Underground cavity sealing property evaluation method and device based on multilayer perceptron

Abstract

The application discloses a kind of underground cavity tightness evaluation method and device based on multilayer perceptron, comprising the following steps: constructing the sample data set for representing the multivariate feature parameters of underground cavity tightness and corresponding tightness grade label and carrying out pretreatment and feature screening to obtain feature data, and the feature vector is obtained by standardizing the feature data;Multilayer perceptron evaluation model is constructed, and the multilayer perceptron evaluation model is trained and verified using the feature vector;The multivariate feature parameters of the cavity to be evaluated are processed to obtain the feature vector to be evaluated and input into the trained multilayer perceptron evaluation model to evaluate the tightness grade, obtain the prediction probability of each sealing grade and generate engineering intervention instruction, and execute corresponding engineering intervention operation on the cavity to be evaluated according to the engineering intervention instruction.The application reduces the dependence on human subjective experience, improves the objectivity, consistency and evaluation efficiency of the evaluation result, and can effectively solve the actual geological hazard problem.

Description

Technical Field

[0001] This invention relates to the field of underground cavity sealing technology, specifically to a method and apparatus for evaluating the sealing performance of underground cavities based on a multilayer sensor. Background Technology

[0002] Urbanization and long-term resource extraction have resulted in numerous abandoned mines, salt caverns, civil defense projects, and natural karst caves in my country, creating numerous underground cavities. If these cavities lack proper sealing, they can easily trigger geological disasters such as surface subsidence and uneven foundation settlement, threatening the safety of superstructures. Inadequately sealed cavities can also become conduits for groundwater pollution, altering regional groundwater dynamics and jeopardizing drinking water safety. Under the dual-carbon strategy, these abandoned underground cavities are often used for carbon dioxide sequestration, energy storage, and waste disposal; inadequate sealing will lead to gas leaks and pollutant spills, causing environmental damage and economic losses.

[0003] Existing methods for assessing airtightness largely rely on expert experience, are highly subjective, and are difficult to standardize. On-site investigations and physical tests are costly and time-consuming, failing to meet the needs for large-scale, rapid screening and monitoring. Current machine learning-based geological assessment methods do not address the airtightness design of the cavity itself, only outputting evaluation results, and cannot improve engineering parameters based on these results.

[0004] Therefore, how to overcome the shortcomings of existing technologies using effective methods has become an urgent technical challenge. Summary of the Invention

[0005] The purpose of this invention is to address the above-mentioned problems by providing a method and apparatus for evaluating the sealing performance of underground cavities based on a multilayer sensor.

[0006] The technical solution of this invention is: a method for evaluating the sealing performance of underground cavities based on a multilayer sensor, comprising the following steps:

[0007] S1: Construct a sample dataset, which includes multi-dimensional feature parameters and corresponding sealing level labels for characterizing the sealing performance of underground cavities. The multi-dimensional feature parameters include geometric structural parameters of underground cavities, surrounding rock condition parameters, cavity wall state parameters, seepage characteristic parameters, and groundwater environment parameters. S2: Preprocess the sample dataset and use the mutual information method to filter features in the preprocessed sample dataset to obtain feature data, and standardize the feature data to obtain feature vectors; S3: Construct a multilayer perceptron evaluation model, and train and validate the multilayer perceptron evaluation model using the feature vector; the multilayer perceptron evaluation model includes an input layer, at least one hidden layer, and an output layer; wherein, the activation function of the output layer is a classification function; S4: Obtain the multi-dimensional feature parameters of the cavity to be evaluated and preprocess the multi-dimensional feature parameters to obtain the feature vector to be evaluated. Input the feature vector to be evaluated into the trained multilayer perceptron evaluation model to evaluate the sealing level of the cavity to be evaluated and obtain the predicted probability of each sealing level. S5: Generate engineering intervention instructions based on the predicted probabilities of each sealing level, and perform corresponding engineering intervention operations on the cavity to be evaluated according to the engineering intervention instructions.

[0008] As an improvement to this embodiment of the invention, in step S1, the geometric structural parameters include burial depth and equivalent volume; the surrounding rock condition parameters include rock mass quality index RQD and joint fracture density per unit length; the cavity wall state parameters include uniaxial compressive strength of the wall rock and average aperture of major fractures; the seepage characteristic parameters include total flow rate of seepage points per unit time; and the groundwater environment parameters include minimum distance from the aquifer and water content of the surrounding rock.

[0009] As an improvement to this embodiment of the invention, in step S2, the step of using mutual information method to perform feature screening on the preprocessed sample dataset to obtain feature data specifically includes: quantitatively calculating the correlation between each feature in the multi-dimensional feature parameters and the sealing level label, and screening out several key features that contribute the most to the sealing evaluation as feature data.

[0010] As an improvement to this embodiment of the invention, in step S3, the number of input layer nodes of the multilayer perceptron evaluation model is consistent with the dimension of the feature vector, and the output layer contains 4 nodes, which correspond to Level I, Level II, Level III and Level IV, respectively. Level I represents good sealing, Level II represents basic sealing, Level III represents poor sealing, and Level IV represents instability risk.

[0011] As an improvement to this embodiment of the invention, in step S3, the training process adopts a labeled smooth cross-entropy loss function, uses the Adam optimizer for parameter optimization, and introduces an early stopping mechanism and a learning rate decay strategy.

[0012] As an improvement of this embodiment of the invention, the early stopping mechanism stops training when the value of the validation set loss function does not decrease for several consecutive training cycles, and the learning rate decay strategy halves the learning rate when the value of the validation set loss function does not improve for several consecutive training cycles.

[0013] As an improvement to this embodiment of the invention, in step S5, the generation of engineering intervention instructions based on the predicted probabilities of each sealing level specifically includes: using the predicted probabilities of each sealing level output by the multilayer perceptron evaluation model as feedforward independent variables, and substituting them into the preset engineering control equation to calculate the target grouting pressure P or gas storage injection rate Q.

[0014] As an improvement to this embodiment of the invention, the formula for calculating the target grouting pressure P is as follows: ,in, The foundation grouting pressure is the pressure required to fill the cavity under the geological conditions described above. , , All are nonlinear pressure compensation coefficients; , , These represent the predicted probabilities of sealing grades II, III, and IV, respectively.

[0015] As an improvement to this embodiment of the invention, the formula for calculating the gas injection rate Q is as follows: ,in, The maximum rated flow rate of the equipment. This is the traffic attenuation penalty coefficient; , These represent the predicted probabilities for sealing levels III and IV, respectively.

[0016] To achieve one of the above-mentioned objectives, one embodiment of the present invention provides a device for evaluating the sealing performance of underground cavities based on a multilayer sensor, comprising the following modules: The dataset construction module is used to construct a sample dataset, which includes multi-dimensional feature parameters and corresponding sealing level labels for characterizing the sealing performance of underground cavities. The multi-dimensional feature parameters include geometric structural parameters of underground cavities, surrounding rock condition parameters, cavity wall state parameters, seepage characteristic parameters, and groundwater environment parameters. The data processing module is used to preprocess the sample dataset and use the mutual information method to filter features of the preprocessed sample dataset to obtain feature data, and to standardize the feature data to obtain feature vectors. The model building module is used to build a multilayer perceptron evaluation model and to train and validate the multilayer perceptron evaluation model using the feature vectors. The multilayer perceptron evaluation model includes an input layer, at least one hidden layer, and an output layer. The activation function of the output layer is a classification function. The sealing performance evaluation module is used to acquire multi-dimensional feature parameters of the cavity to be evaluated and preprocess the multi-dimensional feature parameters to obtain the feature vector to be evaluated. The feature vector to be evaluated is then input into the trained multilayer perceptron evaluation model to evaluate the sealing performance level of the cavity to be evaluated and obtain the predicted probability of each sealing level. The instruction execution module is used to generate engineering intervention instructions based on the predicted probability of each sealing level, and to perform corresponding engineering intervention operations on the cavity to be evaluated according to the engineering intervention instructions.

[0017] The underground cavity sealing evaluation method based on multilayer perceptron provided in this invention has the following advantages: This invention realizes the underground cavity sealing evaluation through the multilayer perceptron evaluation model, reduces the reliance on human subjective experience, improves the objectivity, consistency and evaluation efficiency of the evaluation results, and controls the on-site equipment to adjust the parameters according to the evaluation results, which can effectively solve actual geological hidden danger problems. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the underground cavity sealing performance evaluation method based on a multilayer sensor as described in this invention. Figure 2 This is a schematic diagram of the underground cavity sealing evaluation device based on a multilayer sensor as described in this invention. Detailed Implementation

[0019] The present invention will now be described in detail with reference to the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present invention, and any structural, methodological, or functional modifications made by those skilled in the art based on these embodiments are included within the scope of protection of the present invention.

[0020] The scope of the embodiments described herein includes the entire scope of the claims and all available equivalents thereof. Throughout this document, the terms “first,” “second,” etc., are used only to distinguish one element from another without requiring or implying any actual relationship or order between the elements. Indeed, a first element can also be referred to as a second element, and vice versa. Furthermore, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a structure, apparatus, or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a structure, apparatus, or device. Without further limitations, an element defined by the phrase “comprising one…” does not exclude the presence of other identical elements in the structure, apparatus, or device that includes said element. The various embodiments described herein are presented in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably.

[0021] The terms "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer" used in this document to indicate orientation or positional relationships are based on the orientation or positional relationships shown in the accompanying drawings and are used only for the convenience of describing this document and simplifying the description. They do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting the invention. In the description herein, unless otherwise specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly. For example, they can refer to mechanical or electrical connections, or internal connections between two elements, or direct connections or indirect connections through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms according to the specific circumstances.

[0022] This invention provides a method for evaluating the sealing performance of underground cavities based on a multilayer sensor, such as... Figure 1 As shown, it includes the following steps: Step S1: Construct a sample dataset, which includes multi-dimensional feature parameters for characterizing the sealing performance of underground cavities and corresponding sealing performance level labels. The multi-dimensional feature parameters include the geometric structure parameters of the underground cavity, surrounding rock condition parameters, cavity wall state parameters, seepage characteristic parameters, and groundwater environment parameters. In some specific implementations, the geometric structural parameters include burial depth and equivalent volume. The burial depth determines the stress environment and groundwater pressure conditions of the cavity, while the equivalent volume reflects the overall size of the cavity. The surrounding rock condition parameters include the rock mass quality index (RQD) and the joint and fracture density per unit length. The RQD measures the integrity of the surrounding rock, while the joint and fracture density reflects the degree of rock fragmentation. The cavity wall state parameters include the uniaxial compressive strength of the wall rock and the average aperture of the main fractures. The uniaxial compressive strength reflects the bearing capacity and resistance to damage of the wall rock mass, while the average aperture of the main fractures determines the size of the fluid leakage channels. The seepage characteristic parameters include the total flow rate at the leakage points per unit time, used to quantify the current leakage degree of the cavity. The groundwater environment parameters include the minimum distance to the aquifer and the water content of the surrounding rock. The distance to the aquifer determines the groundwater recharge intensity, while the water content of the surrounding rock reflects the water saturation state of the rock mass.

[0023] After the multi-dimensional characteristic parameters are collected, multiple professionals, in conjunction with engineering specifications, determine the sealing performance level according to the leakage rate and strain value. The sealing performance level includes Level I, Level II, Level III, and Level IV, where Level I indicates good sealing, Level II indicates basic sealing, Level III indicates poor sealing, and Level IV indicates risk of instability. The sealing performance level results are then converted into sealing performance level labels. Specifically, the sealing performance level labels can be 1, 2, 3, and 4, corresponding one-to-one with Level I, Level II, Level III, and Level IV, respectively.

[0024] Step S2: Preprocess the sample dataset and use mutual information to filter features in the preprocessed dataset to obtain feature data. Here, mutual information is used to measure the statistical correlation between two random variables and can assess the nonlinear association strength between a single feature and the target label. Specifically, it includes: quantitatively calculating the correlation between each feature in the multi-dimensional feature parameters and the sealing level label, and selecting several key features that contribute the most to the sealing evaluation as feature data; then, the feature data is standardized to obtain feature vectors. The Z-score standardization method can be used to standardize all feature data to eliminate dimensional differences and make the feature vectors conform to a standard normal distribution with a mean of 0 and a standard deviation of 1. Specifically, the Z-score standardization method is... ,in, These are the original feature parameters. It is the mean of the feature parameter on the sample dataset. It is the standard deviation of the feature parameter on the sample dataset. These are the standardized feature values.

[0025] In practice, preprocessing the sample dataset specifically includes: identifying and removing outliers that exceed the geologically reasonable range; and imputing missing values ​​using the median of the same parameter within the same geological unit. If the number of samples is small, Gaussian noise augmentation techniques can be used to augment the sample dataset, expanding the number of sample data without changing the label distribution and improving the model's generalization ability.

[0026] Step S3: Construct a multilayer perceptron evaluation model, and train and validate the multilayer perceptron evaluation model using the feature vector; the multilayer perceptron evaluation model includes an input layer, at least one hidden layer, and an output layer; wherein, the activation function of the output layer is a classification function; the number of nodes in the input layer of the multilayer perceptron evaluation model is consistent with the dimension of the feature vector, for lossless data reception. In this invention, the dimension of the feature vector and the number of nodes in the input layer are both 8, and the output layer contains 4 nodes, which correspond to levels I, II, III, and IV, respectively.

[0027] Preferably, a batch normalization layer is added after the input layer of the multilayer perceptron evaluation model of the present invention to accelerate the model training convergence speed. The hidden layers consist of two layers: the first layer has 32 neurons and the second layer has 16 neurons. Both hidden layers employ the ReLU activation function to introduce nonlinear transformation capability. The ReLU activation function is... ,in, These are the input values ​​of the hidden layer neurons. The output values ​​of the hidden layer neurons can improve the model's fitting accuracy to the coupling relationship between geomechanics and seepage. Simultaneously, setting a Dropout layer with a dropout rate of 0.2 after each hidden layer and applying L2 weighted regularization constraints can effectively reduce the risk of overfitting and improve the model's generalization ability. The output layer uses a Softmax activation function to transform the network output into a probability distribution of each sealing level. ,in, Let i be the input value of the i-th node in the output layer. Let be the probability of the sealing level corresponding to the i-th node in the output layer. This is the total number of sealing ratings. In this invention, Then, the feature vectors are input into the model for training and validation.

[0028] In some specific implementations, the training process employs a labeled, smoothed cross-entropy loss function. ,in, It is the one-hot code of the genuine seal label. It is the th in the model's predicted sealing probability distribution The probability of each level, This is the total number of sealing ratings. Yes, that's natural. This is the cross-entropy loss value, used to measure the difference between the model's predicted sealing probability distribution and the true label. The Adam optimizer is used for parameter optimization, and an early stopping mechanism and a learning rate decay strategy are introduced. The early stopping mechanism stops training when the validation set loss function value does not decrease for a certain number of consecutive training epochs. The learning rate decay strategy halves the learning rate when the validation set loss function value does not improve for a certain number of consecutive training epochs. Specifically, the early stopping mechanism automatically terminates training and restores the parameters to the optimal sealing discrimination performance when the validation set loss function value does not decrease for 5 consecutive training epochs; the learning rate decay strategy halves the current learning rate when the validation set loss function value does not improve for 3 consecutive training epochs, with the lower limit of the learning rate set to 1×10⁻⁶. -6 It is used to evaluate the model's discriminative performance on the validation set for the level of sealing.

[0029] After training, the final sealing discrimination generalization performance of the model is evaluated using a reserved independent test set that was not involved in training and parameter tuning, thus obtaining the trained multilayer perceptron evaluation model.

[0030] Step S4: Obtain the multi-dimensional feature parameters of the cavity to be evaluated and preprocess the multi-dimensional feature parameters to obtain the feature vector to be evaluated. Input the feature vector to be evaluated into the trained multilayer perceptron evaluation model to evaluate the sealing level of the cavity to be evaluated and obtain the predicted probability of each sealing level. In practice, the multi-dimensional feature parameters of the cavity to be evaluated are collected in the same way as the sample dataset constructed in step S1, in terms of parameter type, source, and data format. This ensures that the features are homogeneous and comparable. Then, the same data preprocessing procedure as in step S2 is used to process the multi-dimensional feature parameters to obtain the feature vector to be evaluated. This feature vector is then input into the trained multilayer perceptron evaluation model. The forward computation of the multilayer perceptron evaluation model outputs the predicted probabilities of each sealing level corresponding to the cavity to be evaluated.

[0031] Step S5: Generate engineering intervention instructions based on the predicted probabilities of each sealing level, and perform corresponding engineering intervention operations on the cavity to be evaluated according to the engineering intervention instructions. Specifically, this includes: using the predicted probabilities of each sealing level output by the multilayer perceptron evaluation model as feedforward independent variables, and substituting them into the preset engineering control equation to calculate the target grouting pressure P or gas injection rate Q. Specifically, the calculation formula for the target grouting pressure P is: ,in, The foundation grouting pressure is the pressure required to fill the cavity under the geological conditions described above. , , All are nonlinear pressure compensation coefficients. Preferably, the nonlinear pressure compensation coefficients satisfy their empirical value ranges as follows: , , ; , , These represent the predicted probabilities of sealing grades II, III, and IV, respectively. The formula for calculating the gas injection rate Q is: ,in, The maximum rated flow rate of the equipment. The flow attenuation penalty coefficient is preferably a specific value used to define the flow attenuation penalty coefficient. ; , These represent the predicted probabilities for sealing levels III and IV, respectively.

[0032] In practice, the sealing performance level with the highest probability can be selected as the final evaluation result, and a structured evaluation report can be generated. The structured evaluation report includes at least: key feature values, predicted probability distributions for each sealing performance level, and the final assessed sealing performance level. Simultaneously, the engineering intervention command is sent to the engineering equipment corresponding to the cavity to be evaluated to perform the corresponding engineering intervention operation. Specifically, this can involve driving the grouting pump, the gas compressor frequency converter, and the wellhead injection valve to perform sealing improvement operations.

[0033] Example This embodiment uses an abandoned coal mine in Huainan as the evaluation object to verify the underground cavity sealing evaluation method described in this invention.

[0034] Step S1: Establish a sample dataset for model training and testing. Extract multi-dimensional feature parameters of the samples from engineering survey reports and monitoring records. Divide the sealing performance into four levels and convert them into sealing performance level labels.

[0035] Step S2: Clean the sample dataset, identify and correct or remove outliers, and impute missing data using the median. Calculate the correlation between each feature and the sealing level using mutual information, and select 8 key features. Use the Z-score standardization method to eliminate differences in the units and numerical ranges of the feature parameters, transforming the features into standard normally distributed data.

[0036] Step S3: Construct a multilayer perceptron evaluation model, including an input layer, two hidden layers, and an output layer. The input layer has 8 nodes, followed by a batch normalization layer. The first hidden layer has 32 neurons, and the second hidden layer has 16 neurons. Each hidden layer uses the ReLU activation function, followed by a Dropout layer with a dropout rate of 0.2 and L2 weighted regularization. The output layer has 4 nodes corresponding to 4 sealing levels, using the Softmax activation function to output the probability distribution of each level. A five-fold cross-validation dataset is used, and a labeled smoothing classification cross-entropy loss function is selected. The model is trained using the Adam optimizer with an initial learning rate of 0.0005. An early stopping mechanism and a learning rate decay strategy are introduced during training. Training is terminated if the validation set loss does not decrease for 5 consecutive epochs, and the learning rate is halved if there is no improvement for 3 consecutive epochs. The lower limit of the learning rate is set to 1×10⁻⁶. -6 After training, the generalization performance of the model was evaluated using an independent test set and a five-fold cross-validation method. The relevant evaluation results are shown in Tables 1 to 4. After the model was validated, the optimal parameter model was used as the evaluation model of the trained multilayer perceptron.

[0037] Table 1. Classification of Independent Test Sets

[0038] Table 2. Confusion Matrix of Independent Test Sets

[0039] As can be seen from Tables 1 and 2, the model achieved a precision, recall, and F1 score of 1.00 for all four sealing levels on the independent test set, with no misclassified samples, demonstrating that the model has accurate discrimination ability on unseen test data.

[0040] Table 3 Summary Classification Table of Five-Fold Cross-Validation

[0041] Table 4. Confusion Matrix of Five-Fold Cross-Validation

[0042] As can be seen from Tables 3 and 4, after five-fold cross-validation, all evaluation indicators of the model are close to 1.00, with only one classification bias. This indicates that the model has stable generalization ability and reliable evaluation results in small sample scenarios, and can be used for the actual evaluation of underground cavity sealing.

[0043] Step S4: Collect multi-dimensional feature data of an abandoned coal mine A in Huainan area on-site. Following the same preprocessing procedure as the training phase, complete data cleaning, feature selection, and standardization to generate feature vectors to be evaluated. Input the feature vectors to be evaluated into the trained multilayer perceptron evaluation model, and obtain the predicted probabilities of the four sealing levels corresponding to coal mine A through forward calculation. The results are shown in Table 5.

[0044] Table 5 Predicted probability of cavity sealing performance to be evaluated

[0045] Table 5 shows the probabilities of each sealing level of the mine to be evaluated output by the model. The level with the highest probability is selected as the final sealing evaluation result. The sealing level of the coal mine A is Level I.

[0046] Step S5: Calculate the target grouting pressure and gas injection rate based on the predicted probabilities of each sealing level output by the model and the preset control equations. Send the calculated parameters to the on-site engineering equipment to drive the grouting pump, gas compressor frequency converter, and wellhead injection valve for sealing improvement operations. The target grouting pressure calculation process is as follows: Set the basic grouting pressure... And retrieve the nonlinear pressure compensation coefficient. , , Substituting into the grouting pressure control equation, the target pressure command is calculated. The calculation process for the gas injection rate is as follows: Set the maximum rated flow rate of the gas injection equipment. Rate decay penalty coefficient Substituting into the gas injection rate control equation, the gas injection rate is calculated. .

[0047] This invention uses a multilayer perceptron evaluation model to evaluate the sealing performance of underground cavities. It can integrate multi-source geological and engineering data to objectively quantify the sealing status of cavities, effectively reducing the reliance on manual experience in traditional methods, and reducing the cost and cycle of on-site investigation and physical testing. It can also achieve sealing performance evaluation of multiple cavities over a large area, identify hidden leakage risks, and map the evaluation results to engineering control parameters such as grouting pressure and gas injection rate, forming a complete control process from evaluation to engineering improvement.

[0048] This invention also provides a device for evaluating the sealing performance of underground cavities based on a multilayer sensor, such as... Figure 2 As shown, it includes the following modules: The dataset construction module 201 is used to construct a sample dataset, which includes multi-dimensional feature parameters and corresponding sealing level labels for characterizing the sealing performance of underground cavities. The multi-dimensional feature parameters include geometric structural parameters of underground cavities, surrounding rock condition parameters, cavity wall state parameters, seepage characteristic parameters, and groundwater environment parameters. Data processing module 202 is used to preprocess the sample dataset and use mutual information method to filter features of the preprocessed sample dataset to obtain feature data, and to standardize the feature data to obtain feature vectors; The model building module 203 is used to build a multilayer perceptron evaluation model and to train and validate the multilayer perceptron evaluation model using the feature vector; the multilayer perceptron evaluation model includes an input layer, at least one hidden layer, and an output layer; wherein the activation function of the output layer is a classification function; The sealing performance evaluation module 204 is used to obtain multi-dimensional feature parameters of the cavity to be evaluated and preprocess the multi-dimensional feature parameters to obtain the feature vector to be evaluated. The feature vector to be evaluated is then input into the trained multilayer perceptron evaluation model to evaluate the sealing performance level of the cavity to be evaluated and obtain the predicted probability of each sealing level. The instruction execution module 205 is used to generate engineering intervention instructions based on the predicted probability of each sealing level, and to perform corresponding engineering intervention operations on the cavity to be evaluated according to the engineering intervention instructions.

[0049] This invention can be an apparatus, method, and / or computer program product. A computer program product may include a readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of the invention.

[0050] Storage media can be tangible devices that hold and store instructions for use by instruction execution devices. Storage media can include, but are not limited to, electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination thereof.

[0051] It should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This way of describing the specification is only for clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0052] The detailed descriptions listed above are merely specific descriptions of feasible embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. All equivalent embodiments or modifications made without departing from the spirit of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for evaluating the sealing performance of underground cavities based on a multilayer perceptron, characterized in that, Includes the following steps: S1: Construct a sample dataset, which includes multi-dimensional feature parameters and corresponding sealing level labels for characterizing the sealing performance of underground cavities. The multi-dimensional feature parameters include geometric structural parameters of underground cavities, surrounding rock condition parameters, cavity wall state parameters, seepage characteristic parameters, and groundwater environment parameters. S2: Preprocess the sample dataset and use the mutual information method to filter features in the preprocessed sample dataset to obtain feature data, and standardize the feature data to obtain feature vectors; S3: Construct a multilayer perceptron evaluation model, and train and validate the multilayer perceptron evaluation model using the feature vector; the multilayer perceptron evaluation model includes an input layer, at least one hidden layer, and an output layer; wherein, the activation function of the output layer is a classification function; S4: Obtain the multi-dimensional feature parameters of the cavity to be evaluated and preprocess the multi-dimensional feature parameters to obtain the feature vector to be evaluated. Input the feature vector to be evaluated into the trained multilayer perceptron evaluation model to evaluate the sealing level of the cavity to be evaluated and obtain the predicted probability of each sealing level. S5: Generate engineering intervention instructions based on the predicted probabilities of each sealing level, and perform corresponding engineering intervention operations on the cavity to be evaluated according to the engineering intervention instructions.

2. The method for evaluating the sealing performance of underground cavities according to claim 1, characterized in that, In step S1, the geometric structural parameters include burial depth and equivalent volume; the surrounding rock condition parameters include rock mass quality index RQD and joint fissure density per unit length; the cavity wall state parameters include uniaxial compressive strength of the wall rock and average aperture of major fissures; the seepage characteristic parameters include total flow rate of seepage points per unit time; and the groundwater environment parameters include minimum distance from the aquifer and water content of the surrounding rock.

3. The method for evaluating the sealing performance of underground cavities according to claim 1, characterized in that, In step S2, the step of using mutual information to filter features in the preprocessed sample dataset to obtain feature data specifically includes: quantitatively calculating the correlation between each feature in the multi-dimensional feature parameters and the sealing level label, and selecting several key features that contribute the most to the sealing evaluation as feature data.

4. The method for evaluating the sealing performance of underground cavities according to claim 1, characterized in that, In step S3, the number of input layer nodes of the multilayer perceptron evaluation model is consistent with the dimension of the feature vector, and the output layer contains 4 nodes, which correspond to Level I, Level II, Level III and Level IV, respectively. Level I indicates good sealing, Level II indicates basic sealing, Level III indicates poor sealing, and Level IV indicates risk of instability.

5. The method for evaluating the sealing performance of underground cavities according to claim 1, characterized in that, In step S3, the training process uses a labeled smooth cross-entropy loss function, optimizes parameters using the Adam optimizer, and introduces an early stopping mechanism and a learning rate decay strategy.

6. The method for evaluating the sealing performance of underground cavities according to claim 5, characterized in that, The early stopping mechanism stops training when the value of the validation set loss function does not decrease for several consecutive training cycles, and the learning rate decay strategy halves the learning rate when the value of the validation set loss function does not improve for several consecutive training cycles.

7. The method for evaluating the sealing performance of underground cavities according to claim 1, characterized in that, In step S5, the process of generating engineering intervention instructions based on the predicted probabilities of each sealing level specifically includes: using the predicted probabilities of each sealing level output by the multilayer perceptron evaluation model as feedforward independent variables, and substituting them into the preset engineering control equations to calculate the target grouting pressure P or gas storage injection rate Q.

8. The method for evaluating the sealing performance of underground cavities according to claim 7, characterized in that, The formula for calculating the target grouting pressure P is: ,in, The foundation grouting pressure is the pressure required to fill the cavity under the geological conditions described above. , , All are nonlinear pressure compensation coefficients; , , These represent the predicted probabilities of sealing grades II, III, and IV, respectively.

9. The method for evaluating the sealing performance of underground cavities according to claim 7, characterized in that, The formula for calculating the gas injection rate Q is: ,in, The maximum rated flow rate of the equipment. This is the traffic attenuation penalty coefficient; , These represent the predicted probabilities for sealing levels III and IV, respectively.

10. A device for evaluating the sealing performance of underground cavities based on a multilayer sensor, characterized in that, Includes the following modules: The dataset construction module is used to construct a sample dataset, which includes multi-dimensional feature parameters and corresponding sealing level labels for characterizing the sealing performance of underground cavities. The multi-dimensional feature parameters include geometric structural parameters of underground cavities, surrounding rock condition parameters, cavity wall state parameters, seepage characteristic parameters, and groundwater environment parameters. The data processing module is used to preprocess the sample dataset and use the mutual information method to filter features of the preprocessed sample dataset to obtain feature data, and to standardize the feature data to obtain feature vectors. The model building module is used to build a multilayer perceptron evaluation model and to train and validate the multilayer perceptron evaluation model using the feature vectors. The multilayer perceptron evaluation model includes an input layer, at least one hidden layer, and an output layer. The activation function of the output layer is a classification function. The sealing performance evaluation module is used to acquire multi-dimensional feature parameters of the cavity to be evaluated and preprocess the multi-dimensional feature parameters to obtain the feature vector to be evaluated. The feature vector to be evaluated is then input into the trained multilayer perceptron evaluation model to evaluate the sealing performance level of the cavity to be evaluated and obtain the predicted probability of each sealing level. The instruction execution module is used to generate engineering intervention instructions based on the predicted probability of each sealing level, and to perform corresponding engineering intervention operations on the cavity to be evaluated according to the engineering intervention instructions.

Images