performance, and then calculate the cumulative sum of the model performance indicators of the multi-class models with the best performance obtained through training. If the model performance indicators are the harmonic mean of the precision and recall rate (that is, the F1 value) and the accuracy (Accuracy), then the cumulative sum of the model performance indicators is the sum of the F1 value and the Accuracy.

[0094] Of course, the model performance indicators may include but are not limited to the above-mentioned model performance indicators.

[0095] Figure 3 A training flow chart of a tunnel disaster warning model provided in an embodiment of the present disclosure. Figure 3 The tunnel disaster warning model training process is specifically shown in FIG.

[0096] S7. Obtain the resistivity of the geological body in front of the tunnel face.

[0097] For example, the following is set in front of the tunnel face: Figure 2 The resistivity measurement structure shown in the figure can obtain the resistivity of the mass in front of the tunnel face.

[0098] S8. Determine the type of geological body in front of the tunnel face based on the resistivity of the geological body in front of the tunnel face and the tunnel disaster early warning model.

[0099] In step S8, the measured resistivity and the distance between the geological body and the tunnel face are input, and the tunnel disaster warning model can determine the type of geological body in front of the tunnel face.

[0100] S9. Predict the type of disaster that may occur in front of the tunnel face based on the type of geological body.

[0101] In the disclosed embodiment, by obtaining the lithological conditions and geological structures along the tunnel and determining the type and resistivity range of the unfavorable geological body, a tunnel finite element simulation model is established according to the type and resistivity range of the unfavorable geological body, and simulation is performed to form sample data that matches the actual tunnel construction situation. Subsequently, a machine learning model is trained based on the sample data, and the machine learning model with the best model performance obtained through training is used as a tunnel disaster warning model, which can achieve advance prediction of disasters during the construction process. By obtaining the resistivity of the geological body in front of the tunnel face and the distance between the geological body and the tunnel face during the tunnel construction process, the tunnel disaster warning model can be used to predict the geological disasters that are about to occur in front of the tunnel face.

[0102] Figure 4 This is another training flow chart of a tunnel disaster warning model provided by an embodiment of the present disclosure. Figure 4 , including the following steps: Step 1: Construct coordinates with the hyperparameter group of the machine learning model to be optimized to represent the position of the sparrows and initialize the sparrow population.

[0103] Step 2: Divide the sparrows in the sparrow population into finders, followers, and scouts.

[0104] Step 3, calculate the fitness value of the sparrow.

[0105] Step 4: Use the butterfly search strategy and triangle walk strategy to update the discoverer's position.

[0106] Step 5: Update the follower's location.

[0107] Step 6: Update the location of the scout.

[0108] Step 7, determine whether the number of iterations is met, if not, return to step 3.

[0109] Step 8: When the number of iterations is met, the coordinates of the sparrow with the best fitness value are output as the optimal hyperparameter combination of the machine learning model to be optimized.

[0110] Figure 4 The detailed implementation process of the training step of the tunnel disaster warning model shown in FIG. 1 has been described in step S6, which will not be repeated here, and the content of step S6 can be used as a reference.

[0111] Figure 5 This is a structural block diagram of a tunnel disaster warning device provided by an embodiment of the present disclosure. Figure 5 ,include: The acquisition module 21 is used to obtain the lithological conditions and geological structures along the tunnel; determine several types of unfavorable geological bodies and their resistivity ranges according to the lithological conditions and geological structures; and obtain the resistivity of the geological body in front of the tunnel face; The finite element simulation module 22 is used to establish a tunnel finite element simulation model according to the bad geological bodies and their resistivity ranges; bad geological bodies are respectively set at different distances in front of the tunnel face of the tunnel finite element simulation model, and simulation is performed, and the type of each bad geological body and the distance between it and the tunnel face are recorded to obtain a number of samples and construct a data set; the samples include distance, sample resistivity and geological body type, and the sample resistivity is the resistivity of the bad geological body measured at the tunnel face; The model training module 23 is used to construct a variety of machine learning models; use the data set to train a variety of machine learning models, and evaluate the model performance, and use the optimal machine learning model as a tunnel disaster warning model; The determination module 24 is used to determine the type of geological body in front of the tunnel face according to the resistivity of the geological body in front of the tunnel face and the tunnel disaster early warning model; and predict the type of disaster occurring in front of the tunnel face according to the geological body type.

[0112] Figure 5 Each module in Figure 1 The execution unit of the method steps described in Figure 1 The detailed process of the method steps can be determined Figure 5 The specific method steps implemented by each module. Figure 1 The specific implementation methods and steps of the tunnel disaster warning device have been described in detail in Figure 1 The method steps described in the above are sufficient and will not be repeated here.

[0113] Figure 6 This is a structural block diagram of an electronic device provided by an embodiment of the present disclosure. Figure 6 , electronic devices may include Figure 5 The tunnel disaster warning device generally includes a processor 31 and a memory 32 .

[0114] The processor 31 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 31 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 31 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. The memory 32 may include one or more computer-readable storage media, which may be non-transitory. The memory 32 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices and flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 32 is used to store at least one instruction, which is used to be executed by the processor 31 to implement the tunnel disaster warning method performed by an electronic device provided in the method embodiment of the present application.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A tunnel disaster early warning method, characterized in that: include: Obtain the lithology conditions and geological structures along the tunnel; Determine several types of adverse geological bodies and their resistivity ranges based on lithological conditions and geological structures; Establish a tunnel finite element simulation model based on the adverse geological body and its resistivity range; Unfavorable geological bodies are set at different distances in front of the tunnel face of the tunnel finite element simulation model, and simulation is performed. The type of each unfavorable geological body and the distance between it and the tunnel face are recorded to obtain several samples and construct a data set. The samples include distance, sample resistivity and geological body type. The sample resistivity is the resistivity of the unfavorable geological body measured at the tunnel face. Build multiple machine learning models; Use the data set to train multiple machine learning models and evaluate the model performance, with the optimal machine learning model used as the tunnel disaster warning model; Obtain the resistivity of the geological body in front of the tunnel face; Determine the type of geological body in front of the tunnel face based on the resistivity of the geological body in front of the tunnel face and the tunnel disaster early warning model; Predict the type of disaster that may occur in front of the tunnel face based on the type of geological body.

2. The tunnel disaster early warning method according to claim 1, characterized in that: The bad geological body is a possible bad geological body determined according to the lithological conditions and geological structure, and the bad geological body includes water-rich faults, complex mixed strata and cavities; the resistivity range of the bad geological body is determined based on the resistivity of the geological structure obtained according to the geological structure and lithological conditions.

3. The tunnel disaster early warning method according to claim 1, characterized in that: During simulation, at least two spacings are set for each unfavorable geological body; at each spacing, each unfavorable geological body is divided into two groups according to the resistivity range of the unfavorable geological body and the resistivity increment is used. Increase the resistivity of the adverse geological body, perform multiple simulations, and obtain multiple groups of samples.

4. The tunnel disaster early warning method according to claim 1, characterized in that: When training multiple machine learning models, the data set is divided into a training set and a test set. The training set is used to train the multiple machine learning models, and the test set is used to test the trained multiple machine learning models to obtain the best machine learning model. In the process of training multiple machine learning models with the training set, the hyperparameters of each machine learning model are optimized using the improved sparrow optimization algorithm to obtain the optimal hyperparameter combination of each model, and the machine learning model with the optimal hyperparameter combination is used for testing.

5. The tunnel disaster early warning method according to claim 4, characterized in that: The method of optimizing the hyperparameters of each machine learning model using the improved sparrow optimization algorithm is as follows: The hyperparameter group of the machine learning model to be optimized is used to construct coordinates to represent the position of the sparrow in the sparrow optimization algorithm, and the performance index of the machine learning model to be optimized obtained by training according to the sparrow's coordinates is used as the fitness value of the sparrow; Classify sparrows in a sparrow population into finders, followers, and scouts; Initialize the sparrow population and calculate the fitness values ​​of all sparrows; Position update: Select the butterfly search strategy or the triangle walk strategy according to the size of the random number to update the discoverer's position; update the positions of the followers and scouts; Recalculate the fitness values ​​of all sparrows; Repeat the position update and fitness value calculation until the set number of iterations is reached; The sparrow position with the best fitness value is output as the optimal hyperparameter group of the machine learning model to be optimized.

6. The tunnel disaster early warning method according to claim 5, characterized in that: The cubic chaos mapping strategy is used to initialize the sparrow population. The initialization formula is as follows: Indicates the first The sparrow's The coordinates of the dimension, i.e. The first of the hyperparameter groups The value of the hyperparameter; Represents the hyperparameter group The upper limit of the hyperparameters; Represents the hyperparameter group The lower bound of the hyperparameters; For the A random number of sparrows.

7. The tunnel disaster early warning method according to claim 5, characterized in that: The method of updating the discoverer's position according to the butterfly search strategy and the triangle walk strategy is as follows: Set a safety threshold. When the random number is not greater than the safety threshold, use the butterfly search strategy to update the discoverer's position; otherwise, use the triangle walk strategy to update the discoverer's position. The update formula of the butterfly search strategy is as follows: Indicates A sparrow in the The position is updated before the iteration Dimensional coordinates; Indicates A sparrow in the After the first discoverer position update in the iteration Dimensional coordinates; Indicates A sparrow in the After the second position update in the iteration process Dimensional coordinates; for The sparrow with the best fitness value in the iteration Dimensional coordinates; is a random number; is the maximum number of iterations; Indicates the number of iterations; Indicates the amount of aroma perception; For the The amount of scent perceived by a sparrow; is the stimulus intensity, The optimal fitness value of the current sparrow population equal; It's the fragrance. It is a form of perception. , is a constant; is a constant; The update formula of the triangle walk strategy is as follows: 、 、 Indicates A sparrow in the The position is updated before the iteration The vector formed by the coordinates of the dimension; Indicates A sparrow in the After the first discoverer position update in the iteration The vector formed by the coordinates of the dimension; Indicates A sparrow in the After the second position update in the iteration process Dimensional coordinates; for The individual with the best fitness value in the iteration The vector formed by the coordinates of the dimension; is a random number that follows a normal distribution; To include elements, All the elements in are 1. The size of is equal to the number of dimensions of the sparrow coordinates; is a unit vector with a random direction.

8. A tunnel disaster warning device, characterized in that: include: Acquisition module, used to obtain the lithology conditions and geological structures along the tunnel; Determine several types of adverse geological bodies and their resistivity ranges based on lithological conditions and geological structures; Obtain the resistivity of the geological body in front of the tunnel face; The finite element simulation module is used to establish a tunnel finite element simulation model based on the unfavorable geological body and its resistivity range; unfavorable geological bodies are respectively set at different distances in front of the tunnel face of the tunnel finite element simulation model, and simulation is performed, and the type of each unfavorable geological body and the distance between it and the tunnel face are recorded to obtain a number of samples and construct a data set; the samples include distance, sample resistivity and geological body type, and the sample resistivity is the resistivity of the unfavorable geological body measured at the tunnel face; Model training module, used to build multiple machine learning models; use data sets to train multiple machine learning models, evaluate model performance, and use the optimal machine learning model as the tunnel disaster warning model; The determination module is used to determine the type of geological body in front of the tunnel face according to the resistivity of the geological body in front of the tunnel face and the tunnel disaster early warning model; and predict the type of disaster occurring in front of the tunnel face according to the geological body type.

9. An electronic device, characterized in that: The electronic equipment includes the tunnel disaster warning device as claimed in claim 8.

10. A computer-readable storage medium, characterized in that: At least one program code is stored in the computer-readable storage medium, and the program code is executed by a processor to implement the tunnel disaster warning method according to any one of claims 1 to 7.

Images