Tunnel coal and gas risk grading prediction method and device

Abstract

The invention provides a tunnel coal and gas risk grading prediction method and device, and relates to the technical field of tunnel engineering prediction, and the method comprises the steps: obtaining historical data; performing data cleaning processing on the historical data, and performing data division on the cleaned historical data to obtain a data set; performing sine and cosine search strategy iterative training optimization on the initial sparrow search algorithm based on a preset cubic chaotic mapping method to obtain an improved sparrow search algorithm; performing hyper-parameter optimization on a preset deep neural network model according to an improved sparrow search algorithm to obtain an improved deep neural network model; inputting the data set into an improved deep neural network model for iterative training, and performing feature parameter screening on the data set to obtain a prediction model; and predicting preset tunnel coal and gas monitoring data based on the prediction model to obtain a risk grading result. According to the method, the problem that a prediction result has relatively high randomness and inaccuracy is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel engineering prediction, and in particular to a method and device for predicting the graded hazard of coal and gas in a tunnel. Background Art

[0002] In the field of tunnel engineering prediction technology, the single index method and the comprehensive index method are currently the main methods used. Both methods preset the critical value of a key indicator and compare it with the actual monitoring data to determine whether it exceeds the limit for prediction. However, as a porous medium, the physical and mechanical properties of coal are easily affected by uncontrollable factors such as artificial mining activities and the heterogeneity of the coal environment, which makes it difficult to accurately predict the behavior pattern of coal. At present, most research focuses on the prediction of hazards in the coal mining field. However, when these methods are applied to tunnel engineering, the introduction of specific parameters such as tunnel excavation methods, face area, support methods, and surrounding rock lithology makes data collection more complicated. This requirement increases the difficulty of data collection and also puts higher requirements on the construction of prediction models. Due to the diversity and uncertainty of tunnel parameters, prediction models often find it difficult to accurately capture the interactions between parameters, resulting in large randomness and inaccuracy in prediction results.

[0003] Therefore, there is an urgent need for a tunnel coal and gas hazard classification prediction method and device to solve the problem of large randomness and inaccuracy in the prediction results. Summary of the invention

[0004] The purpose of the present invention is to provide a method and device for predicting the hazard classification of coal and gas in a tunnel to improve the above problems. In order to achieve the above purpose, the technical solution adopted by the present invention is as follows: In a first aspect, the present application provides a tunnel coal and gas hazard classification prediction method, comprising: Acquiring historical data, wherein the historical data includes gas parameter data, coal body parameter data, tunnel parameter data and hazard classification labels; Performing data cleaning on the historical data, and obtaining a data set by dividing the cleaned historical data; Based on the preset cubic chaos mapping method, the initial sparrow search algorithm is iteratively trained and optimized with sine and cosine search strategies to obtain an improved sparrow search algorithm. According to the improved sparrow search algorithm, hyperparameter optimization is performed on the preset deep neural network model to obtain an improved deep neural network model; Inputting the data set into the improved deep neural network model for iterative training, and obtaining a prediction model by screening feature parameters of the data set; Based on the prediction model, the preset tunnel coal and gas monitoring data are predicted to obtain the hazard classification result.

[0005] In a second aspect, the present application also provides a tunnel coal and gas hazard classification prediction device, comprising: An acquisition model is used to acquire historical data, wherein the historical data includes gas parameter data, coal body parameter data, tunnel parameter data and hazard classification labels; A cleaning model is used to perform data cleaning on the historical data, and obtain a data set by dividing the cleaned historical data; The first optimization model is used to perform iterative training optimization of the sine and cosine search strategy on the initial sparrow search algorithm based on a preset cubic chaos mapping method to obtain an improved sparrow search algorithm; A second optimization model is used to optimize the hyperparameters of a preset deep neural network model according to the improved sparrow search algorithm to obtain an improved deep neural network model; A training model is used to input the data set into the improved deep neural network model for iterative training, and obtain a prediction model by screening feature parameters of the data set; The prediction model is used to predict the preset tunnel coal and gas monitoring data based on the prediction model to obtain a hazard classification result.

[0006] In a third aspect, the present application also provides a tunnel coal and gas hazard classification prediction device, comprising: Memory for storing computer programs; A processor is used to implement the steps of the tunnel coal and gas hazard classification prediction method when executing the computer program.

[0007] In a fourth aspect, the present application further provides a medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned method for predicting the hazard classification of coal and gas in a tunnel are implemented.

[0008] The beneficial effects of the present invention are: The present invention improves the initial sparrow search algorithm through the cubic chaos mapping method and the sinusoidal search strategy, thereby improving the uniformity and randomness of the population distribution, and helping to enhance the exploration ability of the discoverer and avoid premature convergence; the deep neural network model is optimized by improving the sparrow search algorithm, simulating the sparrow's foraging and anti-predation behavior. When the improved sparrow search algorithm optimizes the hyperparameters of the deep neural network model, it can perform a more efficient search in the hyperparameter space, quickly and accurately find a set of better hyperparameter combinations, and obtain an improved deep neural network model. In the construction of the prediction model, by selecting features and analyzing the role of each feature in the model prediction, the prediction performance of the model can be maintained while reducing the number of features. The present invention solves the problem of large randomness and inaccuracy in the prediction results.

[0009] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or be understood by implementing the embodiments of the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 It is a schematic diagram of the process of the tunnel coal and gas hazard classification prediction method described in an embodiment of the present invention; Figure 2 It is a schematic diagram of the structure of the tunnel coal and gas hazard classification prediction equipment described in an embodiment of the present invention.

[0012] Markings in the figure: 800, tunnel coal and gas hazard classification prediction equipment; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component. DETAILED DESCRIPTION

[0013] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0014] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0015] Embodiment 1: This embodiment provides a tunnel coal and gas hazard classification prediction method, see Figure 1 , the figure shows that the method includes steps S1 to S6, including: S1: Acquire historical data, wherein the historical data includes gas parameter data, coal body parameter data, tunnel parameter data and hazard classification labels; In this step, the gas parameter data include gas pressure, gas content and / or initial velocity of gas release, etc.; the coal body parameter data include coal seam thickness, coal seam roof lithology and / or coal seam inclination, etc.; the tunnel parameter data include tunnel excavation progress, excavation method and / or advance support type, etc.

[0016] S2: performing data cleaning processing on the historical data, and obtaining a data set by dividing the cleaned historical data; In order to clarify the specific method of obtaining the data set, step S2 includes S21 to S23, which are: S21: de-duplicating the historical data, and filling missing values ​​in the deleted historical data to obtain pre-processed data; In this step, the missing value filling process is to delete the features whose historical data missing exceeds a preset threshold ratio after deletion, and fill the features whose historical data missing is less than the threshold ratio through a multiple filling (MI) algorithm, and select the best filling by comparing the filling effects.

[0017] S22: performing category balancing on the preprocessed data using a preset synthetic minority sample algorithm according to characteristic properties of the historical data to obtain category balanced data; In this step, the category balance adopts the SMOETNC algorithm, which is an improved algorithm of SMOTE. The characteristic properties of the historical data are interpolated at the same time to generate new adjacent samples to balance the category balance of the sample labels; preferably, the characteristic properties include continuity and discrete features.

[0018] Preferably, the SMOTE algorithm is an algorithm for synthesizing minority class samples.

[0019] S23: Screening characteristic parameters of the category balance data based on a preset hazard level threshold to obtain a parameter set, and performing data division on the parameter set to obtain a data set.

[0020] In this step, the parameter set is divided into a training set and a validation set according to a ratio of 7:3 to the total sample quantity; the continuous variables are Z-Score standardized, the variables of discrete characteristics are Ordinal encoded, and the risk classification labels are One-Hot encoded; The data of the training set is used to improve the optimization and training of the deep neural network model; the data of the validation set is used to verify the performance of the improved deep neural network model under the optimal parameter combination.

[0021] S3: Based on the preset cubic chaos mapping method, the initial sparrow search algorithm is iteratively trained and optimized with the sine and cosine search strategy to obtain an improved sparrow search algorithm; In this step, the improved sparrow search algorithm is the ISSA algorithm.

[0022] In order to clarify the specific acquisition method of the improved sparrow search algorithm, step S3 includes S31 to S35, which are specifically: S31: Initializing the distribution of the population in the initial sparrow search algorithm based on the cubic chaos mapping method to obtain the population initialization position; In this step, when the population is initialized for distribution, the original random distribution of the population in the initial sparrow search algorithm is replaced by using Cubic chaotic mapping; The initial value of the population iteration is: (1); In the above formula (1), For the The initial value of the iteration; For the The initial value of the iteration; The population initialization position formula is: (2); In the above formula (2), Initialize positions for the population; For the The initial value of the iteration; is the lower position boundary of the population; is the upper position boundary of the population; Preferably, the Cubic chaos mapping is a cubic chaos mapping method. This method improves the uniformity and randomness of population distribution; S32: improving and optimizing the sine-cosine search strategy according to the population initialization position and the preset adaptive nonlinear decreasing inertia weight factor to obtain an improved sine-cosine search strategy; In this step, according to the population initialization position, a preset exponential nonlinear decreasing adaptive factor and the preset adaptive nonlinear decreasing inertia weight factor Improve the sine and cosine search strategy; The expression of the exponential nonlinear decreasing adaptive factor is: (3); In the above formula (3), is the adaptive factor; is the adjustment coefficient; is the current iteration number; is the maximum number of iterations; is a natural constant; The expression of the adaptive nonlinear decreasing inertia weight factor is: (4); In the above formula (4), is the inertia weight factor; is the initial value of the weight; is the final value of the weight; is the current iteration number; is the maximum number of iterations; is the adjustment coefficient; is a natural constant; The expression of the improved sine-cosine search strategy is: (5); In the above formula (5), For the A sparrow in the Wei, The position in the iteration; For the A sparrow in the Wei, The position in the iteration; For the The global optimal position in the iteration; is the adaptive factor; is the inertia weight factor; , , are all random numbers that follow a uniform distribution.

[0023] S33: Optimizing and updating the individual positions of the sparrows in the initial sparrow search algorithm based on the improved sine-cosine search strategy to obtain the optimal position; In order to clarify the specific method of obtaining the optimal position, step S33 includes S331 to S333, which are specifically: S331: updating the position of the discoverer sparrow in the individual sparrow position according to the improved sine-cosine search strategy to obtain an update expression of the discoverer sparrow position; In this step, the expression for updating the position of the discoverer sparrow is: (6); In the above formula (6), For the finder sparrow position A sparrow in the Wei, The position in the iteration; For the A sparrow in the Wei, The position in the iteration; For the The global optimal position in the iteration; is the adaptive factor; is the inertia weight factor; is the warning value; is the safety threshold; , are all random numbers that follow a uniform distribution.

[0024] When updating the position of the discoverer, the improved sine-cosine search strategy is introduced to help enhance the discoverer's exploration ability and avoid premature convergence. At the same time, the exponential nonlinear decreasing adaptive factor is introduced. and the preset adaptive nonlinear decreasing inertia weight factor , so that the improved sine-cosine strategy can reduce the search step size more quickly and accelerate the convergence speed.

[0025] S332: updating the follower position in the individual position of the sparrow according to a preset Levy flight strategy to obtain a follower position update expression; In this step, the expression of Levy flight strategy is: (7); In the above formula (7), is a random step length; To obey Normal distribution, where ; is the absolute value; β is the parameter; The follower position update expression is: (8); In the above formula (8), For followers, position the A sparrow in the Wei, The position in the iteration; For the A sparrow in the Wei, The position in the iteration; For the The global optimal position in the iteration; For the The global worst sparrow position in the iteration; Indicates The best position in the iteration; is a constant; represents the exponential function; is the number of individuals in the population; is a random step length; Preferably, the Levy flight strategy is a Levy flight strategy.

[0026] When updating the follower position, the Levy flight strategy is introduced to avoid the sparrow population from being trapped in a local optimal solution due to excessive aggregation.

[0027] S333: Update the individual positions of the sparrows based on the discoverer sparrow position update expression and the follower position update expression to obtain the optimal position.

[0028] S34: Based on the optimal position, a sparrow individual solution is obtained by performing lens imaging reverse learning strategy calculation on a preset adaptive nonlinear increase factor and the sparrow individual in the initial sparrow search algorithm; In this step, based on the optimal position, an adaptive nonlinear increase factor is introduced through an adaptive reverse learning strategy based on lens imaging. , The adaptive nonlinear increase factor expression is: (9); In the above formula (9), is the adaptive nonlinear increase factor; is the adjustment coefficient; is the current iteration number; is the maximum number of iterations; Then the lens reverse learning formula becomes: (10); In the above formula (10), is the reverse solution for individual sparrows; For a particular sparrow; is the adaptive nonlinear increase factor; is the lower position boundary of the population; is the upper position boundary of the population; The reverse solutions (M in total) of the sparrow individual are combined with the forward solutions (M in total) of the sparrow individual (2M in total), and the best first M individuals are selected by fitness sorting, and the next round of iteration is performed to obtain the sparrow individual solution.

[0029] Finally, an adaptive reverse learning strategy based on lens imaging is introduced to rank the fitness of each sparrow individual and its reverse solution, select high-quality individual solutions, and accelerate the convergence speed of the algorithm and the accuracy of optimization.

[0030] S35: judging whether the current number of iterations meets a preset maximum number of iterations, and when the preset maximum number of iterations is met, outputting the sparrow individual solution and the optimal position to obtain an improved sparrow search algorithm.

[0031] In this step, when the preset maximum number of iterations is met, the training is completed by outputting the sparrow individual solution, the optimal position and the preset fitness curve, and an improved sparrow search algorithm is obtained.

[0032] S4: Optimizing hyperparameters of a preset deep neural network model according to the improved sparrow search algorithm to obtain an improved deep neural network model; In order to clarify the specific method of obtaining the improved deep neural network model, step S4 includes S41 to S44, which are: S41: according to the boundary conditions of the improved sparrow search algorithm, input the training set in the data set into the improved sparrow search algorithm for training to obtain an optimal solution, wherein the optimal solution includes the number of neurons in the hidden layer and the batch size; S42: Optimizing the deep neural network according to a preset K-fold cross-validation method to obtain a deep neural network model; In order to clarify the specific method of obtaining the deep neural network model, step S42 includes S421 to S425, which are specifically: S421: Acquisition strategy dynamically adjusts learning rate; S422: Divide the training set in the data set according to the K-fold cross-validation method to obtain multiple subsets; In this step, the K-fold cross validation method is a K-fold cross validation method, which divides the training set in the data set into K subsets.

[0033] S423: dynamically adjusting the learning rate and the preset activation function according to the strategy to optimize the deep neural network to obtain an optimized deep neural network; In this step, the deep neural network is optimized by dynamically adjusting the learning rate ReduceLROnPlateau, the activation function ReLu, the optimizer algorithm, AdamOptimizer, the loss function Categorical_crossentropy, the evaluation criteria, setting the regularization means DropOut rate and the maximum number of iterations Epoch and the early stopping strategy to obtain an optimized deep neural network; the early stopping strategy is used to prevent overfitting in training.

[0034] S424: performing multiple training and validation on the plurality of subsets according to the K-fold cross validation method, wherein in each training, one subset is used as a validation set and the other subsets are used as training sets, to obtain an average accuracy; In this step, the multiple subsets are trained and validated K times according to the K-fold cross-validation method. In each training, one subset is used as the validation set and the remaining K-1 subsets are used as the training set to obtain the average Accuracy of the K-times validation; the average Accuracy of the K-times validation is the average precision.

[0035] S425: Input the training set into the optimized deep neural network, input the number of neurons in the hidden layer and the batch size into the optimized deep neural network for optimization, output the average accuracy, and obtain a deep neural network model.

[0036] In this step, the number of neurons in the hidden layer is N1, N2, N3 and the batch size is BatchSize.

[0037] S43: Inputting the optimal solution into the deep neural network model to perform hyperparameter optimization to obtain an optimal hyperparameter combination set; S44: Inputting the optimal hyperparameter combination set into the deep neural network model, and obtaining an improved deep neural network model by training the training set.

[0038] In this step, the deep neural network model is a DNN deep neural network model, and the improved deep neural network model is an ISSA-DNN model. The deep neural network model is optimized by the improved sparrow search algorithm to simulate the foraging and anti-predation behavior of sparrows. When the improved sparrow search algorithm optimizes the hyperparameters of the deep neural network model, it can perform a more efficient search in the hyperparameter space and quickly and accurately find a set of better hyperparameter combinations.

[0039] S5: inputting the data set into the improved deep neural network model for iterative training, and obtaining a prediction model by screening feature parameters of the data set; In order to clarify the specific method of obtaining the prediction model, step S5 includes S51 to S55, which are specifically: S51: inputting the validation set in the data set into the improved deep neural network model for training, and obtaining the original evaluation index by comparing the predicted label of the validation set with the real label of the validation set; S52: Based on the method for predicting the result of a preset machine learning model, the improved deep neural network model is calculated by the characteristic parameters of the data set to obtain the characteristic parameter contribution; In this step, the method of predicting the results of the machine learning model is the SHAP method.

[0040] S53: sorting the importance of the feature parameters according to the contribution of the feature parameters to obtain a ranking result; In this step, the ranking result is a SHAP feature importance ranking.

[0041] S54: Based on the ranking results, the number of features at the bottom of the ranking results is reduced in sequence, and the data set features at the bottom of the ranking results are deleted. When the number of features is reduced, the improved deep neural network model is retrained and predicted to obtain multiple model evaluation indicators; In this step, the number of features with lower rankings is reduced in turn by combining the traditional recursive feature selection method (RFE method). This step uses an intuitive and more explanatory feature selection strategy to analyze the role of each feature in model prediction. While reducing the number of features, the prediction performance of the model can still be maintained to avoid the "curse of dimensionality".

[0042] S55: Compare the original evaluation index with the plurality of model evaluation indexes, select the optimal model based on the comparison result, and obtain a prediction model.

[0043] S6: Predict the preset tunnel coal and gas monitoring data based on the prediction model to obtain a hazard classification result.

[0044] Embodiment 2: This embodiment provides a tunnel coal and gas hazard classification prediction device, the device comprising: An acquisition model is used to acquire historical data, wherein the historical data includes gas parameter data, coal body parameter data, tunnel parameter data and hazard classification labels; A cleaning model is used to perform data cleaning on the historical data, and obtain a data set by dividing the cleaned historical data; To clarify the specific methods of obtaining the cleaning model, there are: A processing unit, used for deduplicating the historical data and filling missing values ​​in the deleted historical data to obtain preprocessed data; A balancing unit, configured to perform category balancing on the preprocessed data by using a preset synthetic minority class sample algorithm according to characteristic properties of the historical data to obtain category balanced data; The division unit is used to screen the characteristic parameters of the category balance data based on a preset danger level threshold to obtain a parameter set, and to obtain a data set by performing data division on the parameter set.

[0045] The first optimization model is used to perform iterative training optimization of the sine and cosine search strategy on the initial sparrow search algorithm based on a preset cubic chaos mapping method to obtain an improved sparrow search algorithm; To clarify the specific method of obtaining the first optimization model, the specific methods are: An initialization module, used for initializing the distribution of the population in the initial sparrow search algorithm based on the cubic chaos mapping method to obtain the population initialization position; An improved optimization module, used for improving and optimizing the sine-cosine search strategy according to the population initialization position and a preset adaptive nonlinear decreasing inertia weight factor to obtain an improved sine-cosine search strategy; An optimization and updating module, used for optimizing and updating the individual positions of sparrows in the initial sparrow search algorithm based on the improved sine-cosine search strategy to obtain the optimal position; A calculation module, for obtaining a sparrow individual solution based on the optimal position by performing lens imaging reverse learning strategy calculation on a preset adaptive nonlinear increase factor and the sparrow individual in the initial sparrow search algorithm; The training module is used to judge whether the current number of iterations meets the preset maximum number of iterations. When the preset maximum number of iterations is met, the improved sparrow search algorithm is obtained by outputting the sparrow individual solution and the optimal position.

[0046] A second optimization model is used to optimize the hyperparameters of a preset deep neural network model according to the improved sparrow search algorithm to obtain an improved deep neural network model; To clarify the specific method of obtaining the second optimization model, the specific methods are: A second training unit is used for inputting the training set in the data set into the improved sparrow search algorithm for training according to the boundary conditions of the improved sparrow search algorithm to obtain an optimal solution, wherein the optimal solution includes the number of hidden layer neurons and the batch size; A first optimization unit is used to optimize the deep neural network according to a preset K-fold cross-validation method to obtain a deep neural network model; A second optimization unit is used to input the optimal solution into the deep neural network model to perform hyperparameter optimization to obtain an optimal hyperparameter combination set; The third training unit is used to input the optimal hyperparameter combination set into the deep neural network model, and obtain an improved deep neural network model by training the training set.

[0047] A training model is used to input the data set into the improved deep neural network model for iterative training, and obtain a prediction model by screening feature parameters of the data set; To clarify the specific methods of obtaining the training model, there are: A first training unit is used to input a verification set in the data set into the improved deep neural network model for training, and obtain an original evaluation index by comparing the predicted label of the verification set with the real label of the verification set; A calculation unit, for calculating the improved deep neural network model based on the method for predicting the result of a preset machine learning model, by using the characteristic parameters of the data set to obtain the characteristic parameter contribution; A ranking unit, used to rank the importance of feature parameters according to the contribution of the feature parameters to obtain a ranking result; A screening unit is used to reduce the number of features at the bottom of the ranking results in turn based on the ranking results, and delete the data set features at the bottom of the ranking results. When the number of features is reduced, the improved deep neural network model is retrained and predicted to obtain multiple model evaluation indicators; The selection unit is used to compare the original evaluation index with the multiple model evaluation indexes, select the optimal model based on the comparison results, and obtain the prediction model.

[0048] The prediction model is used to predict the preset tunnel coal and gas monitoring data based on the prediction model to obtain a hazard classification result.

[0049] It should be noted that, regarding the device in the above embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment of the method, and will not be elaborated here.

[0050] Embodiment 3: Corresponding to the above method embodiment, a tunnel coal and gas hazard grading prediction device is also provided in this embodiment. The tunnel coal and gas hazard grading prediction device described below and the tunnel coal and gas hazard grading prediction method described above can be referenced to each other.

[0051] Figure 2 FIG. 8 is a block diagram of a tunnel coal and gas hazard classification prediction device 800 according to an exemplary embodiment. Figure 2 As shown, the tunnel coal and gas hazard classification prediction device 800 may include: a processor 801 and a memory 802. The tunnel coal and gas hazard classification prediction device 800 may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0052] The processor 801 is used to control the overall operation of the tunnel coal and gas hazard classification prediction device 800 to complete all or part of the steps in the above-mentioned tunnel coal and gas hazard classification prediction method. The memory 802 is used to store various types of data to support the operation of the tunnel coal and gas hazard classification prediction device 800, and these data may include, for example, instructions for any application or method operating on the tunnel coal and gas hazard classification prediction device 800, and application-related data, such as contact data, sent and received messages, pictures, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 802 or sent via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules can be keyboards, mice, buttons, etc. These buttons can be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the tunnel coal and gas hazard classification prediction device 800 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 805 can include: Wi-Fi module, Bluetooth module, NFC module.

[0053] In an exemplary embodiment, the tunnel coal and gas hazard classification prediction device 800 can be implemented by one or more application specific integrated circuits (Application Specific Integrated Circuit, referred to as ASIC), digital signal processors (Digital Signal Processor, referred to as DSP), digital signal processing devices (Digital Signal Processing Device, referred to as DSPD), programmable logic devices (Programmable Logic Device, referred to as PLD), field programmable gate arrays (Field Programmable Gate Array, referred to as FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned tunnel coal and gas hazard classification prediction method.

[0054] Embodiment 4: Corresponding to the above method embodiment, a medium is also provided in this embodiment. The medium described below and the tunnel coal and gas hazard classification prediction method described above can refer to each other.

[0055] A medium stores a computer program, which, when executed by a processor, implements the steps of the tunnel coal and gas hazard classification prediction method of the above method embodiment.

[0056] The medium may specifically be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or other medium that can store program codes.

[0057] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0058] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for predicting the hazard classification of coal and gas in a tunnel, characterized in that: include: Acquiring historical data, wherein the historical data includes gas parameter data, coal body parameter data, tunnel parameter data and hazard classification labels; Performing data cleaning on the historical data, and obtaining a data set by dividing the cleaned historical data; Based on the preset cubic chaos mapping method, the initial sparrow search algorithm is iteratively trained and optimized with sine and cosine search strategies to obtain an improved sparrow search algorithm. According to the improved sparrow search algorithm, hyperparameter optimization is performed on the preset deep neural network model to obtain an improved deep neural network model; Inputting the data set into the improved deep neural network model for iterative training, and obtaining a prediction model by screening feature parameters of the data set; Based on the prediction model, the preset tunnel coal and gas monitoring data are predicted to obtain the hazard classification result.

2. The tunnel coal and gas hazard classification prediction method according to claim 1 is characterized in that: Based on the preset cubic chaos mapping method, the initial sparrow search algorithm is iteratively trained and optimized with sine and cosine search strategies to obtain an improved sparrow search algorithm, including: Initializing the distribution of the population in the initial sparrow search algorithm based on the cubic chaos mapping method to obtain the population initialization position; The sine-cosine search strategy is improved and optimized according to the population initialization position and the preset adaptive nonlinear decreasing inertia weight factor to obtain an improved sine-cosine search strategy; Based on the improved sine-cosine search strategy, the individual positions of the sparrows in the initial sparrow search algorithm are optimized and updated to obtain the optimal position; Based on the optimal position, a sparrow individual solution is obtained by calculating the lens imaging reverse learning strategy for the preset adaptive nonlinear increase factor and the sparrow individual in the initial sparrow search algorithm; It is judged whether the current number of iterations meets the preset maximum number of iterations. When the preset maximum number of iterations is met, the sparrow individual solution and the optimal position are output to obtain an improved sparrow search algorithm.

3. The tunnel coal and gas hazard classification prediction method according to claim 1 is characterized in that: The data set is input into the improved deep neural network model for iterative training, and a prediction model is obtained by screening feature parameters of the data set, including: Inputting the validation set in the data set into the improved deep neural network model for training, and obtaining the original evaluation index by comparing the predicted labels of the validation set with the real labels of the validation set; Based on the method of predicting the results of a preset machine learning model, the improved deep neural network model is calculated through the characteristic parameters of the data set to obtain the characteristic parameter contribution; The importance of the feature parameters is sorted according to the contribution of the feature parameters to obtain a ranking result; Based on the ranking results, the number of features at the bottom of the ranking results is reduced in sequence, and the features of the data set at the bottom of the ranking results are deleted. When the number of features is reduced, the improved deep neural network model is retrained and predicted to obtain multiple model evaluation indicators; The original evaluation index is compared with the plurality of model evaluation indexes, and the optimal model is selected based on the comparison results to obtain a prediction model.

4. The tunnel coal and gas hazard classification prediction method according to claim 1 is characterized in that: The historical data is cleaned and divided into data sets, including: Deleting duplicate data from the historical data, and filling missing values ​​from the deleted historical data to obtain preprocessed data; According to the characteristic properties of the historical data, class balancing is performed on the preprocessed data by using a preset synthetic minority sample algorithm to obtain class balanced data; The characteristic parameters of the category balance data are screened based on a preset hazard level threshold to obtain a parameter set, and the data set is obtained by dividing the parameter set.

5. The tunnel coal and gas hazard classification prediction method according to claim 1 is characterized in that: The preset deep neural network model is optimized for hyperparameters according to the improved sparrow search algorithm to obtain an improved deep neural network model, including: According to the boundary conditions of the improved sparrow search algorithm, the training set in the data set is input into the improved sparrow search algorithm for training to obtain an optimal solution, wherein the optimal solution includes the number of hidden layer neurons and the batch size; Optimizing the deep neural network according to a preset K-fold cross-validation method to obtain a deep neural network model; Inputting the optimal solution into the deep neural network model to perform hyperparameter optimization to obtain an optimal hyperparameter combination set; The optimal hyperparameter combination set is input into the deep neural network model, and an improved deep neural network model is obtained by training the training set.

6. A tunnel coal and gas hazard classification prediction device, characterized in that: include: An acquisition model is used to acquire historical data, wherein the historical data includes gas parameter data, coal body parameter data, tunnel parameter data and hazard classification labels; A cleaning model is used to perform data cleaning on the historical data, and obtain a data set by dividing the cleaned historical data; The first optimization model is used to perform iterative training optimization of the sine and cosine search strategy on the initial sparrow search algorithm based on a preset cubic chaos mapping method to obtain an improved sparrow search algorithm; A second optimization model is used to optimize the hyperparameters of a preset deep neural network model according to the improved sparrow search algorithm to obtain an improved deep neural network model; A training model is used to input the data set into the improved deep neural network model for iterative training, and obtain a prediction model by screening feature parameters of the data set; The prediction model is used to predict the preset tunnel coal and gas monitoring data based on the prediction model to obtain a hazard classification result.

7. The tunnel coal and gas hazard classification prediction device according to claim 6 is characterized in that: The first optimization model includes: An initialization module, used for initializing the distribution of the population in the initial sparrow search algorithm based on the cubic chaos mapping method to obtain the population initialization position; An improved optimization module, used for improving and optimizing the sine-cosine search strategy according to the population initialization position and a preset adaptive nonlinear decreasing inertia weight factor to obtain an improved sine-cosine search strategy; An optimization and updating module, used for optimizing and updating the individual positions of sparrows in the initial sparrow search algorithm based on the improved sine-cosine search strategy to obtain the optimal position; A calculation module, for obtaining a sparrow individual solution based on the optimal position by performing lens imaging reverse learning strategy calculation on a preset adaptive nonlinear increase factor and the sparrow individual in the initial sparrow search algorithm; The training module is used to judge whether the current number of iterations meets the preset maximum number of iterations. When the preset maximum number of iterations is met, the improved sparrow search algorithm is obtained by outputting the sparrow individual solution and the optimal position.

8. The tunnel coal and gas hazard classification prediction device according to claim 6 is characterized in that: Training the model, including: A first training unit is used to input a verification set in the data set into the improved deep neural network model for training, and obtain an original evaluation index by comparing the predicted label of the verification set with the real label of the verification set; A calculation unit, for calculating the improved deep neural network model based on the method for predicting the result of a preset machine learning model, by using the characteristic parameters of the data set to obtain the characteristic parameter contribution; A ranking unit, used to rank the importance of feature parameters according to the contribution of the feature parameters to obtain a ranking result; A screening unit is used to reduce the number of features at the bottom of the ranking results in turn based on the ranking results, and delete the data set features at the bottom of the ranking results. When the number of features is reduced, the improved deep neural network model is retrained and predicted to obtain multiple model evaluation indicators; The selection unit is used to compare the original evaluation index with the multiple model evaluation indexes, select the optimal model based on the comparison results, and obtain the prediction model.

9. The tunnel coal and gas hazard classification prediction device according to claim 6, characterized in that: Cleaning model, including: A processing unit, used for deduplicating the historical data and filling missing values ​​in the deleted historical data to obtain preprocessed data; A balancing unit, configured to perform category balancing on the preprocessed data by using a preset synthetic minority class sample algorithm according to characteristic properties of the historical data to obtain category balanced data; The division unit is used to screen the characteristic parameters of the category balance data based on a preset danger level threshold to obtain a parameter set, and to obtain a data set by performing data division on the parameter set.

10. The tunnel coal and gas hazard classification prediction device according to claim 6, characterized in that: The second optimization model includes: A second training unit is used for inputting the training set in the data set into the improved sparrow search algorithm for training according to the boundary conditions of the improved sparrow search algorithm to obtain an optimal solution, wherein the optimal solution includes the number of hidden layer neurons and the batch size; A first optimization unit is used to optimize the deep neural network according to a preset K-fold cross-validation method to obtain a deep neural network model; A second optimization unit is used to input the optimal solution into the deep neural network model to perform hyperparameter optimization to obtain an optimal hyperparameter combination set; The third training unit is used to input the optimal hyperparameter combination set into the deep neural network model, and obtain an improved deep neural network model by training the training set.

Images